opensource
/
dify
mirror of https://github.com/langgenius/dify.git
2
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

fix question classifier issue when llm 

is completion mode
This commit is contained in:
jyong 2024-03-27 14:19:23 +08:00
parent e952e01dfe
commit a29d3f2400
2 changed files with 12 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
api/core/workflow/nodes/question_classifier/question_classifier_node.py
View File

@ -188,11 +188,10 @@ class QuestionClassifierNode(LLMNode):
prompt_messages.append(user_prompt_message_3)
return prompt_messages
elif model_mode == ModelMode.COMPLETION:
prompt_messages.append(CompletionModelPromptTemplate(
return CompletionModelPromptTemplate(
text=QUESTION_CLASSIFIER_COMPLETION_PROMPT.format(input_text=input_text, categories=class_names_str,
classification_instructions=instruction)
))
)
return prompt_messages
else:
raise ValueError(f"Model mode {model_mode} not support.")

18
api/core/workflow/nodes/question_classifier/template_prompts.py
View File

@ -48,15 +48,17 @@ Your task is to assign one categories ONLY to the input text and only one catego
The input text is in the variable text_field. Categories are specified as a comma-separated list in the variable categories or left empty for automatic determination. Classification instructions may be included to improve the classification accuracy.
### Constraint
DO NOT include anything other than the JSON array in your response.
### Example
Input:
{{"input_text": ["I recently had a great experience with your company. The service was prompt and the staff was very friendly."],"categories": ["Customer Service, Satisfaction, Sales, Product"], "classification_instructions": ["classify the text based on the feedback provided by customer"]}}
{{"input_text": ["bad service, slow to bring the food"],"categories": ["Food Quality, Experience, Price" ], "classification_instructions": []}}
Output:
{{"keywords": ["recently", "great experience", "company", "service", "prompt", "staff", "friendly"],"categories": ["Customer Service"]}}
{{"keywords": ["bad service", "slow", "food", "tip", "terrible", "waitresses"],"categories": ["Experience""]}}
### Example
Here is the chat example between human and assistant, inside <example></example> XML tags.
<example>
User:{{"input_text": ["I recently had a great experience with your company. The service was prompt and the staff was very friendly."],"categories": ["Customer Service, Satisfaction, Sales, Product"], "classification_instructions": ["classify the text based on the feedback provided by customer"]}}
Assistant:{{"keywords": ["recently", "great experience", "company", "service", "prompt", "staff", "friendly"],"categories": ["Customer Service"]}}
User:{{"input_text": ["bad service, slow to bring the food"],"categories": ["Food Quality, Experience, Price" ], "classification_instructions": []}}
Assistant:{{"keywords": ["recently", "great experience", "company", "service", "prompt", "staff", "friendly"],"categories": ["Customer Service"]}}{{"keywords": ["bad service", "slow", "food", "tip", "terrible", "waitresses"],"categories": ["Experience""]}}
</example>
### Memory
Here is the chat histories between human and assistant, inside <histories></histories> XML tags.
### User Input
{{"input_text" : ["{input_text}"], "class" : ["{class}"],"classification_instruction" : ["{classification_instructions}"]}}
{{"input_text" : ["{input_text}"], "categories" : ["{categories}"],"classification_instruction" : ["{classification_instructions}"]}}
### Assistant Output
"""