diff --git a/.gitignore b/.gitignore index 22aba65364..29374d13dc 100644 --- a/.gitignore +++ b/.gitignore @@ -153,6 +153,9 @@ docker-legacy/volumes/etcd/* docker-legacy/volumes/minio/* docker-legacy/volumes/milvus/* docker-legacy/volumes/chroma/* +docker-legacy/volumes/opensearch/data/* +docker-legacy/volumes/pgvectors/data/* +docker-legacy/volumes/pgvector/data/* docker/volumes/app/storage/* docker/volumes/certbot/* @@ -164,6 +167,12 @@ docker/volumes/etcd/* docker/volumes/minio/* docker/volumes/milvus/* docker/volumes/chroma/* +docker/volumes/opensearch/data/* +docker/volumes/myscale/data/* +docker/volumes/myscale/log/* +docker/volumes/unstructured/* +docker/volumes/pgvector/data/* +docker/volumes/pgvecto_rs/data/* docker/nginx/conf.d/default.conf docker/middleware.env diff --git a/CONTRIBUTING_CN.md b/CONTRIBUTING_CN.md index 7cd2bb60eb..310c55090a 100644 --- a/CONTRIBUTING_CN.md +++ b/CONTRIBUTING_CN.md @@ -36,7 +36,7 @@ | 被团队成员标记为高优先级的功能 | 高优先级 | | 在 [community feedback board](https://github.com/langgenius/dify/discussions/categories/feedbacks) 内反馈的常见功能请求 | 中等优先级 | | 非核心功能和小幅改进 | 低优先级 | - | 有价值当不紧急 | 未来功能 | + | 有价值但不紧急 | 未来功能 | ### 其他任何事情（例如 bug 报告、性能优化、拼写错误更正）： * 立即开始编码。 @@ -138,7 +138,7 @@ Dify 的后端使用 Python 编写，使用 [Flask](https://flask.palletsproject ├── models // 描述数据模型和 API 响应的形状 ├── public // 如 favicon 等元资源 ├── service // 定义 API 操作的形状 -├── test +├── test ├── types // 函数参数和返回值的描述 └── utils // 共享的实用函数 ``` diff --git a/api/core/embedding/cached_embedding.py b/api/core/embedding/cached_embedding.py index 4cc793b0d7..8ce12fd59f 100644 --- a/api/core/embedding/cached_embedding.py +++ b/api/core/embedding/cached_embedding.py @@ -65,7 +65,7 @@ class CacheEmbedding(Embeddings): except IntegrityError: db.session.rollback() except Exception as e: - logging.exception("Failed transform embedding: ", e) + logging.exception("Failed transform embedding: %s", e) cache_embeddings = [] try: for i, embedding in zip(embedding_queue_indices, embedding_queue_embeddings): @@ -85,7 +85,7 @@ class CacheEmbedding(Embeddings): db.session.rollback() except Exception as ex: db.session.rollback() - logger.error("Failed to embed documents: ", ex) + logger.error("Failed to embed documents: %s", ex) raise ex return text_embeddings @@ -116,10 +116,7 @@ class CacheEmbedding(Embeddings): # Transform to string encoded_str = encoded_vector.decode("utf-8") redis_client.setex(embedding_cache_key, 600, encoded_str) - - except IntegrityError: - db.session.rollback() - except: - logging.exception("Failed to add embedding to redis") + except Exception as ex: + logging.exception("Failed to add embedding to redis %s", ex) return embedding_results diff --git a/api/core/model_runtime/model_providers/bedrock/llm/eu.anthropic.claude-3-haiku-v1.yaml b/api/core/model_runtime/model_providers/bedrock/llm/eu.anthropic.claude-3-haiku-v1.yaml index fe5f54de13..24a65ef1bb 100644 --- a/api/core/model_runtime/model_providers/bedrock/llm/eu.anthropic.claude-3-haiku-v1.yaml +++ b/api/core/model_runtime/model_providers/bedrock/llm/eu.anthropic.claude-3-haiku-v1.yaml @@ -1,6 +1,6 @@ model: eu.anthropic.claude-3-haiku-20240307-v1:0 label: - en_US: Claude 3 Haiku(Cross Region Inference) + en_US: Claude 3 Haiku(EU.Cross Region Inference) model_type: llm features: - agent-thought diff --git a/api/core/model_runtime/model_providers/bedrock/llm/eu.anthropic.claude-3-sonnet-v1.5.yaml b/api/core/model_runtime/model_providers/bedrock/llm/eu.anthropic.claude-3-sonnet-v1.5.yaml index 9f8d029a57..e3d25c7d8f 100644 --- a/api/core/model_runtime/model_providers/bedrock/llm/eu.anthropic.claude-3-sonnet-v1.5.yaml +++ b/api/core/model_runtime/model_providers/bedrock/llm/eu.anthropic.claude-3-sonnet-v1.5.yaml @@ -1,6 +1,6 @@ model: eu.anthropic.claude-3-5-sonnet-20240620-v1:0 label: - en_US: Claude 3.5 Sonnet(Cross Region Inference) + en_US: Claude 3.5 Sonnet(EU.Cross Region Inference) model_type: llm features: - agent-thought diff --git a/api/core/model_runtime/model_providers/bedrock/llm/eu.anthropic.claude-3-sonnet-v1.yaml b/api/core/model_runtime/model_providers/bedrock/llm/eu.anthropic.claude-3-sonnet-v1.yaml index bfaf5abb8e..9a06a4ad6d 100644 --- a/api/core/model_runtime/model_providers/bedrock/llm/eu.anthropic.claude-3-sonnet-v1.yaml +++ b/api/core/model_runtime/model_providers/bedrock/llm/eu.anthropic.claude-3-sonnet-v1.yaml @@ -1,6 +1,6 @@ model: eu.anthropic.claude-3-sonnet-20240229-v1:0 label: - en_US: Claude 3 Sonnet(Cross Region Inference) + en_US: Claude 3 Sonnet(EU.Cross Region Inference) model_type: llm features: - agent-thought diff --git a/api/core/model_runtime/model_providers/bedrock/llm/llm.py b/api/core/model_runtime/model_providers/bedrock/llm/llm.py index 06a8606901..77bab0c294 100644 --- a/api/core/model_runtime/model_providers/bedrock/llm/llm.py +++ b/api/core/model_runtime/model_providers/bedrock/llm/llm.py @@ -1,8 +1,8 @@ # standard import import base64 -import io import json import logging +import mimetypes from collections.abc import Generator from typing import Optional, Union, cast @@ -17,7 +17,6 @@ from botocore.exceptions import ( ServiceNotInRegionError, UnknownServiceError, ) -from PIL.Image import Image # local import from core.model_runtime.callbacks.base_callback import Callback @@ -443,8 +442,9 @@ class BedrockLargeLanguageModel(LargeLanguageModel): try: url = message_content.data image_content = requests.get(url).content - with Image.open(io.BytesIO(image_content)) as img: - mime_type = f"image/{img.format.lower()}" + if "?" in url: + url = url.split("?")[0] + mime_type, _ = mimetypes.guess_type(url) base64_data = base64.b64encode(image_content).decode("utf-8") except Exception as ex: raise ValueError(f"Failed to fetch image data from url {message_content.data}, {ex}") diff --git a/api/core/model_runtime/model_providers/bedrock/llm/us.anthropic.claude-3-haiku-v1.yaml b/api/core/model_runtime/model_providers/bedrock/llm/us.anthropic.claude-3-haiku-v1.yaml index 58c1f05779..9247f46974 100644 --- a/api/core/model_runtime/model_providers/bedrock/llm/us.anthropic.claude-3-haiku-v1.yaml +++ b/api/core/model_runtime/model_providers/bedrock/llm/us.anthropic.claude-3-haiku-v1.yaml @@ -1,6 +1,6 @@ model: us.anthropic.claude-3-haiku-20240307-v1:0 label: - en_US: Claude 3 Haiku(Cross Region Inference) + en_US: Claude 3 Haiku(US.Cross Region Inference) model_type: llm features: - agent-thought diff --git a/api/core/model_runtime/model_providers/bedrock/llm/us.anthropic.claude-3-opus-v1.yaml b/api/core/model_runtime/model_providers/bedrock/llm/us.anthropic.claude-3-opus-v1.yaml index 6b9e1ec067..f9854d51f0 100644 --- a/api/core/model_runtime/model_providers/bedrock/llm/us.anthropic.claude-3-opus-v1.yaml +++ b/api/core/model_runtime/model_providers/bedrock/llm/us.anthropic.claude-3-opus-v1.yaml @@ -1,6 +1,6 @@ model: us.anthropic.claude-3-opus-20240229-v1:0 label: - en_US: Claude 3 Opus(Cross Region Inference) + en_US: Claude 3 Opus(US.Cross Region Inference) model_type: llm features: - agent-thought diff --git a/api/core/model_runtime/model_providers/bedrock/llm/us.anthropic.claude-3-sonnet-v1.5.yaml b/api/core/model_runtime/model_providers/bedrock/llm/us.anthropic.claude-3-sonnet-v1.5.yaml index f1e0d6c5a2..fbcab2d5f3 100644 --- a/api/core/model_runtime/model_providers/bedrock/llm/us.anthropic.claude-3-sonnet-v1.5.yaml +++ b/api/core/model_runtime/model_providers/bedrock/llm/us.anthropic.claude-3-sonnet-v1.5.yaml @@ -1,6 +1,6 @@ model: us.anthropic.claude-3-5-sonnet-20240620-v1:0 label: - en_US: Claude 3.5 Sonnet(Cross Region Inference) + en_US: Claude 3.5 Sonnet(US.Cross Region Inference) model_type: llm features: - agent-thought diff --git a/api/core/model_runtime/model_providers/bedrock/llm/us.anthropic.claude-3-sonnet-v1.yaml b/api/core/model_runtime/model_providers/bedrock/llm/us.anthropic.claude-3-sonnet-v1.yaml index dce50bf4b5..9f5a1501f0 100644 --- a/api/core/model_runtime/model_providers/bedrock/llm/us.anthropic.claude-3-sonnet-v1.yaml +++ b/api/core/model_runtime/model_providers/bedrock/llm/us.anthropic.claude-3-sonnet-v1.yaml @@ -1,6 +1,6 @@ model: us.anthropic.claude-3-sonnet-20240229-v1:0 label: - en_US: Claude 3 Sonnet(Cross Region Inference) + en_US: Claude 3 Sonnet(US.Cross Region Inference) model_type: llm features: - agent-thought diff --git a/api/core/model_runtime/model_providers/jina/jina.py b/api/core/model_runtime/model_providers/jina/jina.py index 33977b6a33..186a0a0fa7 100644 --- a/api/core/model_runtime/model_providers/jina/jina.py +++ b/api/core/model_runtime/model_providers/jina/jina.py @@ -18,9 +18,9 @@ class JinaProvider(ModelProvider): try: model_instance = self.get_model_instance(ModelType.TEXT_EMBEDDING) - # Use `jina-embeddings-v2-base-en` model for validate, + # Use `jina-embeddings-v3` model for validate, # no matter what model you pass in, text completion model or chat model - model_instance.validate_credentials(model="jina-embeddings-v2-base-en", credentials=credentials) + model_instance.validate_credentials(model="jina-embeddings-v3", credentials=credentials) except CredentialsValidateFailedError as ex: raise ex except Exception as ex: diff --git a/api/core/model_runtime/model_providers/jina/text_embedding/jina-embeddings-v3.yaml b/api/core/model_runtime/model_providers/jina/text_embedding/jina-embeddings-v3.yaml new file mode 100644 index 0000000000..4e5374dc9d --- /dev/null +++ b/api/core/model_runtime/model_providers/jina/text_embedding/jina-embeddings-v3.yaml @@ -0,0 +1,9 @@ +model: jina-embeddings-v3 +model_type: text-embedding +model_properties: + context_size: 8192 + max_chunks: 2048 +pricing: + input: '0.001' + unit: '0.001' + currency: USD diff --git a/api/core/model_runtime/model_providers/jina/text_embedding/text_embedding.py b/api/core/model_runtime/model_providers/jina/text_embedding/text_embedding.py index ef12e534db..ceb79567d5 100644 --- a/api/core/model_runtime/model_providers/jina/text_embedding/text_embedding.py +++ b/api/core/model_runtime/model_providers/jina/text_embedding/text_embedding.py @@ -56,6 +56,9 @@ class JinaTextEmbeddingModel(TextEmbeddingModel): data = {"model": model, "input": [transform_jina_input_text(model, text) for text in texts]} + if model == "jina-embeddings-v3": + data["task"] = "text-matching" + try: response = post(url, headers=headers, data=dumps(data)) except Exception as e: diff --git a/api/core/model_runtime/model_providers/openai/llm/gpt-3.5-turbo-0613.yaml b/api/core/model_runtime/model_providers/openai/llm/gpt-3.5-turbo-0613.yaml index 31dc53e89f..a1ad07d712 100644 --- a/api/core/model_runtime/model_providers/openai/llm/gpt-3.5-turbo-0613.yaml +++ b/api/core/model_runtime/model_providers/openai/llm/gpt-3.5-turbo-0613.yaml @@ -31,3 +31,4 @@ pricing: output: '0.002' unit: '0.001' currency: USD +deprecated: true diff --git a/api/core/model_runtime/model_providers/openai/llm/gpt-3.5-turbo-16k-0613.yaml b/api/core/model_runtime/model_providers/openai/llm/gpt-3.5-turbo-16k-0613.yaml index 4a0e2ef191..4e30279284 100644 --- a/api/core/model_runtime/model_providers/openai/llm/gpt-3.5-turbo-16k-0613.yaml +++ b/api/core/model_runtime/model_providers/openai/llm/gpt-3.5-turbo-16k-0613.yaml @@ -31,3 +31,4 @@ pricing: output: '0.004' unit: '0.001' currency: USD +deprecated: true diff --git a/api/core/model_runtime/model_providers/openai/llm/o1-mini-2024-09-12.yaml b/api/core/model_runtime/model_providers/openai/llm/o1-mini-2024-09-12.yaml index 07a3bc9a7a..0ade7f8ded 100644 --- a/api/core/model_runtime/model_providers/openai/llm/o1-mini-2024-09-12.yaml +++ b/api/core/model_runtime/model_providers/openai/llm/o1-mini-2024-09-12.yaml @@ -11,9 +11,9 @@ model_properties: parameter_rules: - name: max_tokens use_template: max_tokens - default: 65563 + default: 65536 min: 1 - max: 65563 + max: 65536 - name: response_format label: zh_Hans: 回复格式 diff --git a/api/core/model_runtime/model_providers/openai/llm/o1-mini.yaml b/api/core/model_runtime/model_providers/openai/llm/o1-mini.yaml index 3e83529201..60816c5d1e 100644 --- a/api/core/model_runtime/model_providers/openai/llm/o1-mini.yaml +++ b/api/core/model_runtime/model_providers/openai/llm/o1-mini.yaml @@ -11,9 +11,9 @@ model_properties: parameter_rules: - name: max_tokens use_template: max_tokens - default: 65563 + default: 65536 min: 1 - max: 65563 + max: 65536 - name: response_format label: zh_Hans: 回复格式 diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen-max-0107.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen-max-0107.yaml new file mode 100644 index 0000000000..7c90afecf5 --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen-max-0107.yaml @@ -0,0 +1,81 @@ +model: qwen-max-0107 +label: + en_US: qwen-max-0107 +model_type: llm +features: + - multi-tool-call + - agent-thought + - stream-tool-call +model_properties: + mode: chat + context_size: 8192 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 2000 + min: 1 + max: 2000 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.04' + output: '0.12' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen-max-1201.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen-max-1201.yaml index 0368a4a01e..dc234783cd 100644 --- a/api/core/model_runtime/model_providers/tongyi/llm/qwen-max-1201.yaml +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen-max-1201.yaml @@ -79,3 +79,4 @@ pricing: output: '0.12' unit: '0.001' currency: RMB +deprecated: true diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-0206.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-0206.yaml new file mode 100644 index 0000000000..7940be9e8b --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-0206.yaml @@ -0,0 +1,79 @@ +model: qwen-plus-0206 +label: + en_US: qwen-plus-0206 +model_type: llm +features: + - agent-thought +model_properties: + mode: completion + context_size: 32768 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 2000 + min: 1 + max: 2000 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.004' + output: '0.012' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-0624.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-0624.yaml new file mode 100644 index 0000000000..0e02526beb --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-0624.yaml @@ -0,0 +1,79 @@ +model: qwen-plus-0624 +label: + en_US: qwen-plus-0624 +model_type: llm +features: + - agent-thought +model_properties: + mode: completion + context_size: 32768 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 2000 + min: 1 + max: 2000 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.004' + output: '0.012' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-0723.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-0723.yaml new file mode 100644 index 0000000000..65175f1b10 --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-0723.yaml @@ -0,0 +1,79 @@ +model: qwen-plus-0806 +label: + en_US: qwen-plus-0806 +model_type: llm +features: + - agent-thought +model_properties: + mode: completion + context_size: 32768 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 2000 + min: 1 + max: 2000 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.004' + output: '0.012' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-0806.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-0806.yaml new file mode 100644 index 0000000000..1c530dcba2 --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen-plus-0806.yaml @@ -0,0 +1,79 @@ +model: qwen-plus-0806 +label: + en_US: qwen-plus-0806 +model_type: llm +features: + - agent-thought +model_properties: + mode: completion + context_size: 131072 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 2000 + min: 1 + max: 2000 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.004' + output: '0.012' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen-plus.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen-plus.yaml index 4be78627f0..e78b77c7f2 100644 --- a/api/core/model_runtime/model_providers/tongyi/llm/qwen-plus.yaml +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen-plus.yaml @@ -6,7 +6,7 @@ features: - agent-thought model_properties: mode: completion - context_size: 32768 + context_size: 131072 parameter_rules: - name: temperature use_template: temperature diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen-turbo-0206.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen-turbo-0206.yaml new file mode 100644 index 0000000000..2c9857cf9f --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen-turbo-0206.yaml @@ -0,0 +1,79 @@ +model: qwen-turbo-0206 +label: + en_US: qwen-turbo-0206 +model_type: llm +features: + - agent-thought +model_properties: + mode: completion + context_size: 8192 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 1500 + min: 1 + max: 1500 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.002' + output: '0.006' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen-turbo-0624.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen-turbo-0624.yaml new file mode 100644 index 0000000000..7ea5afc795 --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen-turbo-0624.yaml @@ -0,0 +1,79 @@ +model: qwen-turbo-0624 +label: + en_US: qwen-turbo-0624 +model_type: llm +features: + - agent-thought +model_properties: + mode: completion + context_size: 8192 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 1500 + min: 1 + max: 1500 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.002' + output: '0.006' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-max-0201.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-max-0201.yaml new file mode 100644 index 0000000000..fffd732ca5 --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-max-0201.yaml @@ -0,0 +1,47 @@ +model: qwen-vl-max-0201 +label: + en_US: qwen-vl-max-0201 +model_type: llm +features: + - vision + - agent-thought +model_properties: + mode: chat + context_size: 8192 +parameter_rules: + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: response_format + use_template: response_format +pricing: + input: '0.02' + output: '0.02' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-max-0809.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-max-0809.yaml new file mode 100644 index 0000000000..af8742b981 --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-max-0809.yaml @@ -0,0 +1,47 @@ +model: qwen-vl-max-0809 +label: + en_US: qwen-vl-max-0809 +model_type: llm +features: + - vision + - agent-thought +model_properties: + mode: chat + context_size: 32768 +parameter_rules: + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: response_format + use_template: response_format +pricing: + input: '0.02' + output: '0.02' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-max.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-max.yaml index f917ccaa5d..a93d456428 100644 --- a/api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-max.yaml +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-max.yaml @@ -7,7 +7,7 @@ features: - agent-thought model_properties: mode: chat - context_size: 8192 + context_size: 32768 parameter_rules: - name: top_p use_template: top_p diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-plus-0809.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-plus-0809.yaml new file mode 100644 index 0000000000..12573511b9 --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-plus-0809.yaml @@ -0,0 +1,47 @@ +model: qwen-vl-plus-0809 +label: + en_US: qwen-vl-plus-0809 +model_type: llm +features: + - vision + - agent-thought +model_properties: + mode: chat + context_size: 32768 +parameter_rules: + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: response_format + use_template: response_format +pricing: + input: '0.008' + output: '0.008' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-plus.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-plus.yaml index e2dd8c4e57..13468c44ee 100644 --- a/api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-plus.yaml +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen-vl-plus.yaml @@ -7,7 +7,7 @@ features: - agent-thought model_properties: mode: chat - context_size: 32768 + context_size: 8192 parameter_rules: - name: top_p use_template: top_p diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen2-math-1.5b-instruct.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen2-math-1.5b-instruct.yaml new file mode 100644 index 0000000000..8b204ff1f0 --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen2-math-1.5b-instruct.yaml @@ -0,0 +1,79 @@ +model: qwen2-math-1.5b-instruct +label: + en_US: qwen2-math-1.5b-instruct +model_type: llm +features: + - agent-thought +model_properties: + mode: completion + context_size: 4096 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 2000 + min: 1 + max: 2000 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.004' + output: '0.012' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen2-math-72b-instruct.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen2-math-72b-instruct.yaml new file mode 100644 index 0000000000..3875a274e7 --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen2-math-72b-instruct.yaml @@ -0,0 +1,79 @@ +model: qwen2-math-72b-instruct +label: + en_US: qwen2-math-72b-instruct +model_type: llm +features: + - agent-thought +model_properties: + mode: completion + context_size: 4096 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 2000 + min: 1 + max: 2000 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.004' + output: '0.012' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen2-math-7b-instruct.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen2-math-7b-instruct.yaml new file mode 100644 index 0000000000..0920806845 --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen2-math-7b-instruct.yaml @@ -0,0 +1,79 @@ +model: qwen2-math-7b-instruct +label: + en_US: qwen2-math-7b-instruct +model_type: llm +features: + - agent-thought +model_properties: + mode: completion + context_size: 4096 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 2000 + min: 1 + max: 2000 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.004' + output: '0.012' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-0.5b-instruct.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-0.5b-instruct.yaml new file mode 100644 index 0000000000..824954323b --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-0.5b-instruct.yaml @@ -0,0 +1,79 @@ +model: qwen2.5-0.5b-instruct +label: + en_US: qwen2.5-0.5b-instruct +model_type: llm +features: + - agent-thought +model_properties: + mode: completion + context_size: 32768 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 8192 + min: 1 + max: 8192 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.000' + output: '0.000' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-1.5b-instruct.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-1.5b-instruct.yaml new file mode 100644 index 0000000000..c0a4b45be6 --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-1.5b-instruct.yaml @@ -0,0 +1,79 @@ +model: qwen2.5-1.5b-instruct +label: + en_US: qwen2.5-1.5b-instruct +model_type: llm +features: + - agent-thought +model_properties: + mode: completion + context_size: 32768 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 8192 + min: 1 + max: 8192 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.000' + output: '0.000' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-14b-instruct.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-14b-instruct.yaml new file mode 100644 index 0000000000..92b67804e8 --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-14b-instruct.yaml @@ -0,0 +1,79 @@ +model: qwen2.5-14b-instruct +label: + en_US: qwen2.5-14b-instruct +model_type: llm +features: + - agent-thought +model_properties: + mode: completion + context_size: 131072 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 8192 + min: 1 + max: 8192 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.002' + output: '0.006' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-32b-instruct.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-32b-instruct.yaml new file mode 100644 index 0000000000..960438e3e7 --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-32b-instruct.yaml @@ -0,0 +1,79 @@ +model: qwen2.5-32b-instruct +label: + en_US: qwen2.5-32b-instruct +model_type: llm +features: + - agent-thought +model_properties: + mode: completion + context_size: 131072 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 8192 + min: 1 + max: 8192 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.0035' + output: '0.007' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-3b-instruct.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-3b-instruct.yaml new file mode 100644 index 0000000000..59a8827d9e --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-3b-instruct.yaml @@ -0,0 +1,79 @@ +model: qwen2.5-3b-instruct +label: + en_US: qwen2.5-3b-instruct +model_type: llm +features: + - agent-thought +model_properties: + mode: completion + context_size: 32768 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 8192 + min: 1 + max: 8192 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.000' + output: '0.000' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-72b-instruct.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-72b-instruct.yaml new file mode 100644 index 0000000000..f14ee2daff --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-72b-instruct.yaml @@ -0,0 +1,79 @@ +model: qwen2.5-72b-instruct +label: + en_US: qwen2.5-72b-instruct +model_type: llm +features: + - agent-thought +model_properties: + mode: completion + context_size: 131072 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 8192 + min: 1 + max: 8192 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.004' + output: '0.012' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-7b-instruct.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-7b-instruct.yaml new file mode 100644 index 0000000000..8ea8166358 --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-7b-instruct.yaml @@ -0,0 +1,79 @@ +model: qwen2.5-7b-instruct +label: + en_US: qwen2.5-7b-instruct +model_type: llm +features: + - agent-thought +model_properties: + mode: completion + context_size: 131072 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 8192 + min: 1 + max: 8192 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.001' + output: '0.002' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-coder-7b-instruct.yaml b/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-coder-7b-instruct.yaml new file mode 100644 index 0000000000..8ea8166358 --- /dev/null +++ b/api/core/model_runtime/model_providers/tongyi/llm/qwen2.5-coder-7b-instruct.yaml @@ -0,0 +1,79 @@ +model: qwen2.5-7b-instruct +label: + en_US: qwen2.5-7b-instruct +model_type: llm +features: + - agent-thought +model_properties: + mode: completion + context_size: 131072 +parameter_rules: + - name: temperature + use_template: temperature + type: float + default: 0.3 + min: 0.0 + max: 2.0 + help: + zh_Hans: 用于控制随机性和多样性的程度。具体来说，temperature值控制了生成文本时对每个候选词的概率分布进行平滑的程度。较高的temperature值会降低概率分布的峰值，使得更多的低概率词被选择，生成结果更加多样化；而较低的temperature值则会增强概率分布的峰值，使得高概率词更容易被选择，生成结果更加确定。 + en_US: Used to control the degree of randomness and diversity. Specifically, the temperature value controls the degree to which the probability distribution of each candidate word is smoothed when generating text. A higher temperature value will reduce the peak value of the probability distribution, allowing more low-probability words to be selected, and the generated results will be more diverse; while a lower temperature value will enhance the peak value of the probability distribution, making it easier for high-probability words to be selected. , the generated results are more certain. + - name: max_tokens + use_template: max_tokens + type: int + default: 8192 + min: 1 + max: 8192 + help: + zh_Hans: 用于指定模型在生成内容时token的最大数量，它定义了生成的上限，但不保证每次都会生成到这个数量。 + en_US: It is used to specify the maximum number of tokens when the model generates content. It defines the upper limit of generation, but does not guarantee that this number will be generated every time. + - name: top_p + use_template: top_p + type: float + default: 0.8 + min: 0.1 + max: 0.9 + help: + zh_Hans: 生成过程中核采样方法概率阈值，例如，取值为0.8时，仅保留概率加起来大于等于0.8的最可能token的最小集合作为候选集。取值范围为（0,1.0)，取值越大，生成的随机性越高；取值越低，生成的确定性越高。 + en_US: The probability threshold of the kernel sampling method during the generation process. For example, when the value is 0.8, only the smallest set of the most likely tokens with a sum of probabilities greater than or equal to 0.8 is retained as the candidate set. The value range is (0,1.0). The larger the value, the higher the randomness generated; the lower the value, the higher the certainty generated. + - name: top_k + type: int + min: 0 + max: 99 + label: + zh_Hans: 取样数量 + en_US: Top k + help: + zh_Hans: 生成时，采样候选集的大小。例如，取值为50时，仅将单次生成中得分最高的50个token组成随机采样的候选集。取值越大，生成的随机性越高；取值越小，生成的确定性越高。 + en_US: The size of the sample candidate set when generated. For example, when the value is 50, only the 50 highest-scoring tokens in a single generation form a randomly sampled candidate set. The larger the value, the higher the randomness generated; the smaller the value, the higher the certainty generated. + - name: seed + required: false + type: int + default: 1234 + label: + zh_Hans: 随机种子 + en_US: Random seed + help: + zh_Hans: 生成时使用的随机数种子，用户控制模型生成内容的随机性。支持无符号64位整数，默认值为 1234。在使用seed时，模型将尽可能生成相同或相似的结果，但目前不保证每次生成的结果完全相同。 + en_US: The random number seed used when generating, the user controls the randomness of the content generated by the model. Supports unsigned 64-bit integers, default value is 1234. When using seed, the model will try its best to generate the same or similar results, but there is currently no guarantee that the results will be exactly the same every time. + - name: repetition_penalty + required: false + type: float + default: 1.1 + label: + en_US: Repetition penalty + help: + zh_Hans: 用于控制模型生成时的重复度。提高repetition_penalty时可以降低模型生成的重复度。1.0表示不做惩罚。 + en_US: Used to control the repeatability when generating models. Increasing repetition_penalty can reduce the duplication of model generation. 1.0 means no punishment. + - name: enable_search + type: boolean + default: false + help: + zh_Hans: 模型内置了互联网搜索服务，该参数控制模型在生成文本时是否参考使用互联网搜索结果。启用互联网搜索，模型会将搜索结果作为文本生成过程中的参考信息，但模型会基于其内部逻辑“自行判断”是否使用互联网搜索结果。 + en_US: The model has a built-in Internet search service. This parameter controls whether the model refers to Internet search results when generating text. When Internet search is enabled, the model will use the search results as reference information in the text generation process, but the model will "judge" whether to use Internet search results based on its internal logic. + - name: response_format + use_template: response_format +pricing: + input: '0.001' + output: '0.002' + unit: '0.001' + currency: RMB diff --git a/api/core/model_runtime/model_providers/tongyi/tongyi.yaml b/api/core/model_runtime/model_providers/tongyi/tongyi.yaml index b251391e34..de2c289c94 100644 --- a/api/core/model_runtime/model_providers/tongyi/tongyi.yaml +++ b/api/core/model_runtime/model_providers/tongyi/tongyi.yaml @@ -11,9 +11,9 @@ background: "#EFF1FE" help: title: en_US: Get your API key from AliCloud - zh_Hans: 从阿里云获取 API Key + zh_Hans: 从阿里云百炼获取 API Key url: - en_US: https://dashscope.console.aliyun.com/api-key_management + en_US: https://bailian.console.aliyun.com/?apiKey=1#/api-key supported_model_types: - llm - tts diff --git a/api/core/rag/retrieval/dataset_retrieval.py b/api/core/rag/retrieval/dataset_retrieval.py index 124c58f0fe..286ecd4c03 100644 --- a/api/core/rag/retrieval/dataset_retrieval.py +++ b/api/core/rag/retrieval/dataset_retrieval.py @@ -426,7 +426,7 @@ class DatasetRetrieval: retrieval_method=retrieval_model["search_method"], dataset_id=dataset.id, query=query, - top_k=top_k, + top_k=retrieval_model.get("top_k") or 2, score_threshold=retrieval_model.get("score_threshold", 0.0) if retrieval_model["score_threshold_enabled"] else 0.0, diff --git a/api/core/tools/provider/builtin/brave/brave.yaml b/api/core/tools/provider/builtin/brave/brave.yaml index 93d315f839..2b0dcc0188 100644 --- a/api/core/tools/provider/builtin/brave/brave.yaml +++ b/api/core/tools/provider/builtin/brave/brave.yaml @@ -29,3 +29,11 @@ credentials_for_provider: zh_Hans: 从 Brave 获取您的 Brave Search API key pt_BR: Get your Brave Search API key from Brave url: https://brave.com/search/api/ + base_url: + type: text-input + required: false + label: + en_US: Brave server's Base URL + zh_Hans: Brave服务器的API URL + placeholder: + en_US: https://api.search.brave.com/res/v1/web/search diff --git a/api/core/tools/provider/builtin/brave/tools/brave_search.py b/api/core/tools/provider/builtin/brave/tools/brave_search.py index 94a4d92844..c34362ae52 100644 --- a/api/core/tools/provider/builtin/brave/tools/brave_search.py +++ b/api/core/tools/provider/builtin/brave/tools/brave_search.py @@ -7,6 +7,8 @@ from pydantic import BaseModel, Field from core.tools.entities.tool_entities import ToolInvokeMessage from core.tools.tool.builtin_tool import BuiltinTool +BRAVE_BASE_URL = "https://api.search.brave.com/res/v1/web/search" + class BraveSearchWrapper(BaseModel): """Wrapper around the Brave search engine.""" @@ -15,8 +17,10 @@ class BraveSearchWrapper(BaseModel): """The API key to use for the Brave search engine.""" search_kwargs: dict = Field(default_factory=dict) """Additional keyword arguments to pass to the search request.""" - base_url: str = "https://api.search.brave.com/res/v1/web/search" + base_url: str = BRAVE_BASE_URL """The base URL for the Brave search engine.""" + ensure_ascii: bool = True + """Ensure the JSON output is ASCII encoded.""" def run(self, query: str) -> str: """Query the Brave search engine and return the results as a JSON string. @@ -36,7 +40,7 @@ class BraveSearchWrapper(BaseModel): } for item in web_search_results ] - return json.dumps(final_results) + return json.dumps(final_results, ensure_ascii=self.ensure_ascii) def _search_request(self, query: str) -> list[dict]: headers = { @@ -68,7 +72,9 @@ class BraveSearch(BaseModel): search_wrapper: BraveSearchWrapper @classmethod - def from_api_key(cls, api_key: str, search_kwargs: Optional[dict] = None, **kwargs: Any) -> "BraveSearch": + def from_api_key( + cls, api_key: str, base_url: str, search_kwargs: Optional[dict] = None, ensure_ascii: bool = True, **kwargs: Any + ) -> "BraveSearch": """Create a tool from an api key. Args: @@ -79,7 +85,9 @@ class BraveSearch(BaseModel): Returns: A tool. """ - wrapper = BraveSearchWrapper(api_key=api_key, search_kwargs=search_kwargs or {}) + wrapper = BraveSearchWrapper( + api_key=api_key, base_url=base_url, search_kwargs=search_kwargs or {}, ensure_ascii=ensure_ascii + ) return cls(search_wrapper=wrapper, **kwargs) def _run( @@ -109,11 +117,18 @@ class BraveSearchTool(BuiltinTool): query = tool_parameters.get("query", "") count = tool_parameters.get("count", 3) api_key = self.runtime.credentials["brave_search_api_key"] + base_url = self.runtime.credentials.get("base_url", BRAVE_BASE_URL) + ensure_ascii = tool_parameters.get("ensure_ascii", True) + + if len(base_url) == 0: + base_url = BRAVE_BASE_URL if not query: return self.create_text_message("Please input query") - tool = BraveSearch.from_api_key(api_key=api_key, search_kwargs={"count": count}) + tool = BraveSearch.from_api_key( + api_key=api_key, base_url=base_url, search_kwargs={"count": count}, ensure_ascii=ensure_ascii + ) results = tool._run(query) diff --git a/api/core/tools/provider/builtin/brave/tools/brave_search.yaml b/api/core/tools/provider/builtin/brave/tools/brave_search.yaml index b2a734c12d..5222a375f8 100644 --- a/api/core/tools/provider/builtin/brave/tools/brave_search.yaml +++ b/api/core/tools/provider/builtin/brave/tools/brave_search.yaml @@ -39,3 +39,15 @@ parameters: pt_BR: O número de resultados de pesquisa a serem retornados, permitindo que os usuários controlem a amplitude de sua saída de pesquisa. llm_description: Specifies the amount of search results to be displayed, offering users the ability to adjust the scope of their search findings. form: llm + - name: ensure_ascii + type: boolean + default: true + label: + en_US: Ensure ASCII + zh_Hans: 确保 ASCII + pt_BR: Ensure ASCII + human_description: + en_US: Ensure the JSON output is ASCII encoded + zh_Hans: 确保输出的 JSON 是 ASCII 编码 + pt_BR: Ensure the JSON output is ASCII encoded + form: form diff --git a/api/core/tools/provider/builtin/comfyui/_assets/icon.png b/api/core/tools/provider/builtin/comfyui/_assets/icon.png new file mode 100644 index 0000000000..958ec5c5cf Binary files /dev/null and b/api/core/tools/provider/builtin/comfyui/_assets/icon.png differ diff --git a/api/core/tools/provider/builtin/comfyui/comfyui.py b/api/core/tools/provider/builtin/comfyui/comfyui.py new file mode 100644 index 0000000000..7013a0b93c --- /dev/null +++ b/api/core/tools/provider/builtin/comfyui/comfyui.py @@ -0,0 +1,17 @@ +from typing import Any + +from core.tools.errors import ToolProviderCredentialValidationError +from core.tools.provider.builtin.comfyui.tools.comfyui_stable_diffusion import ComfyuiStableDiffusionTool +from core.tools.provider.builtin_tool_provider import BuiltinToolProviderController + + +class ComfyUIProvider(BuiltinToolProviderController): + def _validate_credentials(self, credentials: dict[str, Any]) -> None: + try: + ComfyuiStableDiffusionTool().fork_tool_runtime( + runtime={ + "credentials": credentials, + } + ).validate_models() + except Exception as e: + raise ToolProviderCredentialValidationError(str(e)) diff --git a/api/core/tools/provider/builtin/comfyui/comfyui.yaml b/api/core/tools/provider/builtin/comfyui/comfyui.yaml new file mode 100644 index 0000000000..066fd85308 --- /dev/null +++ b/api/core/tools/provider/builtin/comfyui/comfyui.yaml @@ -0,0 +1,42 @@ +identity: + author: Qun + name: comfyui + label: + en_US: ComfyUI + zh_Hans: ComfyUI + pt_BR: ComfyUI + description: + en_US: ComfyUI is a tool for generating images which can be deployed locally. + zh_Hans: ComfyUI 是一个可以在本地部署的图片生成的工具。 + pt_BR: ComfyUI is a tool for generating images which can be deployed locally. + icon: icon.png + tags: + - image +credentials_for_provider: + base_url: + type: text-input + required: true + label: + en_US: Base URL + zh_Hans: ComfyUI服务器的Base URL + pt_BR: Base URL + placeholder: + en_US: Please input your ComfyUI server's Base URL + zh_Hans: 请输入你的 ComfyUI 服务器的 Base URL + pt_BR: Please input your ComfyUI server's Base URL + model: + type: text-input + required: true + label: + en_US: Model with suffix + zh_Hans: 模型, 需要带后缀 + pt_BR: Model with suffix + placeholder: + en_US: Please input your model + zh_Hans: 请输入你的模型名称 + pt_BR: Please input your model + help: + en_US: The checkpoint name of the ComfyUI server, e.g. xxx.safetensors + zh_Hans: ComfyUI服务器的模型名称, 比如 xxx.safetensors + pt_BR: The checkpoint name of the ComfyUI server, e.g. xxx.safetensors + url: https://docs.dify.ai/tutorials/tool-configuration/comfyui diff --git a/api/core/tools/provider/builtin/comfyui/tools/comfyui_stable_diffusion.py b/api/core/tools/provider/builtin/comfyui/tools/comfyui_stable_diffusion.py new file mode 100644 index 0000000000..81fc8cc985 --- /dev/null +++ b/api/core/tools/provider/builtin/comfyui/tools/comfyui_stable_diffusion.py @@ -0,0 +1,475 @@ +import json +import os +import random +import uuid +from copy import deepcopy +from enum import Enum +from typing import Any, Union + +import websocket +from httpx import get, post +from yarl import URL + +from core.tools.entities.common_entities import I18nObject +from core.tools.entities.tool_entities import ToolInvokeMessage, ToolParameter, ToolParameterOption +from core.tools.errors import ToolProviderCredentialValidationError +from core.tools.tool.builtin_tool import BuiltinTool + +SD_TXT2IMG_OPTIONS = {} +LORA_NODE = { + "inputs": {"lora_name": "", "strength_model": 1, "strength_clip": 1, "model": ["11", 0], "clip": ["11", 1]}, + "class_type": "LoraLoader", + "_meta": {"title": "Load LoRA"}, +} +FluxGuidanceNode = { + "inputs": {"guidance": 3.5, "conditioning": ["6", 0]}, + "class_type": "FluxGuidance", + "_meta": {"title": "FluxGuidance"}, +} + + +class ModelType(Enum): + SD15 = 1 + SDXL = 2 + SD3 = 3 + FLUX = 4 + + +class ComfyuiStableDiffusionTool(BuiltinTool): + def _invoke( + self, user_id: str, tool_parameters: dict[str, Any] + ) -> Union[ToolInvokeMessage, list[ToolInvokeMessage]]: + """ + invoke tools + """ + # base url + base_url = self.runtime.credentials.get("base_url", "") + if not base_url: + return self.create_text_message("Please input base_url") + + if tool_parameters.get("model"): + self.runtime.credentials["model"] = tool_parameters["model"] + + model = self.runtime.credentials.get("model", None) + if not model: + return self.create_text_message("Please input model") + + # prompt + prompt = tool_parameters.get("prompt", "") + if not prompt: + return self.create_text_message("Please input prompt") + + # get negative prompt + negative_prompt = tool_parameters.get("negative_prompt", "") + + # get size + width = tool_parameters.get("width", 1024) + height = tool_parameters.get("height", 1024) + + # get steps + steps = tool_parameters.get("steps", 1) + + # get sampler_name + sampler_name = tool_parameters.get("sampler_name", "euler") + + # scheduler + scheduler = tool_parameters.get("scheduler", "normal") + + # get cfg + cfg = tool_parameters.get("cfg", 7.0) + + # get model type + model_type = tool_parameters.get("model_type", ModelType.SD15.name) + + # get lora + # supports up to 3 loras + lora_list = [] + lora_strength_list = [] + if tool_parameters.get("lora_1"): + lora_list.append(tool_parameters["lora_1"]) + lora_strength_list.append(tool_parameters.get("lora_strength_1", 1)) + if tool_parameters.get("lora_2"): + lora_list.append(tool_parameters["lora_2"]) + lora_strength_list.append(tool_parameters.get("lora_strength_2", 1)) + if tool_parameters.get("lora_3"): + lora_list.append(tool_parameters["lora_3"]) + lora_strength_list.append(tool_parameters.get("lora_strength_3", 1)) + + return self.text2img( + base_url=base_url, + model=model, + model_type=model_type, + prompt=prompt, + negative_prompt=negative_prompt, + width=width, + height=height, + steps=steps, + sampler_name=sampler_name, + scheduler=scheduler, + cfg=cfg, + lora_list=lora_list, + lora_strength_list=lora_strength_list, + ) + + def get_checkpoints(self) -> list[str]: + """ + get checkpoints + """ + try: + base_url = self.runtime.credentials.get("base_url", None) + if not base_url: + return [] + api_url = str(URL(base_url) / "models" / "checkpoints") + response = get(url=api_url, timeout=(2, 10)) + if response.status_code != 200: + return [] + else: + return response.json() + except Exception as e: + return [] + + def get_loras(self) -> list[str]: + """ + get loras + """ + try: + base_url = self.runtime.credentials.get("base_url", None) + if not base_url: + return [] + api_url = str(URL(base_url) / "models" / "loras") + response = get(url=api_url, timeout=(2, 10)) + if response.status_code != 200: + return [] + else: + return response.json() + except Exception as e: + return [] + + def get_sample_methods(self) -> tuple[list[str], list[str]]: + """ + get sample method + """ + try: + base_url = self.runtime.credentials.get("base_url", None) + if not base_url: + return [], [] + api_url = str(URL(base_url) / "object_info" / "KSampler") + response = get(url=api_url, timeout=(2, 10)) + if response.status_code != 200: + return [], [] + else: + data = response.json()["KSampler"]["input"]["required"] + return data["sampler_name"][0], data["scheduler"][0] + except Exception as e: + return [], [] + + def validate_models(self) -> Union[ToolInvokeMessage, list[ToolInvokeMessage]]: + """ + validate models + """ + try: + base_url = self.runtime.credentials.get("base_url", None) + if not base_url: + raise ToolProviderCredentialValidationError("Please input base_url") + model = self.runtime.credentials.get("model", None) + if not model: + raise ToolProviderCredentialValidationError("Please input model") + + api_url = str(URL(base_url) / "models" / "checkpoints") + response = get(url=api_url, timeout=(2, 10)) + if response.status_code != 200: + raise ToolProviderCredentialValidationError("Failed to get models") + else: + models = response.json() + if len([d for d in models if d == model]) > 0: + return self.create_text_message(json.dumps(models)) + else: + raise ToolProviderCredentialValidationError(f"model {model} does not exist") + except Exception as e: + raise ToolProviderCredentialValidationError(f"Failed to get models, {e}") + + def get_history(self, base_url, prompt_id): + """ + get history + """ + url = str(URL(base_url) / "history") + respond = get(url, params={"prompt_id": prompt_id}, timeout=(2, 10)) + return respond.json() + + def download_image(self, base_url, filename, subfolder, folder_type): + """ + download image + """ + url = str(URL(base_url) / "view") + response = get(url, params={"filename": filename, "subfolder": subfolder, "type": folder_type}, timeout=(2, 10)) + return response.content + + def queue_prompt_image(self, base_url, client_id, prompt): + """ + send prompt task and rotate + """ + # initiate task execution + url = str(URL(base_url) / "prompt") + respond = post(url, data=json.dumps({"client_id": client_id, "prompt": prompt}), timeout=(2, 10)) + prompt_id = respond.json()["prompt_id"] + + ws = websocket.WebSocket() + if "https" in base_url: + ws_url = base_url.replace("https", "ws") + else: + ws_url = base_url.replace("http", "ws") + ws.connect(str(URL(f"{ws_url}") / "ws") + f"?clientId={client_id}", timeout=120) + + # websocket rotate execution status + output_images = {} + while True: + out = ws.recv() + if isinstance(out, str): + message = json.loads(out) + if message["type"] == "executing": + data = message["data"] + if data["node"] is None and data["prompt_id"] == prompt_id: + break # Execution is done + elif message["type"] == "status": + data = message["data"] + if data["status"]["exec_info"]["queue_remaining"] == 0 and data.get("sid"): + break # Execution is done + else: + continue # previews are binary data + + # download image when execution finished + history = self.get_history(base_url, prompt_id)[prompt_id] + for o in history["outputs"]: + for node_id in history["outputs"]: + node_output = history["outputs"][node_id] + if "images" in node_output: + images_output = [] + for image in node_output["images"]: + image_data = self.download_image(base_url, image["filename"], image["subfolder"], image["type"]) + images_output.append(image_data) + output_images[node_id] = images_output + + ws.close() + + return output_images + + def text2img( + self, + base_url: str, + model: str, + model_type: str, + prompt: str, + negative_prompt: str, + width: int, + height: int, + steps: int, + sampler_name: str, + scheduler: str, + cfg: float, + lora_list: list, + lora_strength_list: list, + ) -> Union[ToolInvokeMessage, list[ToolInvokeMessage]]: + """ + generate image + """ + if not SD_TXT2IMG_OPTIONS: + current_dir = os.path.dirname(os.path.realpath(__file__)) + with open(os.path.join(current_dir, "txt2img.json")) as file: + SD_TXT2IMG_OPTIONS.update(json.load(file)) + + draw_options = deepcopy(SD_TXT2IMG_OPTIONS) + draw_options["3"]["inputs"]["steps"] = steps + draw_options["3"]["inputs"]["sampler_name"] = sampler_name + draw_options["3"]["inputs"]["scheduler"] = scheduler + draw_options["3"]["inputs"]["cfg"] = cfg + # generate different image when using same prompt next time + draw_options["3"]["inputs"]["seed"] = random.randint(0, 100000000) + draw_options["4"]["inputs"]["ckpt_name"] = model + draw_options["5"]["inputs"]["width"] = width + draw_options["5"]["inputs"]["height"] = height + draw_options["6"]["inputs"]["text"] = prompt + draw_options["7"]["inputs"]["text"] = negative_prompt + # if the model is SD3 or FLUX series, the Latent class should be corresponding to SD3 Latent + if model_type in {ModelType.SD3.name, ModelType.FLUX.name}: + draw_options["5"]["class_type"] = "EmptySD3LatentImage" + + if lora_list: + # last Lora node link to KSampler node + draw_options["3"]["inputs"]["model"][0] = "10" + # last Lora node link to positive and negative Clip node + draw_options["6"]["inputs"]["clip"][0] = "10" + draw_options["7"]["inputs"]["clip"][0] = "10" + # every Lora node link to next Lora node, and Checkpoints node link to first Lora node + for i, (lora, strength) in enumerate(zip(lora_list, lora_strength_list), 10): + if i - 10 == len(lora_list) - 1: + next_node_id = "4" + else: + next_node_id = str(i + 1) + lora_node = deepcopy(LORA_NODE) + lora_node["inputs"]["lora_name"] = lora + lora_node["inputs"]["strength_model"] = strength + lora_node["inputs"]["strength_clip"] = strength + lora_node["inputs"]["model"][0] = next_node_id + lora_node["inputs"]["clip"][0] = next_node_id + draw_options[str(i)] = lora_node + + # FLUX need to add FluxGuidance Node + if model_type == ModelType.FLUX.name: + last_node_id = str(10 + len(lora_list)) + draw_options[last_node_id] = deepcopy(FluxGuidanceNode) + draw_options[last_node_id]["inputs"]["conditioning"][0] = "6" + draw_options["3"]["inputs"]["positive"][0] = last_node_id + + try: + client_id = str(uuid.uuid4()) + result = self.queue_prompt_image(base_url, client_id, prompt=draw_options) + + # get first image + image = b"" + for node in result: + for img in result[node]: + if img: + image = img + break + + return self.create_blob_message( + blob=image, meta={"mime_type": "image/png"}, save_as=self.VARIABLE_KEY.IMAGE.value + ) + + except Exception as e: + return self.create_text_message(f"Failed to generate image: {str(e)}") + + def get_runtime_parameters(self) -> list[ToolParameter]: + parameters = [ + ToolParameter( + name="prompt", + label=I18nObject(en_US="Prompt", zh_Hans="Prompt"), + human_description=I18nObject( + en_US="Image prompt, you can check the official documentation of Stable Diffusion", + zh_Hans="图像提示词，您可以查看 Stable Diffusion 的官方文档", + ), + type=ToolParameter.ToolParameterType.STRING, + form=ToolParameter.ToolParameterForm.LLM, + llm_description="Image prompt of Stable Diffusion, you should describe the image " + "you want to generate as a list of words as possible as detailed, " + "the prompt must be written in English.", + required=True, + ), + ] + if self.runtime.credentials: + try: + models = self.get_checkpoints() + if len(models) != 0: + parameters.append( + ToolParameter( + name="model", + label=I18nObject(en_US="Model", zh_Hans="Model"), + human_description=I18nObject( + en_US="Model of Stable Diffusion or FLUX, " + "you can check the official documentation of Stable Diffusion or FLUX", + zh_Hans="Stable Diffusion 或者 FLUX 的模型，您可以查看 Stable Diffusion 的官方文档", + ), + type=ToolParameter.ToolParameterType.SELECT, + form=ToolParameter.ToolParameterForm.FORM, + llm_description="Model of Stable Diffusion or FLUX, " + "you can check the official documentation of Stable Diffusion or FLUX", + required=True, + default=models[0], + options=[ + ToolParameterOption(value=i, label=I18nObject(en_US=i, zh_Hans=i)) for i in models + ], + ) + ) + loras = self.get_loras() + if len(loras) != 0: + for n in range(1, 4): + parameters.append( + ToolParameter( + name=f"lora_{n}", + label=I18nObject(en_US=f"Lora {n}", zh_Hans=f"Lora {n}"), + human_description=I18nObject( + en_US="Lora of Stable Diffusion, " + "you can check the official documentation of Stable Diffusion", + zh_Hans="Stable Diffusion 的 Lora 模型，您可以查看 Stable Diffusion 的官方文档", + ), + type=ToolParameter.ToolParameterType.SELECT, + form=ToolParameter.ToolParameterForm.FORM, + llm_description="Lora of Stable Diffusion, " + "you can check the official documentation of " + "Stable Diffusion", + required=False, + options=[ + ToolParameterOption(value=i, label=I18nObject(en_US=i, zh_Hans=i)) for i in loras + ], + ) + ) + sample_methods, schedulers = self.get_sample_methods() + if len(sample_methods) != 0: + parameters.append( + ToolParameter( + name="sampler_name", + label=I18nObject(en_US="Sampling method", zh_Hans="Sampling method"), + human_description=I18nObject( + en_US="Sampling method of Stable Diffusion, " + "you can check the official documentation of Stable Diffusion", + zh_Hans="Stable Diffusion 的Sampling method，您可以查看 Stable Diffusion 的官方文档", + ), + type=ToolParameter.ToolParameterType.SELECT, + form=ToolParameter.ToolParameterForm.FORM, + llm_description="Sampling method of Stable Diffusion, " + "you can check the official documentation of Stable Diffusion", + required=True, + default=sample_methods[0], + options=[ + ToolParameterOption(value=i, label=I18nObject(en_US=i, zh_Hans=i)) + for i in sample_methods + ], + ) + ) + if len(schedulers) != 0: + parameters.append( + ToolParameter( + name="scheduler", + label=I18nObject(en_US="Scheduler", zh_Hans="Scheduler"), + human_description=I18nObject( + en_US="Scheduler of Stable Diffusion, " + "you can check the official documentation of Stable Diffusion", + zh_Hans="Stable Diffusion 的Scheduler，您可以查看 Stable Diffusion 的官方文档", + ), + type=ToolParameter.ToolParameterType.SELECT, + form=ToolParameter.ToolParameterForm.FORM, + llm_description="Scheduler of Stable Diffusion, " + "you can check the official documentation of Stable Diffusion", + required=True, + default=schedulers[0], + options=[ + ToolParameterOption(value=i, label=I18nObject(en_US=i, zh_Hans=i)) for i in schedulers + ], + ) + ) + parameters.append( + ToolParameter( + name="model_type", + label=I18nObject(en_US="Model Type", zh_Hans="Model Type"), + human_description=I18nObject( + en_US="Model Type of Stable Diffusion or Flux, " + "you can check the official documentation of Stable Diffusion or Flux", + zh_Hans="Stable Diffusion 或 FLUX 的模型类型，" + "您可以查看 Stable Diffusion 或 Flux 的官方文档", + ), + type=ToolParameter.ToolParameterType.SELECT, + form=ToolParameter.ToolParameterForm.FORM, + llm_description="Model Type of Stable Diffusion or Flux, " + "you can check the official documentation of Stable Diffusion or Flux", + required=True, + default=ModelType.SD15.name, + options=[ + ToolParameterOption(value=i, label=I18nObject(en_US=i, zh_Hans=i)) + for i in ModelType.__members__ + ], + ) + ) + except: + pass + + return parameters diff --git a/api/core/tools/provider/builtin/comfyui/tools/comfyui_stable_diffusion.yaml b/api/core/tools/provider/builtin/comfyui/tools/comfyui_stable_diffusion.yaml new file mode 100644 index 0000000000..4f4a6942b3 --- /dev/null +++ b/api/core/tools/provider/builtin/comfyui/tools/comfyui_stable_diffusion.yaml @@ -0,0 +1,212 @@ +identity: + name: txt2img workflow + author: Qun + label: + en_US: Txt2Img Workflow + zh_Hans: Txt2Img Workflow + pt_BR: Txt2Img Workflow +description: + human: + en_US: a pre-defined comfyui workflow that can use one model and up to 3 loras to generate images. Support SD1.5, SDXL, SD3 and FLUX which contain text encoders/clip, but does not support models that requires a triple clip loader. + zh_Hans: 一个预定义的 ComfyUI 工作流，可以使用一个模型和最多3个loras来生成图像。支持包含文本编码器/clip的SD1.5、SDXL、SD3和FLUX，但不支持需要clip加载器的模型。 + pt_BR: a pre-defined comfyui workflow that can use one model and up to 3 loras to generate images. Support SD1.5, SDXL, SD3 and FLUX which contain text encoders/clip, but does not support models that requires a triple clip loader. + llm: draw the image you want based on your prompt. +parameters: + - name: prompt + type: string + required: true + label: + en_US: Prompt + zh_Hans: 提示词 + pt_BR: Prompt + human_description: + en_US: Image prompt, you can check the official documentation of Stable Diffusion or FLUX + zh_Hans: 图像提示词，您可以查看 Stable Diffusion 或者 FLUX 的官方文档 + pt_BR: Image prompt, you can check the official documentation of Stable Diffusion or FLUX + llm_description: Image prompt of Stable Diffusion, you should describe the image you want to generate as a list of words as possible as detailed, the prompt must be written in English. + form: llm + - name: model + type: string + required: true + label: + en_US: Model Name + zh_Hans: 模型名称 + pt_BR: Model Name + human_description: + en_US: Model Name + zh_Hans: 模型名称 + pt_BR: Model Name + form: form + - name: model_type + type: string + required: true + label: + en_US: Model Type + zh_Hans: 模型类型 + pt_BR: Model Type + human_description: + en_US: Model Type + zh_Hans: 模型类型 + pt_BR: Model Type + form: form + - name: lora_1 + type: string + required: false + label: + en_US: Lora 1 + zh_Hans: Lora 1 + pt_BR: Lora 1 + human_description: + en_US: Lora 1 + zh_Hans: Lora 1 + pt_BR: Lora 1 + form: form + - name: lora_strength_1 + type: number + required: false + label: + en_US: Lora Strength 1 + zh_Hans: Lora Strength 1 + pt_BR: Lora Strength 1 + human_description: + en_US: Lora Strength 1 + zh_Hans: Lora模型的权重 + pt_BR: Lora Strength 1 + form: form + - name: steps + type: number + required: false + label: + en_US: Steps + zh_Hans: Steps + pt_BR: Steps + human_description: + en_US: Steps + zh_Hans: Steps + pt_BR: Steps + form: form + default: 20 + - name: width + type: number + required: false + label: + en_US: Width + zh_Hans: Width + pt_BR: Width + human_description: + en_US: Width + zh_Hans: Width + pt_BR: Width + form: form + default: 1024 + - name: height + type: number + required: false + label: + en_US: Height + zh_Hans: Height + pt_BR: Height + human_description: + en_US: Height + zh_Hans: Height + pt_BR: Height + form: form + default: 1024 + - name: negative_prompt + type: string + required: false + label: + en_US: Negative prompt + zh_Hans: Negative prompt + pt_BR: Negative prompt + human_description: + en_US: Negative prompt + zh_Hans: Negative prompt + pt_BR: Negative prompt + form: form + default: bad art, ugly, deformed, watermark, duplicated, discontinuous lines + - name: cfg + type: number + required: false + label: + en_US: CFG Scale + zh_Hans: CFG Scale + pt_BR: CFG Scale + human_description: + en_US: CFG Scale + zh_Hans: 提示词相关性(CFG Scale) + pt_BR: CFG Scale + form: form + default: 7.0 + - name: sampler_name + type: string + required: false + label: + en_US: Sampling method + zh_Hans: Sampling method + pt_BR: Sampling method + human_description: + en_US: Sampling method + zh_Hans: Sampling method + pt_BR: Sampling method + form: form + - name: scheduler + type: string + required: false + label: + en_US: Scheduler + zh_Hans: Scheduler + pt_BR: Scheduler + human_description: + en_US: Scheduler + zh_Hans: Scheduler + pt_BR: Scheduler + form: form + - name: lora_2 + type: string + required: false + label: + en_US: Lora 2 + zh_Hans: Lora 2 + pt_BR: Lora 2 + human_description: + en_US: Lora 2 + zh_Hans: Lora 2 + pt_BR: Lora 2 + form: form + - name: lora_strength_2 + type: number + required: false + label: + en_US: Lora Strength 2 + zh_Hans: Lora Strength 2 + pt_BR: Lora Strength 2 + human_description: + en_US: Lora Strength 2 + zh_Hans: Lora模型的权重 + pt_BR: Lora Strength 2 + form: form + - name: lora_3 + type: string + required: false + label: + en_US: Lora 3 + zh_Hans: Lora 3 + pt_BR: Lora 3 + human_description: + en_US: Lora 3 + zh_Hans: Lora 3 + pt_BR: Lora 3 + form: form + - name: lora_strength_3 + type: number + required: false + label: + en_US: Lora Strength 3 + zh_Hans: Lora Strength 3 + pt_BR: Lora Strength 3 + human_description: + en_US: Lora Strength 3 + zh_Hans: Lora模型的权重 + pt_BR: Lora Strength 3 + form: form diff --git a/api/core/tools/provider/builtin/comfyui/tools/txt2img.json b/api/core/tools/provider/builtin/comfyui/tools/txt2img.json new file mode 100644 index 0000000000..8ea869ff10 --- /dev/null +++ b/api/core/tools/provider/builtin/comfyui/tools/txt2img.json @@ -0,0 +1,107 @@ +{ + "3": { + "inputs": { + "seed": 156680208700286, + "steps": 20, + "cfg": 8, + "sampler_name": "euler", + "scheduler": "normal", + "denoise": 1, + "model": [ + "4", + 0 + ], + "positive": [ + "6", + 0 + ], + "negative": [ + "7", + 0 + ], + "latent_image": [ + "5", + 0 + ] + }, + "class_type": "KSampler", + "_meta": { + "title": "KSampler" + } + }, + "4": { + "inputs": { + "ckpt_name": "3dAnimationDiffusion_v10.safetensors" + }, + "class_type": "CheckpointLoaderSimple", + "_meta": { + "title": "Load Checkpoint" + } + }, + "5": { + "inputs": { + "width": 512, + "height": 512, + "batch_size": 1 + }, + "class_type": "EmptyLatentImage", + "_meta": { + "title": "Empty Latent Image" + } + }, + "6": { + "inputs": { + "text": "beautiful scenery nature glass bottle landscape, , purple galaxy bottle,", + "clip": [ + "4", + 1 + ] + }, + "class_type": "CLIPTextEncode", + "_meta": { + "title": "CLIP Text Encode (Prompt)" + } + }, + "7": { + "inputs": { + "text": "text, watermark", + "clip": [ + "4", + 1 + ] + }, + "class_type": "CLIPTextEncode", + "_meta": { + "title": "CLIP Text Encode (Prompt)" + } + }, + "8": { + "inputs": { + "samples": [ + "3", + 0 + ], + "vae": [ + "4", + 2 + ] + }, + "class_type": "VAEDecode", + "_meta": { + "title": "VAE Decode" + } + }, + "9": { + "inputs": { + "filename_prefix": "ComfyUI", + "images": [ + "8", + 0 + ] + }, + "class_type": "SaveImage", + "_meta": { + "title": "Save Image" + } + } +} \ No newline at end of file diff --git a/api/core/tools/provider/builtin/feishu_document/tools/get_document_raw_content.py b/api/core/tools/provider/builtin/feishu_document/tools/get_document_content.py similarity index 79% rename from api/core/tools/provider/builtin/feishu_document/tools/get_document_raw_content.py rename to api/core/tools/provider/builtin/feishu_document/tools/get_document_content.py index 83073e0822..c94a5f70ed 100644 --- a/api/core/tools/provider/builtin/feishu_document/tools/get_document_raw_content.py +++ b/api/core/tools/provider/builtin/feishu_document/tools/get_document_content.py @@ -12,6 +12,8 @@ class GetDocumentRawContentTool(BuiltinTool): client = FeishuRequest(app_id, app_secret) document_id = tool_parameters.get("document_id") + mode = tool_parameters.get("mode") + lang = tool_parameters.get("lang", 0) - res = client.get_document_raw_content(document_id) + res = client.get_document_content(document_id, mode, lang) return self.create_json_message(res) diff --git a/api/core/tools/provider/builtin/feishu_document/tools/get_document_content.yaml b/api/core/tools/provider/builtin/feishu_document/tools/get_document_content.yaml new file mode 100644 index 0000000000..51eda73a60 --- /dev/null +++ b/api/core/tools/provider/builtin/feishu_document/tools/get_document_content.yaml @@ -0,0 +1,49 @@ +identity: + name: get_document_content + author: Doug Lea + label: + en_US: Get Document Content + zh_Hans: 获取飞书云文档的内容 +description: + human: + en_US: Get document content + zh_Hans: 获取飞书云文档的内容 + llm: A tool for retrieving content from Feishu cloud documents. +parameters: + - name: document_id + type: string + required: true + label: + en_US: document_id + zh_Hans: 飞书文档的唯一标识 + human_description: + en_US: Unique identifier for a Feishu document. You can also input the document's URL. + zh_Hans: 飞书文档的唯一标识，支持输入文档的 URL。 + llm_description: 飞书文档的唯一标识，支持输入文档的 URL。 + form: llm + + - name: mode + type: string + required: false + label: + en_US: mode + zh_Hans: 文档返回格式 + human_description: + en_US: Format of the document return, optional values are text, markdown, can be empty, default is markdown. + zh_Hans: 文档返回格式，可选值有 text、markdown，可以为空，默认值为 markdown。 + llm_description: 文档返回格式，可选值有 text、markdown，可以为空，默认值为 markdown。 + form: llm + + - name: lang + type: number + required: false + default: 0 + label: + en_US: lang + zh_Hans: 指定@用户的语言 + human_description: + en_US: | + Specifies the language for MentionUser, optional values are [0, 1]. 0: User's default name, 1: User's English name, default is 0. + zh_Hans: 指定返回的 MentionUser，即 @用户 的语言，可选值有 [0,1]。0：该用户的默认名称，1：该用户的英文名称，默认值为 0。 + llm_description: 指定返回的 MentionUser，即 @用户 的语言，可选值有 [0,1]。0：该用户的默认名称，1：该用户的英文名称，默认值为 0。 + form: llm diff --git a/api/core/tools/provider/builtin/feishu_document/tools/get_document_raw_content.yaml b/api/core/tools/provider/builtin/feishu_document/tools/get_document_raw_content.yaml deleted file mode 100644 index e5b0937e03..0000000000 --- a/api/core/tools/provider/builtin/feishu_document/tools/get_document_raw_content.yaml +++ /dev/null @@ -1,23 +0,0 @@ -identity: - name: get_document_raw_content - author: Doug Lea - label: - en_US: Get Document Raw Content - zh_Hans: 获取文档纯文本内容 -description: - human: - en_US: Get document raw content - zh_Hans: 获取文档纯文本内容 - llm: A tool for getting the plain text content of Feishu documents -parameters: - - name: document_id - type: string - required: true - label: - en_US: document_id - zh_Hans: 飞书文档的唯一标识 - human_description: - en_US: Unique ID of Feishu document document_id - zh_Hans: 飞书文档的唯一标识 document_id - llm_description: 飞书文档的唯一标识 document_id - form: llm diff --git a/api/core/tools/provider/builtin/feishu_document/tools/list_document_block.yaml b/api/core/tools/provider/builtin/feishu_document/tools/list_document_block.yaml deleted file mode 100644 index d51e5a837c..0000000000 --- a/api/core/tools/provider/builtin/feishu_document/tools/list_document_block.yaml +++ /dev/null @@ -1,48 +0,0 @@ -identity: - name: list_document_block - author: Doug Lea - label: - en_US: List Document Block - zh_Hans: 获取飞书文档所有块 -description: - human: - en_US: List document block - zh_Hans: 获取飞书文档所有块的富文本内容并分页返回。 - llm: A tool to get all blocks of Feishu documents -parameters: - - name: document_id - type: string - required: true - label: - en_US: document_id - zh_Hans: 飞书文档的唯一标识 - human_description: - en_US: Unique ID of Feishu document document_id - zh_Hans: 飞书文档的唯一标识 document_id - llm_description: 飞书文档的唯一标识 document_id - form: llm - - - name: page_size - type: number - required: false - default: 500 - label: - en_US: page_size - zh_Hans: 分页大小 - human_description: - en_US: Paging size, the default and maximum value is 500. - zh_Hans: 分页大小, 默认值和最大值为 500。 - llm_description: 分页大小, 表示一次请求最多返回多少条数据，默认值和最大值为 500。 - form: llm - - - name: page_token - type: string - required: false - label: - en_US: page_token - zh_Hans: 分页标记 - human_description: - en_US: Pagination tag, used to paginate query results so that more items can be obtained in the next traversal. - zh_Hans: 分页标记，用于分页查询结果，以便下次遍历时获取更多项。 - llm_description: 分页标记，第一次请求不填，表示从头开始遍历；分页查询结果还有更多项时会同时返回新的 page_token，下次遍历可采用该 page_token 获取查询结果。 - form: llm diff --git a/api/core/tools/provider/builtin/feishu_document/tools/list_document_block.py b/api/core/tools/provider/builtin/feishu_document/tools/list_document_blocks.py similarity index 90% rename from api/core/tools/provider/builtin/feishu_document/tools/list_document_block.py rename to api/core/tools/provider/builtin/feishu_document/tools/list_document_blocks.py index 8c0c4a3c97..572a7abf28 100644 --- a/api/core/tools/provider/builtin/feishu_document/tools/list_document_block.py +++ b/api/core/tools/provider/builtin/feishu_document/tools/list_document_blocks.py @@ -15,5 +15,5 @@ class ListDocumentBlockTool(BuiltinTool): page_size = tool_parameters.get("page_size", 500) page_token = tool_parameters.get("page_token", "") - res = client.list_document_block(document_id, page_token, page_size) + res = client.list_document_blocks(document_id, page_token, page_size) return self.create_json_message(res) diff --git a/api/core/tools/provider/builtin/feishu_document/tools/list_document_blocks.yaml b/api/core/tools/provider/builtin/feishu_document/tools/list_document_blocks.yaml new file mode 100644 index 0000000000..019ac98390 --- /dev/null +++ b/api/core/tools/provider/builtin/feishu_document/tools/list_document_blocks.yaml @@ -0,0 +1,74 @@ +identity: + name: list_document_blocks + author: Doug Lea + label: + en_US: List Document Blocks + zh_Hans: 获取飞书文档所有块 +description: + human: + en_US: List document blocks + zh_Hans: 获取飞书文档所有块的富文本内容并分页返回 + llm: A tool to get all blocks of Feishu documents +parameters: + - name: document_id + type: string + required: true + label: + en_US: document_id + zh_Hans: 飞书文档的唯一标识 + human_description: + en_US: Unique identifier for a Feishu document. You can also input the document's URL. + zh_Hans: 飞书文档的唯一标识，支持输入文档的 URL。 + llm_description: 飞书文档的唯一标识，支持输入文档的 URL。 + form: llm + + - name: user_id_type + type: select + required: false + options: + - value: open_id + label: + en_US: open_id + zh_Hans: open_id + - value: union_id + label: + en_US: union_id + zh_Hans: union_id + - value: user_id + label: + en_US: user_id + zh_Hans: user_id + default: "open_id" + label: + en_US: user_id_type + zh_Hans: 用户 ID 类型 + human_description: + en_US: User ID type, optional values are open_id, union_id, user_id, with a default value of open_id. + zh_Hans: 用户 ID 类型，可选值有 open_id、union_id、user_id，默认值为 open_id。 + llm_description: 用户 ID 类型，可选值有 open_id、union_id、user_id，默认值为 open_id。 + form: llm + + - name: page_size + type: number + required: false + default: "500" + label: + en_US: page_size + zh_Hans: 分页大小 + human_description: + en_US: Paging size, the default and maximum value is 500. + zh_Hans: 分页大小, 默认值和最大值为 500。 + llm_description: 分页大小, 表示一次请求最多返回多少条数据，默认值和最大值为 500。 + form: llm + + - name: page_token + type: string + required: false + label: + en_US: page_token + zh_Hans: 分页标记 + human_description: + en_US: Pagination token used to navigate through query results, allowing retrieval of additional items in subsequent requests. + zh_Hans: 分页标记，用于分页查询结果，以便下次遍历时获取更多项。 + llm_description: 分页标记，第一次请求不填，表示从头开始遍历；分页查询结果还有更多项时会同时返回新的 page_token，下次遍历可采用该 page_token 获取查询结果。 + form: llm diff --git a/api/core/tools/provider/builtin/feishu_document/tools/write_document.yaml b/api/core/tools/provider/builtin/feishu_document/tools/write_document.yaml index 8ee219d4a7..4282e3dcf3 100644 --- a/api/core/tools/provider/builtin/feishu_document/tools/write_document.yaml +++ b/api/core/tools/provider/builtin/feishu_document/tools/write_document.yaml @@ -17,33 +17,35 @@ parameters: en_US: document_id zh_Hans: 飞书文档的唯一标识 human_description: - en_US: Unique ID of Feishu document document_id - zh_Hans: 飞书文档的唯一标识 document_id - llm_description: 飞书文档的唯一标识 document_id + en_US: Unique identifier for a Feishu document. You can also input the document's URL. + zh_Hans: 飞书文档的唯一标识，支持输入文档的 URL。 + llm_description: 飞书文档的唯一标识，支持输入文档的 URL。 form: llm - name: content type: string required: true label: - en_US: document content - zh_Hans: 文档内容 + en_US: Plain text or Markdown content + zh_Hans: 纯文本或 Markdown 内容 human_description: - en_US: Document content, supports markdown syntax, can be empty. - zh_Hans: 文档内容，支持 markdown 语法，可以为空。 - llm_description: + en_US: Plain text or Markdown content. Note that embedded tables in the document should not have merged cells. + zh_Hans: 纯文本或 Markdown 内容。注意文档的内嵌套表格不允许有单元格合并。 + llm_description: 纯文本或 Markdown 内容，注意文档的内嵌套表格不允许有单元格合并。 form: llm - name: position - type: select - required: true - default: start + type: string + required: false label: - en_US: Choose where to add content - zh_Hans: 选择添加内容的位置 + en_US: position + zh_Hans: 添加位置 human_description: - en_US: Please fill in start or end to add content at the beginning or end of the document respectively. - zh_Hans: 请填入 start 或 end, 分别表示在文档开头(start)或结尾(end)添加内容。 + en_US: | + Enumeration values: start or end. Use 'start' to add content at the beginning of the document, and 'end' to add content at the end. The default value is 'end'. + zh_Hans: 枚举值：start 或 end。使用 'start' 在文档开头添加内容，使用 'end' 在文档结尾添加内容，默认值为 'end'。 + llm_description: | + 枚举值 start、end，start: 在文档开头添加内容；end: 在文档结尾添加内容，默认值为 end。 form: llm options: - value: start @@ -54,3 +56,4 @@ parameters: label: en_US: end zh_Hans: 在文档结尾添加内容 + default: start diff --git a/api/core/tools/provider/builtin/siliconflow/tools/flux.py b/api/core/tools/provider/builtin/siliconflow/tools/flux.py index 1b846624bd..0d16ff385e 100644 --- a/api/core/tools/provider/builtin/siliconflow/tools/flux.py +++ b/api/core/tools/provider/builtin/siliconflow/tools/flux.py @@ -5,7 +5,10 @@ import requests from core.tools.entities.tool_entities import ToolInvokeMessage from core.tools.tool.builtin_tool import BuiltinTool -FLUX_URL = "https://api.siliconflow.cn/v1/black-forest-labs/FLUX.1-schnell/text-to-image" +FLUX_URL = { + "schnell": "https://api.siliconflow.cn/v1/black-forest-labs/FLUX.1-schnell/text-to-image", + "dev": "https://api.siliconflow.cn/v1/image/generations", +} class FluxTool(BuiltinTool): @@ -24,8 +27,12 @@ class FluxTool(BuiltinTool): "seed": tool_parameters.get("seed"), "num_inference_steps": tool_parameters.get("num_inference_steps", 20), } + model = tool_parameters.get("model", "schnell") + url = FLUX_URL.get(model) + if model == "dev": + payload["model"] = "black-forest-labs/FLUX.1-dev" - response = requests.post(FLUX_URL, json=payload, headers=headers) + response = requests.post(url, json=payload, headers=headers) if response.status_code != 200: return self.create_text_message(f"Got Error Response:{response.text}") diff --git a/api/core/tools/provider/builtin/siliconflow/tools/flux.yaml b/api/core/tools/provider/builtin/siliconflow/tools/flux.yaml index 2a0698700c..d06b9bf3e1 100644 --- a/api/core/tools/provider/builtin/siliconflow/tools/flux.yaml +++ b/api/core/tools/provider/builtin/siliconflow/tools/flux.yaml @@ -6,8 +6,8 @@ identity: icon: icon.svg description: human: - en_US: Generate image via SiliconFlow's flux schnell. - llm: This tool is used to generate image from prompt via SiliconFlow's flux schnell model. + en_US: Generate image via SiliconFlow's flux model. + llm: This tool is used to generate image from prompt via SiliconFlow's flux model. parameters: - name: prompt type: string @@ -17,9 +17,24 @@ parameters: zh_Hans: 提示词 human_description: en_US: The text prompt used to generate the image. - zh_Hans: 用于生成图片的文字提示词 + zh_Hans: 建议用英文的生成图片提示词以获得更好的生成效果。 llm_description: this prompt text will be used to generate image. form: llm + - name: model + type: select + required: true + options: + - value: schnell + label: + en_US: Flux.1-schnell + - value: dev + label: + en_US: Flux.1-dev + default: schnell + label: + en_US: Choose Image Model + zh_Hans: 选择生成图片的模型 + form: form - name: image_size type: select required: true diff --git a/api/core/tools/provider/tool_provider.py b/api/core/tools/provider/tool_provider.py index bfdc161af6..f07af649ad 100644 --- a/api/core/tools/provider/tool_provider.py +++ b/api/core/tools/provider/tool_provider.py @@ -67,7 +67,7 @@ class ToolProviderController(BaseModel, ABC): # check type credential_schema = credentials_need_to_validate[credential_name] - if credential_schema in {ProviderConfig.Type.SECRET_INPUT, ProviderConfig.Type.TEXT_INPUT}: + if credential_schema.type in {ProviderConfig.Type.SECRET_INPUT, ProviderConfig.Type.TEXT_INPUT}: if not isinstance(credentials[credential_name], str): raise ToolProviderCredentialValidationError(f"credential {credential_name} should be string") diff --git a/api/core/tools/tool/dataset_retriever/dataset_multi_retriever_tool.py b/api/core/tools/tool/dataset_retriever/dataset_multi_retriever_tool.py index 6073b8e92e..ab7b40a253 100644 --- a/api/core/tools/tool/dataset_retriever/dataset_multi_retriever_tool.py +++ b/api/core/tools/tool/dataset_retriever/dataset_multi_retriever_tool.py @@ -165,7 +165,10 @@ class DatasetMultiRetrieverTool(DatasetRetrieverBaseTool): if dataset.indexing_technique == "economy": # use keyword table query documents = RetrievalService.retrieve( - retrieval_method="keyword_search", dataset_id=dataset.id, query=query, top_k=self.top_k + retrieval_method="keyword_search", + dataset_id=dataset.id, + query=query, + top_k=retrieval_model.get("top_k") or 2, ) if documents: all_documents.extend(documents) @@ -176,7 +179,7 @@ class DatasetMultiRetrieverTool(DatasetRetrieverBaseTool): retrieval_method=retrieval_model["search_method"], dataset_id=dataset.id, query=query, - top_k=self.top_k, + top_k=retrieval_model.get("top_k") or 2, score_threshold=retrieval_model.get("score_threshold", 0.0) if retrieval_model["score_threshold_enabled"] else 0.0, diff --git a/api/core/tools/utils/feishu_api_utils.py b/api/core/tools/utils/feishu_api_utils.py index 44803d7d65..ffdb06498f 100644 --- a/api/core/tools/utils/feishu_api_utils.py +++ b/api/core/tools/utils/feishu_api_utils.py @@ -76,9 +76,9 @@ class FeishuRequest: url = "https://lark-plugin-api.solutionsuite.cn/lark-plugin/document/write_document" payload = {"document_id": document_id, "content": content, "position": position} res = self._send_request(url, payload=payload) - return res.get("data") + return res - def get_document_raw_content(self, document_id: str) -> dict: + def get_document_content(self, document_id: str, mode: str, lang: int = 0) -> dict: """ API url: https://open.larkoffice.com/document/server-docs/docs/docs/docx-v1/document/raw_content Example Response: @@ -92,16 +92,18 @@ class FeishuRequest: """ # noqa: E501 params = { "document_id": document_id, + "mode": mode, + "lang": lang, } - url = "https://lark-plugin-api.solutionsuite.cn/lark-plugin/document/get_document_raw_content" + url = "https://lark-plugin-api.solutionsuite.cn/lark-plugin/document/get_document_content" res = self._send_request(url, method="get", params=params) return res.get("data").get("content") - def list_document_block(self, document_id: str, page_token: str, page_size: int = 500) -> dict: + def list_document_blocks(self, document_id: str, page_token: str, page_size: int = 500) -> dict: """ API url: https://open.larkoffice.com/document/server-docs/docs/docs/docx-v1/document/list """ - url = "https://lark-plugin-api.solutionsuite.cn/lark-plugin/document/list_document_block" + url = "https://lark-plugin-api.solutionsuite.cn/lark-plugin/document/list_document_blocks" params = { "document_id": document_id, "page_size": page_size, diff --git a/api/core/workflow/graph_engine/entities/graph.py b/api/core/workflow/graph_engine/entities/graph.py index 1d7e9158d8..1175f4af2a 100644 --- a/api/core/workflow/graph_engine/entities/graph.py +++ b/api/core/workflow/graph_engine/entities/graph.py @@ -689,23 +689,11 @@ class Graph(BaseModel): parallel_start_node_ids[graph_edge.source_node_id].append(branch_node_id) - parallel_start_node_id = None - for p_start_node_id, branch_node_ids in parallel_start_node_ids.items(): + for _, branch_node_ids in parallel_start_node_ids.items(): if set(branch_node_ids) == set(routes_node_ids.keys()): - parallel_start_node_id = p_start_node_id return True - if not parallel_start_node_id: - raise Exception("Parallel start node id not found") - - for graph_edge in reverse_edge_mapping[start_node_id]: - if ( - graph_edge.source_node_id not in all_routes_node_ids - or graph_edge.source_node_id != parallel_start_node_id - ): - return False - - return True + return False @classmethod def _is_node2_after_node1(cls, node1_id: str, node2_id: str, edge_mapping: dict[str, list[GraphEdge]]) -> bool: diff --git a/api/core/workflow/graph_engine/graph_engine.py b/api/core/workflow/graph_engine/graph_engine.py index 1db9b690ab..57e4f716fd 100644 --- a/api/core/workflow/graph_engine/graph_engine.py +++ b/api/core/workflow/graph_engine/graph_engine.py @@ -61,6 +61,9 @@ class GraphEngineThreadPool(ThreadPoolExecutor): return super().submit(fn, *args, **kwargs) + def task_done_callback(self, future): + self.submit_count -= 1 + def check_is_full(self) -> None: print(f"submit_count: {self.submit_count}, max_submit_count: {self.max_submit_count}") if self.submit_count > self.max_submit_count: @@ -426,20 +429,22 @@ class GraphEngine: ): continue - futures.append( - self.thread_pool.submit( - self._run_parallel_node, - **{ - "flask_app": current_app._get_current_object(), # type: ignore[attr-defined] - "q": q, - "parallel_id": parallel_id, - "parallel_start_node_id": edge.target_node_id, - "parent_parallel_id": in_parallel_id, - "parent_parallel_start_node_id": parallel_start_node_id, - }, - ) + future = self.thread_pool.submit( + self._run_parallel_node, + **{ + "flask_app": current_app._get_current_object(), # type: ignore[attr-defined] + "q": q, + "parallel_id": parallel_id, + "parallel_start_node_id": edge.target_node_id, + "parent_parallel_id": in_parallel_id, + "parent_parallel_start_node_id": parallel_start_node_id, + }, ) + future.add_done_callback(self.thread_pool.task_done_callback) + + futures.append(future) + succeeded_count = 0 while True: try: diff --git a/api/core/workflow/nodes/answer/answer_stream_generate_router.py b/api/core/workflow/nodes/answer/answer_stream_generate_router.py index 5e6de8fb15..bbd1f88867 100644 --- a/api/core/workflow/nodes/answer/answer_stream_generate_router.py +++ b/api/core/workflow/nodes/answer/answer_stream_generate_router.py @@ -152,6 +152,7 @@ class AnswerStreamGeneratorRouter: NodeType.ANSWER.value, NodeType.IF_ELSE.value, NodeType.QUESTION_CLASSIFIER.value, + NodeType.ITERATION.value, }: answer_dependencies[answer_node_id].append(source_node_id) else: diff --git a/api/core/workflow/nodes/iteration/iteration_node.py b/api/core/workflow/nodes/iteration/iteration_node.py index 4d944e93db..6f20745daf 100644 --- a/api/core/workflow/nodes/iteration/iteration_node.py +++ b/api/core/workflow/nodes/iteration/iteration_node.py @@ -20,11 +20,9 @@ from core.workflow.graph_engine.entities.event import ( NodeRunSucceededEvent, ) from core.workflow.graph_engine.entities.graph import Graph -from core.workflow.graph_engine.entities.run_condition import RunCondition from core.workflow.nodes.base_node import BaseNode from core.workflow.nodes.event import RunCompletedEvent, RunEvent from core.workflow.nodes.iteration.entities import IterationNodeData -from core.workflow.utils.condition.entities import Condition from models.workflow import WorkflowNodeExecutionStatus logger = logging.getLogger(__name__) @@ -68,38 +66,6 @@ class IterationNode(BaseNode): if not iteration_graph: raise ValueError("iteration graph not found") - leaf_node_ids = iteration_graph.get_leaf_node_ids() - iteration_leaf_node_ids = [] - for leaf_node_id in leaf_node_ids: - node_config = iteration_graph.node_id_config_mapping.get(leaf_node_id) - if not node_config: - continue - - leaf_node_iteration_id = node_config.get("data", {}).get("iteration_id") - if not leaf_node_iteration_id: - continue - - if leaf_node_iteration_id != self.node_id: - continue - - iteration_leaf_node_ids.append(leaf_node_id) - - # add condition of end nodes to root node - iteration_graph.add_extra_edge( - source_node_id=leaf_node_id, - target_node_id=root_node_id, - run_condition=RunCondition( - type="condition", - conditions=[ - Condition( - variable_selector=[self.node_id, "index"], - comparison_operator="<", - value=str(len(iterator_list_value)), - ) - ], - ), - ) - variable_pool = self.graph_runtime_state.variable_pool # append iteration variable (item, index) to variable pool @@ -149,91 +115,90 @@ class IterationNode(BaseNode): outputs: list[Any] = [] try: - # run workflow - rst = graph_engine.run() - for event in rst: - if isinstance(event, (BaseNodeEvent | BaseParallelBranchEvent)) and not event.in_iteration_id: - event.in_iteration_id = self.node_id + for _ in range(len(iterator_list_value)): + # run workflow + rst = graph_engine.run() + for event in rst: + if isinstance(event, (BaseNodeEvent | BaseParallelBranchEvent)) and not event.in_iteration_id: + event.in_iteration_id = self.node_id - if ( - isinstance(event, BaseNodeEvent) - and event.node_type == NodeType.ITERATION_START - and not isinstance(event, NodeRunStreamChunkEvent) - ): - continue + if ( + isinstance(event, BaseNodeEvent) + and event.node_type == NodeType.ITERATION_START + and not isinstance(event, NodeRunStreamChunkEvent) + ): + continue - if isinstance(event, NodeRunSucceededEvent): - if event.route_node_state.node_run_result: - metadata = event.route_node_state.node_run_result.metadata - if not metadata: - metadata = {} + if isinstance(event, NodeRunSucceededEvent): + if event.route_node_state.node_run_result: + metadata = event.route_node_state.node_run_result.metadata + if not metadata: + metadata = {} - if NodeRunMetadataKey.ITERATION_ID not in metadata: - metadata[NodeRunMetadataKey.ITERATION_ID] = self.node_id - metadata[NodeRunMetadataKey.ITERATION_INDEX] = variable_pool.get_any( - [self.node_id, "index"] - ) - event.route_node_state.node_run_result.metadata = metadata + if NodeRunMetadataKey.ITERATION_ID not in metadata: + metadata[NodeRunMetadataKey.ITERATION_ID] = self.node_id + metadata[NodeRunMetadataKey.ITERATION_INDEX] = variable_pool.get_any( + [self.node_id, "index"] + ) + event.route_node_state.node_run_result.metadata = metadata - yield event - - # handle iteration run result - if event.route_node_state.node_id in iteration_leaf_node_ids: - # append to iteration output variable list - current_iteration_output = variable_pool.get_any(self.node_data.output_selector) - outputs.append(current_iteration_output) - - # remove all nodes outputs from variable pool - for node_id in iteration_graph.node_ids: - variable_pool.remove_node(node_id) - - # move to next iteration - current_index = variable_pool.get([self.node_id, "index"]) - if current_index is None: - raise ValueError(f"iteration {self.node_id} current index not found") - - next_index = int(current_index.to_object()) + 1 - variable_pool.add([self.node_id, "index"], next_index) - - if next_index < len(iterator_list_value): - variable_pool.add([self.node_id, "item"], iterator_list_value[next_index]) - - yield IterationRunNextEvent( - iteration_id=self.id, - iteration_node_id=self.node_id, - iteration_node_type=self.node_type, - iteration_node_data=self.node_data, - index=next_index, - pre_iteration_output=jsonable_encoder(current_iteration_output) - if current_iteration_output - else None, - ) - elif isinstance(event, BaseGraphEvent): - if isinstance(event, GraphRunFailedEvent): - # iteration run failed - yield IterationRunFailedEvent( - iteration_id=self.id, - iteration_node_id=self.node_id, - iteration_node_type=self.node_type, - iteration_node_data=self.node_data, - start_at=start_at, - inputs=inputs, - outputs={"output": jsonable_encoder(outputs)}, - steps=len(iterator_list_value), - metadata={"total_tokens": graph_engine.graph_runtime_state.total_tokens}, - error=event.error, - ) - - yield RunCompletedEvent( - run_result=NodeRunResult( - status=WorkflowNodeExecutionStatus.FAILED, + yield event + elif isinstance(event, BaseGraphEvent): + if isinstance(event, GraphRunFailedEvent): + # iteration run failed + yield IterationRunFailedEvent( + iteration_id=self.id, + iteration_node_id=self.node_id, + iteration_node_type=self.node_type, + iteration_node_data=self.node_data, + start_at=start_at, + inputs=inputs, + outputs={"output": jsonable_encoder(outputs)}, + steps=len(iterator_list_value), + metadata={"total_tokens": graph_engine.graph_runtime_state.total_tokens}, error=event.error, ) - ) - break - else: - event = cast(InNodeEvent, event) - yield event + + yield RunCompletedEvent( + run_result=NodeRunResult( + status=WorkflowNodeExecutionStatus.FAILED, + error=event.error, + ) + ) + return + else: + event = cast(InNodeEvent, event) + yield event + + # append to iteration output variable list + current_iteration_output = variable_pool.get_any(self.node_data.output_selector) + outputs.append(current_iteration_output) + + # remove all nodes outputs from variable pool + for node_id in iteration_graph.node_ids: + variable_pool.remove_node(node_id) + + # move to next iteration + current_index = variable_pool.get([self.node_id, "index"]) + if current_index is None: + raise ValueError(f"iteration {self.node_id} current index not found") + + next_index = int(current_index.to_object()) + 1 + variable_pool.add([self.node_id, "index"], next_index) + + if next_index < len(iterator_list_value): + variable_pool.add([self.node_id, "item"], iterator_list_value[next_index]) + + yield IterationRunNextEvent( + iteration_id=self.id, + iteration_node_id=self.node_id, + iteration_node_type=self.node_type, + iteration_node_data=self.node_data, + index=next_index, + pre_iteration_output=jsonable_encoder(current_iteration_output) + if current_iteration_output + else None, + ) yield IterationRunSucceededEvent( iteration_id=self.id, diff --git a/api/extensions/ext_sentry.py b/api/extensions/ext_sentry.py index c2dc736038..e255e7eb35 100644 --- a/api/extensions/ext_sentry.py +++ b/api/extensions/ext_sentry.py @@ -5,6 +5,8 @@ from sentry_sdk.integrations.celery import CeleryIntegration from sentry_sdk.integrations.flask import FlaskIntegration from werkzeug.exceptions import HTTPException +from core.model_runtime.errors.invoke import InvokeRateLimitError + def before_send(event, hint): if "exc_info" in hint: @@ -20,7 +22,13 @@ def init_app(app): sentry_sdk.init( dsn=app.config.get("SENTRY_DSN"), integrations=[FlaskIntegration(), CeleryIntegration()], - ignore_errors=[HTTPException, ValueError, openai.APIStatusError, parse_error.defaultErrorResponse], + ignore_errors=[ + HTTPException, + ValueError, + openai.APIStatusError, + InvokeRateLimitError, + parse_error.defaultErrorResponse, + ], traces_sample_rate=app.config.get("SENTRY_TRACES_SAMPLE_RATE", 1.0), profiles_sample_rate=app.config.get("SENTRY_PROFILES_SAMPLE_RATE", 1.0), environment=app.config.get("DEPLOY_ENV"), diff --git a/api/extensions/ext_storage.py b/api/extensions/ext_storage.py index 5ce18b7292..1e6530f6f4 100644 --- a/api/extensions/ext_storage.py +++ b/api/extensions/ext_storage.py @@ -1,3 +1,4 @@ +import logging from collections.abc import Generator from typing import Union @@ -40,28 +41,56 @@ class Storage: self.storage_runner = LocalStorage(app=app) def save(self, filename, data): - self.storage_runner.save(filename, data) + try: + self.storage_runner.save(filename, data) + except Exception as e: + logging.exception("Failed to save file: %s", e) + raise e def load(self, filename: str, stream: bool = False) -> Union[bytes, Generator]: - if stream: - return self.load_stream(filename) - else: - return self.load_once(filename) + try: + if stream: + return self.load_stream(filename) + else: + return self.load_once(filename) + except Exception as e: + logging.exception("Failed to load file: %s", e) + raise e def load_once(self, filename: str) -> bytes: - return self.storage_runner.load_once(filename) + try: + return self.storage_runner.load_once(filename) + except Exception as e: + logging.exception("Failed to load_once file: %s", e) + raise e def load_stream(self, filename: str) -> Generator: - return self.storage_runner.load_stream(filename) + try: + return self.storage_runner.load_stream(filename) + except Exception as e: + logging.exception("Failed to load_stream file: %s", e) + raise e def download(self, filename, target_filepath): - self.storage_runner.download(filename, target_filepath) + try: + self.storage_runner.download(filename, target_filepath) + except Exception as e: + logging.exception("Failed to download file: %s", e) + raise e def exists(self, filename): - return self.storage_runner.exists(filename) + try: + return self.storage_runner.exists(filename) + except Exception as e: + logging.exception("Failed to check file exists: %s", e) + raise e def delete(self, filename): - return self.storage_runner.delete(filename) + try: + return self.storage_runner.delete(filename) + except Exception as e: + logging.exception("Failed to delete file: %s", e) + raise e storage = Storage() diff --git a/api/extensions/storage/aliyun_storage.py b/api/extensions/storage/aliyun_storage.py index bee237fc17..2677912aa9 100644 --- a/api/extensions/storage/aliyun_storage.py +++ b/api/extensions/storage/aliyun_storage.py @@ -31,54 +31,34 @@ class AliyunStorage(BaseStorage): ) def save(self, filename, data): - if not self.folder or self.folder.endswith("/"): - filename = self.folder + filename - else: - filename = self.folder + "/" + filename - self.client.put_object(filename, data) + self.client.put_object(self.__wrapper_folder_filename(filename), data) def load_once(self, filename: str) -> bytes: - if not self.folder or self.folder.endswith("/"): - filename = self.folder + filename - else: - filename = self.folder + "/" + filename - - with closing(self.client.get_object(filename)) as obj: + with closing(self.client.get_object(self.__wrapper_folder_filename(filename))) as obj: data = obj.read() return data def load_stream(self, filename: str) -> Generator: def generate(filename: str = filename) -> Generator: - if not self.folder or self.folder.endswith("/"): - filename = self.folder + filename - else: - filename = self.folder + "/" + filename - - with closing(self.client.get_object(filename)) as obj: + with closing(self.client.get_object(self.__wrapper_folder_filename(filename))) as obj: while chunk := obj.read(4096): yield chunk return generate() def download(self, filename, target_filepath): - if not self.folder or self.folder.endswith("/"): - filename = self.folder + filename - else: - filename = self.folder + "/" + filename - - self.client.get_object_to_file(filename, target_filepath) + self.client.get_object_to_file(self.__wrapper_folder_filename(filename), target_filepath) def exists(self, filename): - if not self.folder or self.folder.endswith("/"): - filename = self.folder + filename - else: - filename = self.folder + "/" + filename - - return self.client.object_exists(filename) + return self.client.object_exists(self.__wrapper_folder_filename(filename)) def delete(self, filename): - if not self.folder or self.folder.endswith("/"): - filename = self.folder + filename - else: - filename = self.folder + "/" + filename - self.client.delete_object(filename) + self.client.delete_object(self.__wrapper_folder_filename(filename)) + + def __wrapper_folder_filename(self, filename) -> str: + if self.folder: + if self.folder.endswith("/"): + filename = self.folder + filename + else: + filename = self.folder + "/" + filename + return filename diff --git a/api/poetry.lock b/api/poetry.lock index 191db600e4..28c688cc9c 100644 --- a/api/poetry.lock +++ b/api/poetry.lock @@ -2296,18 +2296,18 @@ files = [ [[package]] name = "duckduckgo-search" -version = "6.2.11" +version = "6.2.12" description = "Search for words, documents, images, news, maps and text translation using the DuckDuckGo.com search engine." optional = false python-versions = ">=3.8" files = [ - {file = "duckduckgo_search-6.2.11-py3-none-any.whl", hash = "sha256:6fb7069b79e8928f487001de6859034ade19201bdcd257ec198802430e374bfe"}, - {file = "duckduckgo_search-6.2.11.tar.gz", hash = "sha256:6b6ef1b552c5e67f23e252025d2504caf6f9fc14f70e86c6dd512200f386c673"}, + {file = "duckduckgo_search-6.2.12-py3-none-any.whl", hash = "sha256:0d379c1f845b632a41553efb13d571788f19ad289229e641a27b5710d92097a6"}, + {file = "duckduckgo_search-6.2.12.tar.gz", hash = "sha256:04f9f1459763668d268344c7a32d943173d0e060dad53a5c2df4b4d3ca9a74cf"}, ] [package.dependencies] click = ">=8.1.7" -primp = ">=0.6.1" +primp = ">=0.6.2" [package.extras] dev = ["mypy (>=1.11.1)", "pytest (>=8.3.1)", "pytest-asyncio (>=0.23.8)", "ruff (>=0.6.1)"] @@ -6356,19 +6356,19 @@ dill = ["dill (>=0.3.8)"] [[package]] name = "primp" -version = "0.6.1" +version = "0.6.2" description = "HTTP client that can impersonate web browsers, mimicking their headers and `TLS/JA3/JA4/HTTP2` fingerprints" optional = false python-versions = ">=3.8" files = [ - {file = "primp-0.6.1-cp38-abi3-macosx_10_12_x86_64.whl", hash = "sha256:60cfe95e0bdf154b0f9036d38acaddc9aef02d6723ed125839b01449672d3946"}, - {file = "primp-0.6.1-cp38-abi3-macosx_11_0_arm64.whl", hash = "sha256:e1e92433ecf32639f9e800bc3a5d58b03792bdec99421b7fb06500e2fae63c85"}, - {file = "primp-0.6.1-cp38-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:6e02353f13f07fb5a6f91df9e2f4d8ec9f41312de95088744dce1c9729a3865d"}, - {file = "primp-0.6.1-cp38-abi3-manylinux_2_34_aarch64.whl", hash = "sha256:c5a2ccfdf488b17be225a529a31e2b22724b2e22fba8e1ae168a222f857c2dc0"}, - {file = "primp-0.6.1-cp38-abi3-musllinux_1_2_aarch64.whl", hash = "sha256:f335c2ace907800a23bbb7bc6e15acc7fff659b86a2d5858817f6ed79cea07cf"}, - {file = "primp-0.6.1-cp38-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:5dc15bd9d47ded7bc356fcb5d8321972dcbeba18e7d3b7250e12bb7365447b2b"}, - {file = "primp-0.6.1-cp38-abi3-win_amd64.whl", hash = "sha256:eebf0412ebba4089547b16b97b765d83f69f1433d811bb02b02cdcdbca20f672"}, - {file = "primp-0.6.1.tar.gz", hash = "sha256:64b3c12e3d463a887518811c46f3ec37cca02e6af1ddf1287e548342de436301"}, + {file = "primp-0.6.2-cp38-abi3-macosx_10_12_x86_64.whl", hash = "sha256:4a35d441462a55d9a9525bf170e2ffd2fcb3db6039b23e802859fa22c18cdd51"}, + {file = "primp-0.6.2-cp38-abi3-macosx_11_0_arm64.whl", hash = "sha256:f67ccade95bdbca3cf9b96b93aa53f9617d85ddbf988da4e9c523aa785fd2d54"}, + {file = "primp-0.6.2-cp38-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:8074b93befaf36567e4cf3d4a1a8cd6ab9cc6e4dd4ff710650678daa405aee71"}, + {file = "primp-0.6.2-cp38-abi3-manylinux_2_34_aarch64.whl", hash = "sha256:7d3e2a3f8c6262e9b883651b79c4ff2b7677a76f47293a139f541c9ea333ce3b"}, + {file = "primp-0.6.2-cp38-abi3-musllinux_1_2_aarch64.whl", hash = "sha256:a460ea389371c6d04839b4b50b5805d99da8ebe281a2e8b534d27377c6d44f0e"}, + {file = "primp-0.6.2-cp38-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:5b6b27e89d3c05c811aff0e4fde7a36d6957b15b3112f4ce28b6b99e8ca1e725"}, + {file = "primp-0.6.2-cp38-abi3-win_amd64.whl", hash = "sha256:1006a40a85f88a4c5222094813a1ebc01f85a63e9a33d2c443288c0720bed321"}, + {file = "primp-0.6.2.tar.gz", hash = "sha256:5a96a6b65195a8a989157e67d23bd171c49be238654e02bdf1b1fda36cbcc068"}, ] [package.extras] diff --git a/api/pyproject.toml b/api/pyproject.toml index 166ddcec50..8c10f1dad9 100644 --- a/api/pyproject.toml +++ b/api/pyproject.toml @@ -209,7 +209,7 @@ zhipuai = "1.0.7" # Before adding new dependency, consider place it in alphabet order (a-z) and suitable group. ############################################################ -# Related transparent dependencies with pinned verion +# Related transparent dependencies with pinned version # required by main implementations ############################################################ azure-ai-ml = "^1.19.0" diff --git a/docker/.env.example b/docker/.env.example index 7e4430a37d..c892c15636 100644 --- a/docker/.env.example +++ b/docker/.env.example @@ -568,6 +568,13 @@ SSRF_PROXY_HTTP_URL=http://ssrf_proxy:3128 # SSRF Proxy server HTTPS URL SSRF_PROXY_HTTPS_URL=http://ssrf_proxy:3128 +# ------------------------------ +# Environment Variables for web Service +# ------------------------------ + +# The timeout for the text generation in millisecond +TEXT_GENERATION_TIMEOUT_MS=60000 + # ------------------------------ # Environment Variables for db Service # ------------------------------ diff --git a/docker/docker-compose.middleware.yaml b/docker/docker-compose.middleware.yaml index 00faa2960a..251c62fee1 100644 --- a/docker/docker-compose.middleware.yaml +++ b/docker/docker-compose.middleware.yaml @@ -89,6 +89,7 @@ services: weaviate: image: semitechnologies/weaviate:1.19.0 profiles: + - "" - weaviate restart: always volumes: diff --git a/docker/docker-compose.yaml b/docker/docker-compose.yaml index d080731a28..0fbc695177 100644 --- a/docker/docker-compose.yaml +++ b/docker/docker-compose.yaml @@ -75,7 +75,7 @@ x-shared-env: &shared-api-worker-env ALIYUN_OSS_ENDPOINT: ${ALIYUN_OSS_ENDPOINT:-} ALIYUN_OSS_REGION: ${ALIYUN_OSS_REGION:-} ALIYUN_OSS_AUTH_VERSION: ${ALIYUN_OSS_AUTH_VERSION:-v4} - ALIYUN_OSS_PATHS: ${ALIYUN_OSS_PATH:-} + ALIYUN_OSS_PATH: ${ALIYUN_OSS_PATH:-} TENCENT_COS_BUCKET_NAME: ${TENCENT_COS_BUCKET_NAME:-} TENCENT_COS_SECRET_KEY: ${TENCENT_COS_SECRET_KEY:-} TENCENT_COS_SECRET_ID: ${TENCENT_COS_SECRET_ID:-} @@ -254,6 +254,7 @@ services: APP_API_URL: ${APP_API_URL:-} SENTRY_DSN: ${WEB_SENTRY_DSN:-} NEXT_TELEMETRY_DISABLED: ${NEXT_TELEMETRY_DISABLED:-0} + TEXT_GENERATION_TIMEOUT_MS: ${TEXT_GENERATION_TIMEOUT_MS:-60000} # The postgres database. db: diff --git a/sdks/python-client/dify_client/client.py b/sdks/python-client/dify_client/client.py index a8da1d7cae..2be079bdf3 100644 --- a/sdks/python-client/dify_client/client.py +++ b/sdks/python-client/dify_client/client.py @@ -1,3 +1,4 @@ +import json import requests @@ -131,3 +132,284 @@ class WorkflowClient(DifyClient): def stop(self, task_id, user): data = {"user": user} return self._send_request("POST", f"/workflows/tasks/{task_id}/stop", data) + + def get_result(self, workflow_run_id): + return self._send_request("GET", f"/workflows/run/{workflow_run_id}") + + + +class KnowledgeBaseClient(DifyClient): + + def __init__(self, api_key, base_url: str = 'https://api.dify.ai/v1', dataset_id: str = None): + """ + Construct a KnowledgeBaseClient object. + + Args: + api_key (str): API key of Dify. + base_url (str, optional): Base URL of Dify API. Defaults to 'https://api.dify.ai/v1'. + dataset_id (str, optional): ID of the dataset. Defaults to None. You don't need this if you just want to + create a new dataset. or list datasets. otherwise you need to set this. + """ + super().__init__( + api_key=api_key, + base_url=base_url + ) + self.dataset_id = dataset_id + + def _get_dataset_id(self): + if self.dataset_id is None: + raise ValueError("dataset_id is not set") + return self.dataset_id + + def create_dataset(self, name: str, **kwargs): + return self._send_request('POST', '/datasets', {'name': name}, **kwargs) + + def list_datasets(self, page: int = 1, page_size: int = 20, **kwargs): + return self._send_request('GET', f'/datasets?page={page}&limit={page_size}', **kwargs) + + def create_document_by_text(self, name, text, extra_params: dict = None, **kwargs): + """ + Create a document by text. + + :param name: Name of the document + :param text: Text content of the document + :param extra_params: extra parameters pass to the API, such as indexing_technique, process_rule. (optional) + e.g. + { + 'indexing_technique': 'high_quality', + 'process_rule': { + 'rules': { + 'pre_processing_rules': [ + {'id': 'remove_extra_spaces', 'enabled': True}, + {'id': 'remove_urls_emails', 'enabled': True} + ], + 'segmentation': { + 'separator': '\n', + 'max_tokens': 500 + } + }, + 'mode': 'custom' + } + } + :return: Response from the API + """ + data = { + 'indexing_technique': 'high_quality', + 'process_rule': { + 'mode': 'automatic' + }, + 'name': name, + 'text': text + } + if extra_params is not None and isinstance(extra_params, dict): + data.update(extra_params) + url = f"/datasets/{self._get_dataset_id()}/document/create_by_text" + return self._send_request("POST", url, json=data, **kwargs) + + def update_document_by_text(self, document_id, name, text, extra_params: dict = None, **kwargs): + """ + Update a document by text. + + :param document_id: ID of the document + :param name: Name of the document + :param text: Text content of the document + :param extra_params: extra parameters pass to the API, such as indexing_technique, process_rule. (optional) + e.g. + { + 'indexing_technique': 'high_quality', + 'process_rule': { + 'rules': { + 'pre_processing_rules': [ + {'id': 'remove_extra_spaces', 'enabled': True}, + {'id': 'remove_urls_emails', 'enabled': True} + ], + 'segmentation': { + 'separator': '\n', + 'max_tokens': 500 + } + }, + 'mode': 'custom' + } + } + :return: Response from the API + """ + data = { + 'name': name, + 'text': text + } + if extra_params is not None and isinstance(extra_params, dict): + data.update(extra_params) + url = f"/datasets/{self._get_dataset_id()}/documents/{document_id}/update_by_text" + return self._send_request("POST", url, json=data, **kwargs) + + def create_document_by_file(self, file_path, original_document_id=None, extra_params: dict = None): + """ + Create a document by file. + + :param file_path: Path to the file + :param original_document_id: pass this ID if you want to replace the original document (optional) + :param extra_params: extra parameters pass to the API, such as indexing_technique, process_rule. (optional) + e.g. + { + 'indexing_technique': 'high_quality', + 'process_rule': { + 'rules': { + 'pre_processing_rules': [ + {'id': 'remove_extra_spaces', 'enabled': True}, + {'id': 'remove_urls_emails', 'enabled': True} + ], + 'segmentation': { + 'separator': '\n', + 'max_tokens': 500 + } + }, + 'mode': 'custom' + } + } + :return: Response from the API + """ + files = {"file": open(file_path, "rb")} + data = { + 'process_rule': { + 'mode': 'automatic' + }, + 'indexing_technique': 'high_quality' + } + if extra_params is not None and isinstance(extra_params, dict): + data.update(extra_params) + if original_document_id is not None: + data['original_document_id'] = original_document_id + url = f"/datasets/{self._get_dataset_id()}/document/create_by_file" + return self._send_request_with_files("POST", url, {"data": json.dumps(data)}, files) + + def update_document_by_file(self, document_id, file_path, extra_params: dict = None): + """ + Update a document by file. + + :param document_id: ID of the document + :param file_path: Path to the file + :param extra_params: extra parameters pass to the API, such as indexing_technique, process_rule. (optional) + e.g. + { + 'indexing_technique': 'high_quality', + 'process_rule': { + 'rules': { + 'pre_processing_rules': [ + {'id': 'remove_extra_spaces', 'enabled': True}, + {'id': 'remove_urls_emails', 'enabled': True} + ], + 'segmentation': { + 'separator': '\n', + 'max_tokens': 500 + } + }, + 'mode': 'custom' + } + } + :return: + """ + files = {"file": open(file_path, "rb")} + data = {} + if extra_params is not None and isinstance(extra_params, dict): + data.update(extra_params) + url = f"/datasets/{self._get_dataset_id()}/documents/{document_id}/update_by_file" + return self._send_request_with_files("POST", url, {"data": json.dumps(data)}, files) + + def batch_indexing_status(self, batch_id: str, **kwargs): + """ + Get the status of the batch indexing. + + :param batch_id: ID of the batch uploading + :return: Response from the API + """ + url = f"/datasets/{self._get_dataset_id()}/documents/{batch_id}/indexing-status" + return self._send_request("GET", url, **kwargs) + + def delete_dataset(self): + """ + Delete this dataset. + + :return: Response from the API + """ + url = f"/datasets/{self._get_dataset_id()}" + return self._send_request("DELETE", url) + + def delete_document(self, document_id): + """ + Delete a document. + + :param document_id: ID of the document + :return: Response from the API + """ + url = f"/datasets/{self._get_dataset_id()}/documents/{document_id}" + return self._send_request("DELETE", url) + + def list_documents(self, page: int = None, page_size: int = None, keyword: str = None, **kwargs): + """ + Get a list of documents in this dataset. + + :return: Response from the API + """ + params = {} + if page is not None: + params['page'] = page + if page_size is not None: + params['limit'] = page_size + if keyword is not None: + params['keyword'] = keyword + url = f"/datasets/{self._get_dataset_id()}/documents" + return self._send_request("GET", url, params=params, **kwargs) + + def add_segments(self, document_id, segments, **kwargs): + """ + Add segments to a document. + + :param document_id: ID of the document + :param segments: List of segments to add, example: [{"content": "1", "answer": "1", "keyword": ["a"]}] + :return: Response from the API + """ + data = {"segments": segments} + url = f"/datasets/{self._get_dataset_id()}/documents/{document_id}/segments" + return self._send_request("POST", url, json=data, **kwargs) + + def query_segments(self, document_id, keyword: str = None, status: str = None, **kwargs): + """ + Query segments in this document. + + :param document_id: ID of the document + :param keyword: query keyword, optional + :param status: status of the segment, optional, e.g. completed + """ + url = f"/datasets/{self._get_dataset_id()}/documents/{document_id}/segments" + params = {} + if keyword is not None: + params['keyword'] = keyword + if status is not None: + params['status'] = status + if "params" in kwargs: + params.update(kwargs["params"]) + return self._send_request("GET", url, params=params, **kwargs) + + def delete_document_segment(self, document_id, segment_id): + """ + Delete a segment from a document. + + :param document_id: ID of the document + :param segment_id: ID of the segment + :return: Response from the API + """ + url = f"/datasets/{self._get_dataset_id()}/documents/{document_id}/segments/{segment_id}" + return self._send_request("DELETE", url) + + def update_document_segment(self, document_id, segment_id, segment_data, **kwargs): + """ + Update a segment in a document. + + :param document_id: ID of the document + :param segment_id: ID of the segment + :param segment_data: Data of the segment, example: {"content": "1", "answer": "1", "keyword": ["a"], "enabled": True} + :return: Response from the API + """ + data = {"segment": segment_data} + url = f"/datasets/{self._get_dataset_id()}/documents/{document_id}/segments/{segment_id}" + return self._send_request("POST", url, json=data, **kwargs) diff --git a/sdks/python-client/setup.py b/sdks/python-client/setup.py index e74748377e..bb8ca46d97 100644 --- a/sdks/python-client/setup.py +++ b/sdks/python-client/setup.py @@ -5,7 +5,7 @@ with open("README.md", "r", encoding="utf-8") as fh: setup( name="dify-client", - version="0.1.10", + version="0.1.12", author="Dify", author_email="hello@dify.ai", description="A package for interacting with the Dify Service-API", diff --git a/sdks/python-client/tests/test_client.py b/sdks/python-client/tests/test_client.py index 5259d082ca..301e733b6b 100644 --- a/sdks/python-client/tests/test_client.py +++ b/sdks/python-client/tests/test_client.py @@ -1,10 +1,157 @@ import os +import time import unittest -from dify_client.client import ChatClient, CompletionClient, DifyClient +from dify_client.client import ChatClient, CompletionClient, DifyClient, KnowledgeBaseClient API_KEY = os.environ.get("API_KEY") APP_ID = os.environ.get("APP_ID") +API_BASE_URL = os.environ.get("API_BASE_URL", "https://api.dify.ai/v1") +FILE_PATH_BASE = os.path.dirname(__file__) + + +class TestKnowledgeBaseClient(unittest.TestCase): + def setUp(self): + self.knowledge_base_client = KnowledgeBaseClient(API_KEY, base_url=API_BASE_URL) + self.README_FILE_PATH = os.path.abspath(os.path.join(FILE_PATH_BASE, "../README.md")) + self.dataset_id = None + self.document_id = None + self.segment_id = None + self.batch_id = None + + def _get_dataset_kb_client(self): + self.assertIsNotNone(self.dataset_id) + return KnowledgeBaseClient(API_KEY, base_url=API_BASE_URL, dataset_id=self.dataset_id) + + def test_001_create_dataset(self): + response = self.knowledge_base_client.create_dataset(name="test_dataset") + data = response.json() + self.assertIn("id", data) + self.dataset_id = data["id"] + self.assertEqual("test_dataset", data["name"]) + + # the following tests require to be executed in order because they use + # the dataset/document/segment ids from the previous test + self._test_002_list_datasets() + self._test_003_create_document_by_text() + time.sleep(1) + self._test_004_update_document_by_text() + # self._test_005_batch_indexing_status() + time.sleep(1) + self._test_006_update_document_by_file() + time.sleep(1) + self._test_007_list_documents() + self._test_008_delete_document() + self._test_009_create_document_by_file() + time.sleep(1) + self._test_010_add_segments() + self._test_011_query_segments() + self._test_012_update_document_segment() + self._test_013_delete_document_segment() + self._test_014_delete_dataset() + + def _test_002_list_datasets(self): + response = self.knowledge_base_client.list_datasets() + data = response.json() + self.assertIn("data", data) + self.assertIn("total", data) + + def _test_003_create_document_by_text(self): + client = self._get_dataset_kb_client() + response = client.create_document_by_text("test_document", "test_text") + data = response.json() + self.assertIn("document", data) + self.document_id = data["document"]["id"] + self.batch_id = data["batch"] + + def _test_004_update_document_by_text(self): + client = self._get_dataset_kb_client() + self.assertIsNotNone(self.document_id) + response = client.update_document_by_text(self.document_id, "test_document_updated", "test_text_updated") + data = response.json() + self.assertIn("document", data) + self.assertIn("batch", data) + self.batch_id = data["batch"] + + def _test_005_batch_indexing_status(self): + client = self._get_dataset_kb_client() + response = client.batch_indexing_status(self.batch_id) + data = response.json() + self.assertEqual(response.status_code, 200) + + def _test_006_update_document_by_file(self): + client = self._get_dataset_kb_client() + self.assertIsNotNone(self.document_id) + response = client.update_document_by_file(self.document_id, self.README_FILE_PATH) + data = response.json() + self.assertIn("document", data) + self.assertIn("batch", data) + self.batch_id = data["batch"] + + def _test_007_list_documents(self): + client = self._get_dataset_kb_client() + response = client.list_documents() + data = response.json() + self.assertIn("data", data) + + def _test_008_delete_document(self): + client = self._get_dataset_kb_client() + self.assertIsNotNone(self.document_id) + response = client.delete_document(self.document_id) + data = response.json() + self.assertIn("result", data) + self.assertEqual("success", data["result"]) + + def _test_009_create_document_by_file(self): + client = self._get_dataset_kb_client() + response = client.create_document_by_file(self.README_FILE_PATH) + data = response.json() + self.assertIn("document", data) + self.document_id = data["document"]["id"] + self.batch_id = data["batch"] + + def _test_010_add_segments(self): + client = self._get_dataset_kb_client() + response = client.add_segments(self.document_id, [ + {"content": "test text segment 1"} + ]) + data = response.json() + self.assertIn("data", data) + self.assertGreater(len(data["data"]), 0) + segment = data["data"][0] + self.segment_id = segment["id"] + + def _test_011_query_segments(self): + client = self._get_dataset_kb_client() + response = client.query_segments(self.document_id) + data = response.json() + self.assertIn("data", data) + self.assertGreater(len(data["data"]), 0) + + def _test_012_update_document_segment(self): + client = self._get_dataset_kb_client() + self.assertIsNotNone(self.segment_id) + response = client.update_document_segment(self.document_id, self.segment_id, + {"content": "test text segment 1 updated"} + ) + data = response.json() + self.assertIn("data", data) + self.assertGreater(len(data["data"]), 0) + segment = data["data"] + self.assertEqual("test text segment 1 updated", segment["content"]) + + def _test_013_delete_document_segment(self): + client = self._get_dataset_kb_client() + self.assertIsNotNone(self.segment_id) + response = client.delete_document_segment(self.document_id, self.segment_id) + data = response.json() + self.assertIn("result", data) + self.assertEqual("success", data["result"]) + + def _test_014_delete_dataset(self): + client = self._get_dataset_kb_client() + response = client.delete_dataset() + self.assertEqual(204, response.status_code) class TestChatClient(unittest.TestCase): diff --git a/web/.env.example b/web/.env.example index 3045cef2f9..8e254082b3 100644 --- a/web/.env.example +++ b/web/.env.example @@ -19,3 +19,6 @@ NEXT_TELEMETRY_DISABLED=1 # Disable Upload Image as WebApp icon default is false NEXT_PUBLIC_UPLOAD_IMAGE_AS_ICON=false + +# The timeout for the text generation in millisecond +NEXT_PUBLIC_TEXT_GENERATION_TIMEOUT_MS=60000 diff --git a/web/.husky/pre-commit b/web/.husky/pre-commit index 6df8b24b61..d9290e1853 100755 --- a/web/.husky/pre-commit +++ b/web/.husky/pre-commit @@ -51,5 +51,32 @@ if $web_modified; then echo "Running ESLint on web module" cd ./web || exit 1 npx lint-staged + + echo "Running unit tests check" + modified_files=$(git diff --cached --name-only -- utils | grep -v '\.spec\.ts$' || true) + + if [ -n "$modified_files" ]; then + for file in $modified_files; do + test_file="${file%.*}.spec.ts" + echo "Checking for test file: $test_file" + + # check if the test file exists + if [ -f "../$test_file" ]; then + echo "Detected changes in $file, running corresponding unit tests..." + npm run test "../$test_file" + + if [ $? -ne 0 ]; then + echo "Unit tests failed. Please fix the errors before committing." + exit 1 + fi + echo "Unit tests for $file passed." + else + echo "Warning: $file does not have a corresponding test file." + fi + + done + echo "All unit tests for modified web/utils files have passed." + fi + cd ../ fi diff --git a/web/README.md b/web/README.md index 867d822e27..a84ef21007 100644 --- a/web/README.md +++ b/web/README.md @@ -18,6 +18,10 @@ yarn install --frozen-lockfile Then, configure the environment variables. Create a file named `.env.local` in the current directory and copy the contents from `.env.example`. Modify the values of these environment variables according to your requirements: +```bash +cp .env.example .env.local +``` + ``` # For production release, change this to PRODUCTION NEXT_PUBLIC_DEPLOY_ENV=DEVELOPMENT @@ -78,7 +82,7 @@ If your IDE is VSCode, rename `web/.vscode/settings.example.json` to `web/.vscod We start to use [Jest](https://jestjs.io/) and [React Testing Library](https://testing-library.com/docs/react-testing-library/intro/) for Unit Testing. -You can create a test file with a suffix of `.spec` beside the file that to be tested. For example, if you want to test a file named `util.ts`. The test file name should be `util.spec.ts`. +You can create a test file with a suffix of `.spec` beside the file that to be tested. For example, if you want to test a file named `util.ts`. The test file name should be `util.spec.ts`. Run test: diff --git a/web/app/(commonLayout)/app/(appDetailLayout)/[appId]/layout.tsx b/web/app/(commonLayout)/app/(appDetailLayout)/[appId]/layout.tsx index e728749b85..96ee874d53 100644 --- a/web/app/(commonLayout)/app/(appDetailLayout)/[appId]/layout.tsx +++ b/web/app/(commonLayout)/app/(appDetailLayout)/[appId]/layout.tsx @@ -109,6 +109,11 @@ const AppDetailLayout: FC = (props) => { setAppDetail() fetchAppDetail({ url: '/apps', id: appId }).then((res) => { // redirection + const canIEditApp = isCurrentWorkspaceEditor + if (!canIEditApp && (pathname.endsWith('configuration') || pathname.endsWith('workflow') || pathname.endsWith('logs'))) { + router.replace(`/app/${appId}/overview`) + return + } if ((res.mode === 'workflow' || res.mode === 'advanced-chat') && (pathname).endsWith('configuration')) { router.replace(`/app/${appId}/workflow`) } @@ -118,7 +123,7 @@ const AppDetailLayout: FC = (props) => { else { setAppDetail({ ...res, enable_sso: false }) setNavigation(getNavigations(appId, isCurrentWorkspaceEditor, res.mode)) - if (systemFeatures.enable_web_sso_switch_component) { + if (systemFeatures.enable_web_sso_switch_component && canIEditApp) { fetchAppSSO({ appId }).then((ssoRes) => { setAppDetail({ ...res, enable_sso: ssoRes.enabled }) }) @@ -128,7 +133,7 @@ const AppDetailLayout: FC = (props) => { if (e.status === 404) router.replace('/apps') }) - }, [appId, isCurrentWorkspaceEditor, systemFeatures]) + }, [appId, isCurrentWorkspaceEditor, systemFeatures, getNavigations, pathname, router, setAppDetail]) useUnmount(() => { setAppDetail() diff --git a/web/app/components/app/configuration/toolbox/annotation/annotation-ctrl-btn/index.tsx b/web/app/components/app/configuration/toolbox/annotation/annotation-ctrl-btn/index.tsx index 111c380afc..809b907d62 100644 --- a/web/app/components/app/configuration/toolbox/annotation/annotation-ctrl-btn/index.tsx +++ b/web/app/components/app/configuration/toolbox/annotation/annotation-ctrl-btn/index.tsx @@ -73,7 +73,7 @@ const CacheCtrlBtn: FC = ({ setShowModal(false) } return ( -
+
{cached ? ( @@ -101,7 +101,6 @@ const CacheCtrlBtn: FC = ({ ? (
= ({ }
{/* Panel Header */} -
+
{isChatMode ? t('appLog.detail.conversationId') : t('appLog.detail.time')}
{isChatMode && ( @@ -725,7 +725,7 @@ const ConversationList: FC = ({ logs, appDetail, onRefresh }) onClose={onCloseDrawer} mask={isMobile} footer={null} - panelClassname='mt-16 mx-2 sm:mr-2 mb-4 !p-0 !max-w-[640px] rounded-xl' + panelClassname='mt-16 mx-2 sm:mr-2 mb-4 !p-0 !max-w-[640px] rounded-xl bg-background-gradient-bg-fill-chat-bg-1' > = ({ onSave, }) => { const systemFeatures = useContextSelector(AppContext, state => state.systemFeatures) + const { isCurrentWorkspaceEditor } = useAppContext() const { notify } = useToastContext() const [isShowMore, setIsShowMore] = useState(false) const { @@ -265,7 +266,7 @@ const SettingsModal: FC = ({ } asChild={false} > - setInputInfo({ ...inputInfo, enable_sso: v })}> + setInputInfo({ ...inputInfo, enable_sso: v })}>

{t(`${prefixSettings}.sso.description`)}

diff --git a/web/app/components/base/chat/chat/answer/index.tsx b/web/app/components/base/chat/chat/answer/index.tsx index 270cd553c2..5fe2a7bad5 100644 --- a/web/app/components/base/chat/chat/answer/index.tsx +++ b/web/app/components/base/chat/chat/answer/index.tsx @@ -8,7 +8,6 @@ import type { ChatConfig, ChatItem, } from '../../types' -import { useChatContext } from '../context' import Operation from './operation' import AgentContent from './agent-content' import BasicContent from './basic-content' @@ -16,13 +15,13 @@ import SuggestedQuestions from './suggested-questions' import More from './more' import WorkflowProcess from './workflow-process' import { AnswerTriangle } from '@/app/components/base/icons/src/vender/solid/general' -import { MessageFast } from '@/app/components/base/icons/src/vender/solid/communication' import LoadingAnim from '@/app/components/base/chat/chat/loading-anim' import Citation from '@/app/components/base/chat/chat/citation' import { EditTitle } from '@/app/components/app/annotation/edit-annotation-modal/edit-item' import type { Emoji } from '@/app/components/tools/types' import type { AppData } from '@/models/share' import AnswerIcon from '@/app/components/base/answer-icon' +import cn from '@/utils/classnames' type AnswerProps = { item: ChatItem @@ -61,26 +60,24 @@ const Answer: FC = ({ } = item const hasAgentThoughts = !!agent_thoughts?.length - const [containerWidth] = useState(0) + const [containerWidth, setContainerWidth] = useState(0) const [contentWidth, setContentWidth] = useState(0) const containerRef = useRef(null) const contentRef = useRef(null) - const { - config: chatContextConfig, - } = useChatContext() + const getContainerWidth = () => { + if (containerRef.current) + setContainerWidth(containerRef.current?.clientWidth + 16) + } + useEffect(() => { + getContainerWidth() + }, []) - const voiceRef = useRef(chatContextConfig?.text_to_speech?.voice) const getContentWidth = () => { if (contentRef.current) setContentWidth(contentRef.current?.clientWidth) } - useEffect(() => { - voiceRef.current = chatContextConfig?.text_to_speech?.voice - } - , [chatContextConfig?.text_to_speech?.voice]) - useEffect(() => { if (!responding) getContentWidth() @@ -89,36 +86,20 @@ const Answer: FC = ({ return (
- { - answerIcon || - } - { - responding && ( -
- -
- ) - } + {answerIcon || } + {responding && ( +
+ +
+ )}
-
+
- {annotation?.id && ( -
-
- -
-
- )} { !responding && ( = ({ /> )} { - !positionRight && annotation?.id && ( + annotation?.id && (
diff --git a/web/app/components/base/chat/chat/index.tsx b/web/app/components/base/chat/chat/index.tsx index 65e49eff67..68194193c4 100644 --- a/web/app/components/base/chat/chat/index.tsx +++ b/web/app/components/base/chat/chat/index.tsx @@ -109,9 +109,9 @@ const Chat: FC = ({ const userScrolledRef = useRef(false) const handleScrollToBottom = useCallback(() => { - if (chatContainerRef.current && !userScrolledRef.current) + if (chatList.length > 1 && chatContainerRef.current && !userScrolledRef.current) chatContainerRef.current.scrollTop = chatContainerRef.current.scrollHeight - }, []) + }, [chatList.length]) const handleWindowResize = useCallback(() => { if (chatContainerRef.current) diff --git a/web/app/components/base/image-uploader/image-preview.tsx b/web/app/components/base/image-uploader/image-preview.tsx index 41f29fda2e..e5bd4c1bbc 100644 --- a/web/app/components/base/image-uploader/image-preview.tsx +++ b/web/app/components/base/image-uploader/image-preview.tsx @@ -1,26 +1,42 @@ import type { FC } from 'react' -import { useRef } from 'react' +import React, { useCallback, useEffect, useRef, useState } from 'react' import { t } from 'i18next' import { createPortal } from 'react-dom' -import { RiCloseLine, RiExternalLinkLine } from '@remixicon/react' +import { RiAddBoxLine, RiCloseLine, RiDownloadCloud2Line, RiFileCopyLine, RiZoomInLine, RiZoomOutLine } from '@remixicon/react' import Tooltip from '@/app/components/base/tooltip' -import { randomString } from '@/utils' +import Toast from '@/app/components/base/toast' type ImagePreviewProps = { url: string title: string onCancel: () => void } + +const isBase64 = (str: string): boolean => { + try { + return btoa(atob(str)) === str + } + catch (err) { + return false + } +} + const ImagePreview: FC = ({ url, title, onCancel, }) => { - const selector = useRef(`copy-tooltip-${randomString(4)}`) + const [scale, setScale] = useState(1) + const [position, setPosition] = useState({ x: 0, y: 0 }) + const [isDragging, setIsDragging] = useState(false) + const imgRef = useRef(null) + const dragStartRef = useRef({ x: 0, y: 0 }) + const [isCopied, setIsCopied] = useState(false) + const containerRef = useRef(null) const openInNewTab = () => { // Open in a new window, considering the case when the page is inside an iframe - if (url.startsWith('http')) { + if (url.startsWith('http') || url.startsWith('https')) { window.open(url, '_blank') } else if (url.startsWith('data:image')) { @@ -29,34 +45,224 @@ const ImagePreview: FC = ({ win?.document.write(`${title}`) } else { - console.error('Unable to open image', url) + Toast.notify({ + type: 'error', + message: `Unable to open image: ${url}`, + }) + } + } + const downloadImage = () => { + // Open in a new window, considering the case when the page is inside an iframe + if (url.startsWith('http') || url.startsWith('https')) { + const a = document.createElement('a') + a.href = url + a.download = title + a.click() + } + else if (url.startsWith('data:image')) { + // Base64 image + const a = document.createElement('a') + a.href = url + a.download = title + a.click() + } + else { + Toast.notify({ + type: 'error', + message: `Unable to open image: ${url}`, + }) } } + const zoomIn = () => { + setScale(prevScale => Math.min(prevScale * 1.2, 15)) + } + + const zoomOut = () => { + setScale((prevScale) => { + const newScale = Math.max(prevScale / 1.2, 0.5) + if (newScale === 1) + setPosition({ x: 0, y: 0 }) // Reset position when fully zoomed out + + return newScale + }) + } + + const imageTobase64ToBlob = (base64: string, type = 'image/png'): Blob => { + const byteCharacters = atob(base64) + const byteArrays = [] + + for (let offset = 0; offset < byteCharacters.length; offset += 512) { + const slice = byteCharacters.slice(offset, offset + 512) + const byteNumbers = new Array(slice.length) + for (let i = 0; i < slice.length; i++) + byteNumbers[i] = slice.charCodeAt(i) + + const byteArray = new Uint8Array(byteNumbers) + byteArrays.push(byteArray) + } + + return new Blob(byteArrays, { type }) + } + + const imageCopy = useCallback(() => { + const shareImage = async () => { + try { + const base64Data = url.split(',')[1] + const blob = imageTobase64ToBlob(base64Data, 'image/png') + + await navigator.clipboard.write([ + new ClipboardItem({ + [blob.type]: blob, + }), + ]) + setIsCopied(true) + + Toast.notify({ + type: 'success', + message: t('common.operation.imageCopied'), + }) + } + catch (err) { + console.error('Failed to copy image:', err) + + const link = document.createElement('a') + link.href = url + link.download = `${title}.png` + document.body.appendChild(link) + link.click() + document.body.removeChild(link) + + Toast.notify({ + type: 'info', + message: t('common.operation.imageDownloaded'), + }) + } + } + shareImage() + }, [title, url]) + + const handleWheel = useCallback((e: React.WheelEvent) => { + if (e.deltaY < 0) + zoomIn() + else + zoomOut() + }, []) + + const handleMouseDown = useCallback((e: React.MouseEvent) => { + if (scale > 1) { + setIsDragging(true) + dragStartRef.current = { x: e.clientX - position.x, y: e.clientY - position.y } + } + }, [scale, position]) + + const handleMouseMove = useCallback((e: React.MouseEvent) => { + if (isDragging && scale > 1) { + const deltaX = e.clientX - dragStartRef.current.x + const deltaY = e.clientY - dragStartRef.current.y + + // Calculate boundaries + const imgRect = imgRef.current?.getBoundingClientRect() + const containerRect = imgRef.current?.parentElement?.getBoundingClientRect() + + if (imgRect && containerRect) { + const maxX = (imgRect.width * scale - containerRect.width) / 2 + const maxY = (imgRect.height * scale - containerRect.height) / 2 + + setPosition({ + x: Math.max(-maxX, Math.min(maxX, deltaX)), + y: Math.max(-maxY, Math.min(maxY, deltaY)), + }) + } + } + }, [isDragging, scale]) + + const handleMouseUp = useCallback(() => { + setIsDragging(false) + }, []) + + useEffect(() => { + document.addEventListener('mouseup', handleMouseUp) + return () => { + document.removeEventListener('mouseup', handleMouseUp) + } + }, [handleMouseUp]) + + useEffect(() => { + const handleKeyDown = (event: KeyboardEvent) => { + if (event.key === 'Escape') + onCancel() + } + + window.addEventListener('keydown', handleKeyDown) + + // Set focus to the container element + if (containerRef.current) + containerRef.current.focus() + + // Cleanup function + return () => { + window.removeEventListener('keydown', handleKeyDown) + } + }, [onCancel]) + return createPortal( -
e.stopPropagation()}> +
e.stopPropagation()} + onWheel={handleWheel} + onMouseDown={handleMouseDown} + onMouseMove={handleMouseMove} + onMouseUp={handleMouseUp} + style={{ cursor: scale > 1 ? 'move' : 'default' }} + tabIndex={-1}> {/* eslint-disable-next-line @next/next/no-img-element */} {title} -
- -
- + +
+ {isCopied + ? + : } +
+ + +
+ +
+ + +
+ +
+ + +
+ +
+ + +
+ +
+ +
- + className='absolute top-6 right-6 flex items-center justify-center w-8 h-8 bg-white/8 rounded-lg backdrop-blur-[2px] cursor-pointer' + onClick={onCancel}> +
, diff --git a/web/app/components/base/markdown.tsx b/web/app/components/base/markdown.tsx index d4e7dac4ae..443ee3410c 100644 --- a/web/app/components/base/markdown.tsx +++ b/web/app/components/base/markdown.tsx @@ -5,6 +5,7 @@ import RemarkMath from 'remark-math' import RemarkBreaks from 'remark-breaks' import RehypeKatex from 'rehype-katex' import RemarkGfm from 'remark-gfm' +import RehypeRaw from 'rehype-raw' import SyntaxHighlighter from 'react-syntax-highlighter' import { atelierHeathLight } from 'react-syntax-highlighter/dist/esm/styles/hljs' import type { RefObject } from 'react' @@ -18,6 +19,7 @@ import ImageGallery from '@/app/components/base/image-gallery' import { useChatContext } from '@/app/components/base/chat/chat/context' import VideoGallery from '@/app/components/base/video-gallery' import AudioGallery from '@/app/components/base/audio-gallery' +import SVGRenderer from '@/app/components/base/svg-gallery' // Available language https://github.com/react-syntax-highlighter/react-syntax-highlighter/blob/master/AVAILABLE_LANGUAGES_HLJS.MD const capitalizationLanguageNameMap: Record = { @@ -40,6 +42,7 @@ const capitalizationLanguageNameMap: Record = { powershell: 'PowerShell', json: 'JSON', latex: 'Latex', + svg: 'SVG', } const getCorrectCapitalizationLanguageName = (language: string) => { if (!language) @@ -107,6 +110,7 @@ const useLazyLoad = (ref: RefObject): boolean => { // Error: Minified React error 185; // visit https://reactjs.org/docs/error-decoder.html?invariant=185 for the full message // or use the non-minified dev environment for full errors and additional helpful warnings. + const CodeBlock: CodeComponent = memo(({ inline, className, children, ...props }) => { const [isSVG, setIsSVG] = useState(true) const match = /language-(\w+)/.exec(className || '') @@ -134,7 +138,7 @@ const CodeBlock: CodeComponent = memo(({ inline, className, children, ...props } >
{languageShowName}
- {language === 'mermaid' && } + {language === 'mermaid' && } {(language === 'mermaid' && isSVG) ? () - : ( - (language === 'echarts') - ? (
-
) + : (language === 'echarts' + ? (
) + : (language === 'svg' + ? () : ( {String(children).replace(/\n$/, '')} - ))} + )))}
) - : ( - - {children} - - ) + : ({children}) }, [chartData, children, className, inline, isSVG, language, languageShowName, match, props]) }) - CodeBlock.displayName = 'CodeBlock' const VideoBlock: CodeComponent = memo(({ node }) => { @@ -230,6 +227,7 @@ export function Markdown(props: { content: string; className?: string }) { remarkPlugins={[[RemarkGfm, RemarkMath, { singleDollarTextMath: false }], RemarkBreaks]} rehypePlugins={[ RehypeKatex, + RehypeRaw as any, // The Rehype plug-in is used to remove the ref attribute of an element () => { return (tree) => { @@ -244,6 +242,7 @@ export function Markdown(props: { content: string; className?: string }) { } }, ]} + disallowedElements={['script', 'iframe', 'head', 'html', 'meta', 'link', 'style', 'body']} components={{ code: CodeBlock, img: Img, @@ -266,19 +265,23 @@ export function Markdown(props: { content: string; className?: string }) { // This can happen when a component attempts to access an undefined object that references an unregistered map, causing the program to crash. export default class ErrorBoundary extends Component { - constructor(props) { + constructor(props: any) { super(props) this.state = { hasError: false } } - componentDidCatch(error, errorInfo) { + componentDidCatch(error: any, errorInfo: any) { this.setState({ hasError: true }) console.error(error, errorInfo) } render() { + // eslint-disable-next-line @typescript-eslint/ban-ts-comment + // @ts-expect-error if (this.state.hasError) - return
Oops! ECharts reported a runtime error.
(see the browser console for more information)
+ return
Oops! An error occurred. This could be due to an ECharts runtime error or invalid SVG content.
(see the browser console for more information)
+ // eslint-disable-next-line @typescript-eslint/ban-ts-comment + // @ts-expect-error return this.props.children } } diff --git a/web/app/components/base/svg-gallery/index.tsx b/web/app/components/base/svg-gallery/index.tsx new file mode 100644 index 0000000000..81e8e87655 --- /dev/null +++ b/web/app/components/base/svg-gallery/index.tsx @@ -0,0 +1,79 @@ +import { useEffect, useRef, useState } from 'react' +import { SVG } from '@svgdotjs/svg.js' +import ImagePreview from '@/app/components/base/image-uploader/image-preview' + +export const SVGRenderer = ({ content }: { content: string }) => { + const svgRef = useRef(null) + const [imagePreview, setImagePreview] = useState('') + const [windowSize, setWindowSize] = useState({ + width: typeof window !== 'undefined' ? window.innerWidth : 0, + height: typeof window !== 'undefined' ? window.innerHeight : 0, + }) + + const svgToDataURL = (svgElement: Element): string => { + const svgString = new XMLSerializer().serializeToString(svgElement) + const base64String = Buffer.from(svgString).toString('base64') + return `data:image/svg+xml;base64,${base64String}` + } + + useEffect(() => { + const handleResize = () => { + setWindowSize({ width: window.innerWidth, height: window.innerHeight }) + } + + window.addEventListener('resize', handleResize) + return () => window.removeEventListener('resize', handleResize) + }, []) + + useEffect(() => { + if (svgRef.current) { + try { + svgRef.current.innerHTML = '' + const draw = SVG().addTo(svgRef.current).size('100%', '100%') + + const parser = new DOMParser() + const svgDoc = parser.parseFromString(content, 'image/svg+xml') + const svgElement = svgDoc.documentElement + + if (!(svgElement instanceof SVGElement)) + throw new Error('Invalid SVG content') + + const originalWidth = parseInt(svgElement.getAttribute('width') || '400', 10) + const originalHeight = parseInt(svgElement.getAttribute('height') || '600', 10) + const scale = Math.min(windowSize.width / originalWidth, windowSize.height / originalHeight, 1) + const scaledWidth = originalWidth * scale + const scaledHeight = originalHeight * scale + draw.size(scaledWidth, scaledHeight) + + const rootElement = draw.svg(content) + rootElement.scale(scale) + + rootElement.click(() => { + setImagePreview(svgToDataURL(svgElement as Element)) + }) + } + catch (error) { + if (svgRef.current) + svgRef.current.innerHTML = 'Error rendering SVG. Wait for the image content to complete.' + } + } + }, [content, windowSize]) + + return ( + <> +
+ {imagePreview && ( setImagePreview('')} />)} + + ) +} + +export default SVGRenderer diff --git a/web/app/components/datasets/create/step-two/escape.ts b/web/app/components/datasets/create/step-two/escape.ts new file mode 100644 index 0000000000..098f43bc7f --- /dev/null +++ b/web/app/components/datasets/create/step-two/escape.ts @@ -0,0 +1,18 @@ +function escape(input: string): string { + if (!input || typeof input !== 'string') + return '' + + const res = input + .replaceAll('\\', '\\\\') + .replaceAll('\0', '\\0') + .replaceAll('\b', '\\b') + .replaceAll('\f', '\\f') + .replaceAll('\n', '\\n') + .replaceAll('\r', '\\r') + .replaceAll('\t', '\\t') + .replaceAll('\v', '\\v') + .replaceAll('\'', '\\\'') + return res +} + +export default escape diff --git a/web/app/components/datasets/create/step-two/index.tsx b/web/app/components/datasets/create/step-two/index.tsx index 15332b944d..94614918db 100644 --- a/web/app/components/datasets/create/step-two/index.tsx +++ b/web/app/components/datasets/create/step-two/index.tsx @@ -1,5 +1,5 @@ 'use client' -import React, { useEffect, useLayoutEffect, useRef, useState } from 'react' +import React, { useCallback, useEffect, useLayoutEffect, useRef, useState } from 'react' import { useTranslation } from 'react-i18next' import { useContext } from 'use-context-selector' import { useBoolean } from 'ahooks' @@ -13,6 +13,8 @@ import { groupBy } from 'lodash-es' import PreviewItem, { PreviewType } from './preview-item' import LanguageSelect from './language-select' import s from './index.module.css' +import unescape from './unescape' +import escape from './escape' import cn from '@/utils/classnames' import type { CrawlOptions, CrawlResultItem, CreateDocumentReq, CustomFile, FileIndexingEstimateResponse, FullDocumentDetail, IndexingEstimateParams, NotionInfo, PreProcessingRule, ProcessRule, Rules, createDocumentResponse } from '@/models/datasets' import { @@ -78,6 +80,8 @@ enum IndexingType { ECONOMICAL = 'economy', } +const DEFAULT_SEGMENT_IDENTIFIER = '\\n\\n' + const StepTwo = ({ isSetting, documentDetail, @@ -110,8 +114,11 @@ const StepTwo = ({ const previewScrollRef = useRef(null) const [previewScrolled, setPreviewScrolled] = useState(false) const [segmentationType, setSegmentationType] = useState(SegmentType.AUTO) - const [segmentIdentifier, setSegmentIdentifier] = useState('\\n') - const [max, setMax] = useState(5000) // default chunk length + const [segmentIdentifier, doSetSegmentIdentifier] = useState(DEFAULT_SEGMENT_IDENTIFIER) + const setSegmentIdentifier = useCallback((value: string) => { + doSetSegmentIdentifier(value ? escape(value) : DEFAULT_SEGMENT_IDENTIFIER) + }, []) + const [max, setMax] = useState(4000) // default chunk length const [overlap, setOverlap] = useState(50) const [rules, setRules] = useState([]) const [defaultConfig, setDefaultConfig] = useState() @@ -183,7 +190,7 @@ const StepTwo = ({ } const resetRules = () => { if (defaultConfig) { - setSegmentIdentifier((defaultConfig.segmentation.separator === '\n' ? '\\n' : defaultConfig.segmentation.separator) || '\\n') + setSegmentIdentifier(defaultConfig.segmentation.separator) setMax(defaultConfig.segmentation.max_tokens) setOverlap(defaultConfig.segmentation.chunk_overlap) setRules(defaultConfig.pre_processing_rules) @@ -217,7 +224,7 @@ const StepTwo = ({ const ruleObj = { pre_processing_rules: rules, segmentation: { - separator: segmentIdentifier === '\\n' ? '\n' : segmentIdentifier, + separator: unescape(segmentIdentifier), max_tokens: max, chunk_overlap: overlap, }, @@ -394,7 +401,7 @@ const StepTwo = ({ try { const res = await fetchDefaultProcessRule({ url: '/datasets/process-rule' }) const separator = res.rules.segmentation.separator - setSegmentIdentifier((separator === '\n' ? '\\n' : separator) || '\\n') + setSegmentIdentifier(separator) setMax(res.rules.segmentation.max_tokens) setOverlap(res.rules.segmentation.chunk_overlap) setRules(res.rules.pre_processing_rules) @@ -411,7 +418,7 @@ const StepTwo = ({ const separator = rules.segmentation.separator const max = rules.segmentation.max_tokens const overlap = rules.segmentation.chunk_overlap - setSegmentIdentifier((separator === '\n' ? '\\n' : separator) || '\\n') + setSegmentIdentifier(separator) setMax(max) setOverlap(overlap) setRules(rules.pre_processing_rules) @@ -616,12 +623,22 @@ const StepTwo = ({
-
{t('datasetCreation.stepTwo.separator')}
+
+ {t('datasetCreation.stepTwo.separator')} + + {t('datasetCreation.stepTwo.separatorTip')} +
+ } + /> +
setSegmentIdentifier(e.target.value)} + placeholder={t('datasetCreation.stepTwo.separatorPlaceholder') || ''} + value={segmentIdentifier} + onChange={e => doSetSegmentIdentifier(e.target.value)} />
diff --git a/web/app/components/datasets/create/step-two/unescape.ts b/web/app/components/datasets/create/step-two/unescape.ts new file mode 100644 index 0000000000..5c0f9e426a --- /dev/null +++ b/web/app/components/datasets/create/step-two/unescape.ts @@ -0,0 +1,54 @@ +// https://github.com/iamakulov/unescape-js/blob/master/src/index.js + +/** + * \\ - matches the backslash which indicates the beginning of an escape sequence + * ( + * u\{([0-9A-Fa-f]+)\} - first alternative; matches the variable-length hexadecimal escape sequence (\u{ABCD0}) + * | + * u([0-9A-Fa-f]{4}) - second alternative; matches the 4-digit hexadecimal escape sequence (\uABCD) + * | + * x([0-9A-Fa-f]{2}) - third alternative; matches the 2-digit hexadecimal escape sequence (\xA5) + * | + * ([1-7][0-7]{0,2}|[0-7]{2,3}) - fourth alternative; matches the up-to-3-digit octal escape sequence (\5 or \512) + * | + * (['"tbrnfv0\\]) - fifth alternative; matches the special escape characters (\t, \n and so on) + * | + * \U([0-9A-Fa-f]+) - sixth alternative; matches the 8-digit hexadecimal escape sequence used by python (\U0001F3B5) + * ) + */ +const jsEscapeRegex = /\\(u\{([0-9A-Fa-f]+)\}|u([0-9A-Fa-f]{4})|x([0-9A-Fa-f]{2})|([1-7][0-7]{0,2}|[0-7]{2,3})|(['"tbrnfv0\\]))|\\U([0-9A-Fa-f]{8})/g + +const usualEscapeSequences: Record = { + '0': '\0', + 'b': '\b', + 'f': '\f', + 'n': '\n', + 'r': '\r', + 't': '\t', + 'v': '\v', + '\'': '\'', + '"': '"', + '\\': '\\', +} + +const fromHex = (str: string) => String.fromCodePoint(parseInt(str, 16)) +const fromOct = (str: string) => String.fromCodePoint(parseInt(str, 8)) + +const unescape = (str: string) => { + return str.replace(jsEscapeRegex, (_, __, varHex, longHex, shortHex, octal, specialCharacter, python) => { + if (varHex !== undefined) + return fromHex(varHex) + else if (longHex !== undefined) + return fromHex(longHex) + else if (shortHex !== undefined) + return fromHex(shortHex) + else if (octal !== undefined) + return fromOct(octal) + else if (python !== undefined) + return fromHex(python) + else + return usualEscapeSequences[specialCharacter] + }) +} + +export default unescape diff --git a/web/app/components/share/text-generation/result/index.tsx b/web/app/components/share/text-generation/result/index.tsx index 2d5546f9b4..a61302fc98 100644 --- a/web/app/components/share/text-generation/result/index.tsx +++ b/web/app/components/share/text-generation/result/index.tsx @@ -269,8 +269,10 @@ const Result: FC = ({ })) }, onWorkflowFinished: ({ data }) => { - if (isTimeout) + if (isTimeout) { + notify({ type: 'warning', message: t('appDebug.warningMessage.timeoutExceeded') }) return + } if (data.error) { notify({ type: 'error', message: data.error }) setWorkflowProcessData(produce(getWorkflowProcessData()!, (draft) => { @@ -326,8 +328,10 @@ const Result: FC = ({ setCompletionRes(res.join('')) }, onCompleted: () => { - if (isTimeout) + if (isTimeout) { + notify({ type: 'warning', message: t('appDebug.warningMessage.timeoutExceeded') }) return + } setRespondingFalse() setMessageId(tempMessageId) onCompleted(getCompletionRes(), taskId, true) @@ -338,8 +342,10 @@ const Result: FC = ({ setCompletionRes(res.join('')) }, onError() { - if (isTimeout) + if (isTimeout) { + notify({ type: 'warning', message: t('appDebug.warningMessage.timeoutExceeded') }) return + } setRespondingFalse() onCompleted(getCompletionRes(), taskId, false) isEnd = true diff --git a/web/app/components/share/text-generation/run-once/index.tsx b/web/app/components/share/text-generation/run-once/index.tsx index e3b8d6a0a4..e74616b08d 100644 --- a/web/app/components/share/text-generation/run-once/index.tsx +++ b/web/app/components/share/text-generation/run-once/index.tsx @@ -1,4 +1,4 @@ -import type { FC } from 'react' +import type { FC, FormEvent } from 'react' import React from 'react' import { useTranslation } from 'react-i18next' import { @@ -39,11 +39,16 @@ const RunOnce: FC = ({ onInputsChange(newInputs) } + const onSubmit = (e: FormEvent) => { + e.preventDefault() + onSend() + } + return (
{/* input form */} -
+ {promptConfig.prompt_variables.map(item => (
@@ -65,12 +70,6 @@ const RunOnce: FC = ({ placeholder={`${item.name}${!item.required ? `(${t('appDebug.variableTable.optional')})` : ''}`} value={inputs[item.key]} onChange={(e) => { onInputsChange({ ...inputs, [item.key]: e.target.value }) }} - onKeyDown={(e) => { - if (e.key === 'Enter') { - e.preventDefault() - onSend() - } - }} maxLength={item.max_length || DEFAULT_VALUE_MAX_LEN} /> )} @@ -124,8 +123,8 @@ const RunOnce: FC = ({ {t('common.operation.clear')}
)} - needsDelay >
{ const nodes = useNodes() const node = useMemo(() => { - if (isSystemVar(valueSelector)) - return nodes.find(node => node.data.type === BlockEnum.Start) + if (isSystemVar(valueSelector)) { + const startNode = availableNodes?.find(n => n.data.type === BlockEnum.Start) + if (startNode) + return startNode + } + return getNodeInfoById(availableNodes || nodes, valueSelector[0]) + }, [nodes, valueSelector, availableNodes]) - return nodes.find(node => node.id === valueSelector[0]) - }, [nodes, valueSelector]) const isEnv = isENV(valueSelector) const isChatVar = isConversationVar(valueSelector) + const isValid = Boolean(node) || isEnv || isChatVar const variableName = isSystemVar(valueSelector) ? valueSelector.slice(0).join('.') : valueSelector.slice(1).join('.') + const { t } = useTranslation() return ( -
- {!isEnv && !isChatVar && ( - <> + +
+ {(!isEnv && !isChatVar && <> {node && ( - + <> + +
+ {node?.data.title} +
+ )} -
- {node?.data.title} -
- - )} - {isEnv && } - {isChatVar && } -
- {variableName} + )} + {isEnv && } + {isChatVar && } +
+ {variableName} +
+ { + varType && ( +
{capitalize(varType)}
+ ) + } + {!isValid && }
- { - varType && ( -
{capitalize(varType)}
- ) - } -
+ ) } diff --git a/web/app/components/workflow/nodes/_base/components/variable/var-reference-picker.tsx b/web/app/components/workflow/nodes/_base/components/variable/var-reference-picker.tsx index 7fb4ad68d8..5e51173673 100644 --- a/web/app/components/workflow/nodes/_base/components/variable/var-reference-picker.tsx +++ b/web/app/components/workflow/nodes/_base/components/variable/var-reference-picker.tsx @@ -5,9 +5,10 @@ import { useTranslation } from 'react-i18next' import { RiArrowDownSLine, RiCloseLine, + RiErrorWarningFill, } from '@remixicon/react' import produce from 'immer' -import { useStoreApi } from 'reactflow' +import { useEdges, useStoreApi } from 'reactflow' import useAvailableVarList from '../../hooks/use-available-var-list' import VarReferencePopup from './var-reference-popup' import { getNodeInfoById, isConversationVar, isENV, isSystemVar } from './utils' @@ -33,6 +34,8 @@ import { VarType as VarKindType } from '@/app/components/workflow/nodes/tool/typ import TypeSelector from '@/app/components/workflow/nodes/_base/components/selector' import AddButton from '@/app/components/base/button/add-button' import Badge from '@/app/components/base/badge' +import Tooltip from '@/app/components/base/tooltip' + const TRIGGER_DEFAULT_WIDTH = 227 type Props = { @@ -77,6 +80,7 @@ const VarReferencePicker: FC = ({ const { getNodes, } = store.getState() + const edges = useEdges() const isChatMode = useIsChatMode() const { getCurrentVariableType } = useWorkflowVariables() @@ -179,7 +183,7 @@ const VarReferencePicker: FC = ({ const handleVarReferenceChange = useCallback((value: ValueSelector, varInfo: Var) => { // sys var not passed to backend const newValue = produce(value, (draft) => { - if (draft[1] && draft[1].startsWith('sys')) { + if (draft[1] && draft[1].startsWith('sys.')) { draft.shift() const paths = draft[0].split('.') paths.forEach((p, i) => { @@ -206,8 +210,16 @@ const VarReferencePicker: FC = ({ isConstant: !!isConstant, }) - const isEnv = isENV(value as ValueSelector) - const isChatVar = isConversationVar(value as ValueSelector) + const { isEnv, isChatVar, isValidVar } = useMemo(() => { + const isEnv = isENV(value as ValueSelector) + const isChatVar = isConversationVar(value as ValueSelector) + const isValidVar = Boolean(outputVarNode) || isEnv || isChatVar + return { + isEnv, + isChatVar, + isValidVar, + } + }, [value, edges, outputVarNode]) // 8(left/right-padding) + 14(icon) + 4 + 14 + 2 = 42 + 17 buff const availableWidth = triggerWidth - 56 @@ -285,39 +297,44 @@ const VarReferencePicker: FC = ({ className='grow h-full' >
-
- {hasValue - ? ( - <> - {isShowNodeName && !isEnv && !isChatVar && ( -
-
- + +
+ {hasValue + ? ( + <> + {isShowNodeName && !isEnv && !isChatVar && ( +
+
+ {outputVarNode?.type && } +
+
{outputVarNode?.title}
+
-
{outputVarNode?.title}
- + )} +
+ {!hasValue && } + {isEnv && } + {isChatVar && } +
{varName}
- )} -
- {!hasValue && } - {isEnv && } - {isChatVar && } -
{varName}
-
-
{type}
- - ) - :
{t('workflow.common.setVarValuePlaceholder')}
} -
+
{type}
+ {!isValidVar && } + + ) + :
{t('workflow.common.setVarValuePlaceholder')}
} +
+
diff --git a/web/app/components/workflow/nodes/if-else/components/condition-list/condition-item.tsx b/web/app/components/workflow/nodes/if-else/components/condition-list/condition-item.tsx index c6cb580118..c96f6b3ef6 100644 --- a/web/app/components/workflow/nodes/if-else/components/condition-list/condition-item.tsx +++ b/web/app/components/workflow/nodes/if-else/components/condition-list/condition-item.tsx @@ -80,6 +80,7 @@ const ConditionItem = ({
diff --git a/web/app/components/workflow/run/tracing-panel.tsx b/web/app/components/workflow/run/tracing-panel.tsx index 7684dc84b6..613c10198d 100644 --- a/web/app/components/workflow/run/tracing-panel.tsx +++ b/web/app/components/workflow/run/tracing-panel.tsx @@ -11,6 +11,7 @@ import { RiArrowDownSLine, RiMenu4Line, } from '@remixicon/react' +import { useTranslation } from 'react-i18next' import NodePanel from './node' import { BlockEnum, @@ -37,7 +38,7 @@ type TracingNodeProps = { hideNodeProcessDetail?: boolean } -function buildLogTree(nodes: NodeTracing[]): TracingNodeProps[] { +function buildLogTree(nodes: NodeTracing[], t: (key: string) => string): TracingNodeProps[] { const rootNodes: TracingNodeProps[] = [] const parallelStacks: { [key: string]: TracingNodeProps } = {} const levelCounts: { [key: string]: number } = {} @@ -58,7 +59,7 @@ function buildLogTree(nodes: NodeTracing[]): TracingNodeProps[] { const parentTitle = parentId ? parallelStacks[parentId]?.parallelTitle : '' const levelNumber = parentTitle ? parseInt(parentTitle.split('-')[1]) + 1 : 1 const letter = parallelChildCounts[levelKey]?.size > 1 ? String.fromCharCode(64 + levelCounts[levelKey]) : '' - return `PARALLEL-${levelNumber}${letter}` + return `${t('workflow.common.parallel')}-${levelNumber}${letter}` } const getBranchTitle = (parentId: string | null, branchNum: number): string => { @@ -67,7 +68,7 @@ function buildLogTree(nodes: NodeTracing[]): TracingNodeProps[] { const levelNumber = parentTitle ? parseInt(parentTitle.split('-')[1]) + 1 : 1 const letter = parallelChildCounts[levelKey]?.size > 1 ? String.fromCharCode(64 + levelCounts[levelKey]) : '' const branchLetter = String.fromCharCode(64 + branchNum) - return `BRANCH-${levelNumber}${letter}-${branchLetter}` + return `${t('workflow.common.branch')}-${levelNumber}${letter}-${branchLetter}` } // Count parallel children (for figuring out if we need to use letters) @@ -163,7 +164,8 @@ const TracingPanel: FC = ({ hideNodeInfo = false, hideNodeProcessDetail = false, }) => { - const treeNodes = buildLogTree(list) + const { t } = useTranslation() + const treeNodes = buildLogTree(list, t) const [collapsedNodes, setCollapsedNodes] = useState>(new Set()) const [hoveredParallel, setHoveredParallel] = useState(null) diff --git a/web/app/layout.tsx b/web/app/layout.tsx index 008fdefa0b..e9242edfad 100644 --- a/web/app/layout.tsx +++ b/web/app/layout.tsx @@ -43,6 +43,7 @@ const LocaleLayout = ({ data-public-sentry-dsn={process.env.NEXT_PUBLIC_SENTRY_DSN} data-public-maintenance-notice={process.env.NEXT_PUBLIC_MAINTENANCE_NOTICE} data-public-site-about={process.env.NEXT_PUBLIC_SITE_ABOUT} + data-public-text-generation-timeout-ms={process.env.NEXT_PUBLIC_TEXT_GENERATION_TIMEOUT_MS} > diff --git a/web/config/index.ts b/web/config/index.ts index 71edf8f939..21fc2f211c 100644 --- a/web/config/index.ts +++ b/web/config/index.ts @@ -246,6 +246,13 @@ Thought: {{agent_scratchpad}} export const VAR_REGEX = /\{\{(#[a-zA-Z0-9_-]{1,50}(\.[a-zA-Z_][a-zA-Z0-9_]{0,29}){1,10}#)\}\}/gi -export const TEXT_GENERATION_TIMEOUT_MS = 60000 +export let textGenerationTimeoutMs = 60000 + +if (process.env.NEXT_PUBLIC_TEXT_GENERATION_TIMEOUT_MS && process.env.NEXT_PUBLIC_TEXT_GENERATION_TIMEOUT_MS !== '') + textGenerationTimeoutMs = parseInt(process.env.NEXT_PUBLIC_TEXT_GENERATION_TIMEOUT_MS) +else if (globalThis.document?.body?.getAttribute('data-public-text-generation-timeout-ms') && globalThis.document.body.getAttribute('data-public-text-generation-timeout-ms') !== '') + textGenerationTimeoutMs = parseInt(globalThis.document.body.getAttribute('data-public-text-generation-timeout-ms') as string) + +export const TEXT_GENERATION_TIMEOUT_MS = textGenerationTimeoutMs export const DISABLE_UPLOAD_IMAGE_AS_ICON = process.env.NEXT_PUBLIC_DISABLE_UPLOAD_IMAGE_AS_ICON === 'true' diff --git a/web/docker/entrypoint.sh b/web/docker/entrypoint.sh index a19c543d68..fc4a8f45bc 100755 --- a/web/docker/entrypoint.sh +++ b/web/docker/entrypoint.sh @@ -21,4 +21,6 @@ export NEXT_PUBLIC_SENTRY_DSN=${SENTRY_DSN} export NEXT_PUBLIC_SITE_ABOUT=${SITE_ABOUT} export NEXT_TELEMETRY_DISABLED=${NEXT_TELEMETRY_DISABLED} +export NEXT_PUBLIC_TEXT_GENERATION_TIMEOUT_MS=${TEXT_GENERATION_TIMEOUT_MS} + pm2 start ./pm2.json --no-daemon diff --git a/web/i18n/de-DE/workflow.ts b/web/i18n/de-DE/workflow.ts index 5e154aeca5..c01d0e6f99 100644 --- a/web/i18n/de-DE/workflow.ts +++ b/web/i18n/de-DE/workflow.ts @@ -93,6 +93,8 @@ const translation = { disconnect: 'Trennen', jumpToNode: 'Zu diesem Knoten springen', addParallelNode: 'Parallelen Knoten hinzufügen', + parallel: 'PARALLEL', + branch: 'ZWEIG', }, env: { envPanelTitle: 'Umgebungsvariablen', diff --git a/web/i18n/en-US/app-debug.ts b/web/i18n/en-US/app-debug.ts index b1f3f33cd8..3156c9f6cc 100644 --- a/web/i18n/en-US/app-debug.ts +++ b/web/i18n/en-US/app-debug.ts @@ -268,6 +268,9 @@ const translation = { notSelectModel: 'Please choose a model', waitForImgUpload: 'Please wait for the image to upload', }, + warningMessage: { + timeoutExceeded: 'Results are not displayed due to timeout. Please refer to the logs to gather complete results.', + }, chatSubTitle: 'Instructions', completionSubTitle: 'Prefix Prompt', promptTip: diff --git a/web/i18n/en-US/dataset-creation.ts b/web/i18n/en-US/dataset-creation.ts index 40463593f9..32f9d596ca 100644 --- a/web/i18n/en-US/dataset-creation.ts +++ b/web/i18n/en-US/dataset-creation.ts @@ -87,7 +87,8 @@ const translation = { custom: 'Custom', customDescription: 'Customize chunks rules, chunks length, and preprocessing rules, etc.', separator: 'Delimiter', - separatorPlaceholder: 'For example, newline (\\\\n) or special separator (such as "***")', + separatorTip: 'A delimiter is the character used to separate text. \\n\\n and \\n are commonly used delimiters for separating paragraphs and lines. Combined with commas (\\n\\n,\\n), paragraphs will be segmented by lines when exceeding the maximum chunk length. You can also use special delimiters defined by yourself (e.g. ***).', + separatorPlaceholder: '\\n\\n for separating paragraphs; \\n for separating lines', maxLength: 'Maximum chunk length', overlap: 'Chunk overlap', overlapTip: 'Setting the chunk overlap can maintain the semantic relevance between them, enhancing the retrieve effect. It is recommended to set 10%-25% of the maximum chunk size.', diff --git a/web/i18n/en-US/workflow.ts b/web/i18n/en-US/workflow.ts index b83d213cb8..a7e768911f 100644 --- a/web/i18n/en-US/workflow.ts +++ b/web/i18n/en-US/workflow.ts @@ -93,6 +93,8 @@ const translation = { disconnect: 'Disconnect', jumpToNode: 'Jump to this node', addParallelNode: 'Add Parallel Node', + parallel: 'PARALLEL', + branch: 'BRANCH', }, env: { envPanelTitle: 'Environment Variables', diff --git a/web/i18n/es-ES/workflow.ts b/web/i18n/es-ES/workflow.ts index 0efc996f91..2260631d0f 100644 --- a/web/i18n/es-ES/workflow.ts +++ b/web/i18n/es-ES/workflow.ts @@ -93,6 +93,8 @@ const translation = { disconnect: 'Desconectar', jumpToNode: 'Saltar a este nodo', addParallelNode: 'Agregar nodo paralelo', + parallel: 'PARALELO', + branch: 'RAMA', }, env: { envPanelTitle: 'Variables de Entorno', diff --git a/web/i18n/fa-IR/workflow.ts b/web/i18n/fa-IR/workflow.ts index 67020f5025..eb36dfdc88 100644 --- a/web/i18n/fa-IR/workflow.ts +++ b/web/i18n/fa-IR/workflow.ts @@ -93,6 +93,8 @@ const translation = { jumpToNode: 'پرش به این گره', parallelRun: 'اجرای موازی', addParallelNode: 'افزودن گره موازی', + parallel: 'موازی', + branch: 'شاخه', }, env: { envPanelTitle: 'متغیرهای محیطی', diff --git a/web/i18n/fr-FR/workflow.ts b/web/i18n/fr-FR/workflow.ts index 3e56246e0c..878d25804e 100644 --- a/web/i18n/fr-FR/workflow.ts +++ b/web/i18n/fr-FR/workflow.ts @@ -93,6 +93,8 @@ const translation = { disconnect: 'Déconnecter', jumpToNode: 'Aller à ce nœud', addParallelNode: 'Ajouter un nœud parallèle', + parallel: 'PARALLÈLE', + branch: 'BRANCHE', }, env: { envPanelTitle: 'Variables d\'Environnement', diff --git a/web/i18n/hi-IN/workflow.ts b/web/i18n/hi-IN/workflow.ts index 072e4874e3..ac356c2067 100644 --- a/web/i18n/hi-IN/workflow.ts +++ b/web/i18n/hi-IN/workflow.ts @@ -96,6 +96,8 @@ const translation = { parallelRun: 'समानांतर रन', jumpToNode: 'इस नोड पर जाएं', addParallelNode: 'समानांतर नोड जोड़ें', + parallel: 'समानांतर', + branch: 'शाखा', }, env: { envPanelTitle: 'पर्यावरण चर', diff --git a/web/i18n/it-IT/workflow.ts b/web/i18n/it-IT/workflow.ts index f5d6fc8bf5..0427a45cd9 100644 --- a/web/i18n/it-IT/workflow.ts +++ b/web/i18n/it-IT/workflow.ts @@ -97,6 +97,8 @@ const translation = { disconnect: 'Disconnettere', jumpToNode: 'Vai a questo nodo', addParallelNode: 'Aggiungi nodo parallelo', + parallel: 'PARALLELO', + branch: 'RAMO', }, env: { envPanelTitle: 'Variabili d\'Ambiente', diff --git a/web/i18n/ja-JP/workflow.ts b/web/i18n/ja-JP/workflow.ts index 755061e8f6..48c2019601 100644 --- a/web/i18n/ja-JP/workflow.ts +++ b/web/i18n/ja-JP/workflow.ts @@ -93,6 +93,8 @@ const translation = { disconnect: '切る', jumpToNode: 'このノードにジャンプします', addParallelNode: '並列ノードを追加', + parallel: '並列', + branch: 'ブランチ', }, env: { envPanelTitle: '環境変数', diff --git a/web/i18n/ko-KR/workflow.ts b/web/i18n/ko-KR/workflow.ts index 8fed0e0417..4a97943790 100644 --- a/web/i18n/ko-KR/workflow.ts +++ b/web/i18n/ko-KR/workflow.ts @@ -93,6 +93,8 @@ const translation = { disconnect: '분리하다', jumpToNode: '이 노드로 이동', addParallelNode: '병렬 노드 추가', + parallel: '병렬', + branch: '브랜치', }, env: { envPanelTitle: '환경 변수', diff --git a/web/i18n/pl-PL/workflow.ts b/web/i18n/pl-PL/workflow.ts index de05ee7169..41927668f7 100644 --- a/web/i18n/pl-PL/workflow.ts +++ b/web/i18n/pl-PL/workflow.ts @@ -93,6 +93,8 @@ const translation = { jumpToNode: 'Przejdź do tego węzła', disconnect: 'Odłączyć', addParallelNode: 'Dodaj węzeł równoległy', + parallel: 'RÓWNOLEGŁY', + branch: 'GAŁĄŹ', }, env: { envPanelTitle: 'Zmienne Środowiskowe', diff --git a/web/i18n/pt-BR/workflow.ts b/web/i18n/pt-BR/workflow.ts index ef589c0bde..222fc788bf 100644 --- a/web/i18n/pt-BR/workflow.ts +++ b/web/i18n/pt-BR/workflow.ts @@ -93,6 +93,8 @@ const translation = { disconnect: 'Desligar', jumpToNode: 'Ir para este nó', addParallelNode: 'Adicionar nó paralelo', + parallel: 'PARALELO', + branch: 'RAMIFICAÇÃO', }, env: { envPanelTitle: 'Variáveis de Ambiente', diff --git a/web/i18n/ro-RO/workflow.ts b/web/i18n/ro-RO/workflow.ts index 689ebdead9..ac4b718b07 100644 --- a/web/i18n/ro-RO/workflow.ts +++ b/web/i18n/ro-RO/workflow.ts @@ -93,6 +93,8 @@ const translation = { disconnect: 'Deconecta', jumpToNode: 'Sari la acest nod', addParallelNode: 'Adăugare nod paralel', + parallel: 'PARALEL', + branch: 'RAMURĂ', }, env: { envPanelTitle: 'Variabile de Mediu', diff --git a/web/i18n/ru-RU/workflow.ts b/web/i18n/ru-RU/workflow.ts index 9d3ce1235c..1931863895 100644 --- a/web/i18n/ru-RU/workflow.ts +++ b/web/i18n/ru-RU/workflow.ts @@ -93,6 +93,8 @@ const translation = { disconnect: 'Разъединять', jumpToNode: 'Перейти к этому узлу', addParallelNode: 'Добавить параллельный узел', + parallel: 'ПАРАЛЛЕЛЬНЫЙ', + branch: 'ВЕТКА', }, env: { envPanelTitle: 'Переменные среды', diff --git a/web/i18n/tr-TR/workflow.ts b/web/i18n/tr-TR/workflow.ts index 96313d6d6b..a33a3724ad 100644 --- a/web/i18n/tr-TR/workflow.ts +++ b/web/i18n/tr-TR/workflow.ts @@ -93,6 +93,8 @@ const translation = { addParallelNode: 'Paralel Düğüm Ekle', disconnect: 'Ayırmak', parallelRun: 'Paralel Koşu', + parallel: 'PARALEL', + branch: 'DAL', }, env: { envPanelTitle: 'Çevre Değişkenleri', diff --git a/web/i18n/uk-UA/workflow.ts b/web/i18n/uk-UA/workflow.ts index 03471348c8..e1bea99bcd 100644 --- a/web/i18n/uk-UA/workflow.ts +++ b/web/i18n/uk-UA/workflow.ts @@ -93,6 +93,8 @@ const translation = { parallelRun: 'Паралельний біг', jumpToNode: 'Перейти до цього вузла', addParallelNode: 'Додати паралельний вузол', + parallel: 'ПАРАЛЕЛЬНИЙ', + branch: 'ГІЛКА', }, env: { envPanelTitle: 'Змінні середовища', diff --git a/web/i18n/vi-VN/workflow.ts b/web/i18n/vi-VN/workflow.ts index 5be19ab7fd..5e6f0e0063 100644 --- a/web/i18n/vi-VN/workflow.ts +++ b/web/i18n/vi-VN/workflow.ts @@ -93,6 +93,8 @@ const translation = { disconnect: 'Ngắt kết nối', jumpToNode: 'Chuyển đến nút này', addParallelNode: 'Thêm nút song song', + parallel: 'SONG SONG', + branch: 'NHÁNH', }, env: { envPanelTitle: 'Biến Môi Trường', diff --git a/web/i18n/zh-Hans/dataset-creation.ts b/web/i18n/zh-Hans/dataset-creation.ts index 47a15921f7..78f5170791 100644 --- a/web/i18n/zh-Hans/dataset-creation.ts +++ b/web/i18n/zh-Hans/dataset-creation.ts @@ -87,7 +87,8 @@ const translation = { custom: '自定义', customDescription: '自定义分段规则、分段长度以及预处理规则等参数', separator: '分段标识符', - separatorPlaceholder: '例如换行符（\n）或特定的分隔符（如 "***"）', + separatorTip: '分隔符是用于分隔文本的字符。\\n\\n 和 \\n 是常用于分隔段落和行的分隔符。用逗号连接分隔符（\\n\\n,\\n），当段落超过最大块长度时，会按行进行分割。你也可以使用自定义的特殊分隔符（例如 ***）。', + separatorPlaceholder: '\\n\\n 用于分段；\\n 用于分行', maxLength: '分段最大长度', overlap: '分段重叠长度', overlapTip: '设置分段之间的重叠长度可以保留分段之间的语义关系，提升召回效果。建议设置为最大分段长度的10%-25%', diff --git a/web/i18n/zh-Hans/workflow.ts b/web/i18n/zh-Hans/workflow.ts index 39311649f4..3579ec5df3 100644 --- a/web/i18n/zh-Hans/workflow.ts +++ b/web/i18n/zh-Hans/workflow.ts @@ -93,6 +93,8 @@ const translation = { disconnect: '断开连接', jumpToNode: '跳转到节点', addParallelNode: '添加并行节点', + parallel: '并行', + branch: '分支', }, env: { envPanelTitle: '环境变量', diff --git a/web/i18n/zh-Hant/workflow.ts b/web/i18n/zh-Hant/workflow.ts index eef3ffaebd..8e1b7529fe 100644 --- a/web/i18n/zh-Hant/workflow.ts +++ b/web/i18n/zh-Hant/workflow.ts @@ -93,6 +93,8 @@ const translation = { disconnect: '斷開', jumpToNode: '跳轉到此節點', addParallelNode: '添加並行節點', + parallel: '並行', + branch: '分支', }, env: { envPanelTitle: '環境變數', diff --git a/web/package.json b/web/package.json index 197a1e7e05..bc532fb242 100644 --- a/web/package.json +++ b/web/package.json @@ -44,6 +44,7 @@ "classnames": "^2.3.2", "copy-to-clipboard": "^3.3.3", "crypto-js": "^4.2.0", + "@svgdotjs/svg.js": "^3.2.4", "dayjs": "^1.11.7", "echarts": "^5.4.1", "echarts-for-react": "^3.0.2", diff --git a/web/utils/format.spec.ts b/web/utils/format.spec.ts new file mode 100644 index 0000000000..f349efa4e4 --- /dev/null +++ b/web/utils/format.spec.ts @@ -0,0 +1,61 @@ +import { formatFileSize, formatNumber, formatTime } from './format' +describe('formatNumber', () => { + test('should correctly format integers', () => { + expect(formatNumber(1234567)).toBe('1,234,567') + }) + test('should correctly format decimals', () => { + expect(formatNumber(1234567.89)).toBe('1,234,567.89') + }) + test('should correctly handle string input', () => { + expect(formatNumber('1234567')).toBe('1,234,567') + }) + test('should correctly handle zero', () => { + expect(formatNumber(0)).toBe(0) + }) + test('should correctly handle negative numbers', () => { + expect(formatNumber(-1234567)).toBe('-1,234,567') + }) + test('should correctly handle empty input', () => { + expect(formatNumber('')).toBe('') + }) +}) +describe('formatFileSize', () => { + test('should return the input if it is falsy', () => { + expect(formatFileSize(0)).toBe(0) + }) + test('should format bytes correctly', () => { + expect(formatFileSize(500)).toBe('500.00B') + }) + test('should format kilobytes correctly', () => { + expect(formatFileSize(1500)).toBe('1.46KB') + }) + test('should format megabytes correctly', () => { + expect(formatFileSize(1500000)).toBe('1.43MB') + }) + test('should format gigabytes correctly', () => { + expect(formatFileSize(1500000000)).toBe('1.40GB') + }) + test('should format terabytes correctly', () => { + expect(formatFileSize(1500000000000)).toBe('1.36TB') + }) + test('should format petabytes correctly', () => { + expect(formatFileSize(1500000000000000)).toBe('1.33PB') + }) +}) +describe('formatTime', () => { + test('should return the input if it is falsy', () => { + expect(formatTime(0)).toBe(0) + }) + test('should format seconds correctly', () => { + expect(formatTime(30)).toBe('30.00 sec') + }) + test('should format minutes correctly', () => { + expect(formatTime(90)).toBe('1.50 min') + }) + test('should format hours correctly', () => { + expect(formatTime(3600)).toBe('1.00 h') + }) + test('should handle large numbers', () => { + expect(formatTime(7200)).toBe('2.00 h') + }) +}) diff --git a/web/yarn.lock b/web/yarn.lock index 3c020c9664..bec2059a47 100644 --- a/web/yarn.lock +++ b/web/yarn.lock @@ -1489,6 +1489,11 @@ dependencies: "@sinonjs/commons" "^3.0.0" +"@svgdotjs/svg.js@^3.2.4": + version "3.2.4" + resolved "https://registry.yarnpkg.com/@svgdotjs/svg.js/-/svg.js-3.2.4.tgz#4716be92a64c66b29921b63f7235fcfb953fb13a" + integrity sha512-BjJ/7vWNowlX3Z8O4ywT58DqbNRyYlkk6Yz/D13aB7hGmfQTvGX4Tkgtm/ApYlu9M7lCQi15xUEidqMUmdMYwg== + "@swc/counter@^0.1.3": version "0.1.3" resolved "https://registry.npmjs.org/@swc/counter/-/counter-0.1.3.tgz"