mirror of
https://gitee.com/mateos/mateclaw.git
synced 2026-09-13 03:13:41 +08:00
86 lines
3.5 KiB
Java
86 lines
3.5 KiB
Java
package vip.mate.llm.model;
|
|
|
|
import lombok.AllArgsConstructor;
|
|
import lombok.Data;
|
|
import lombok.NoArgsConstructor;
|
|
|
|
@Data
|
|
@NoArgsConstructor
|
|
@AllArgsConstructor
|
|
public class ModelInfoDTO {
|
|
private String id;
|
|
private String name;
|
|
|
|
/**
|
|
* Discovery probe result. true = passed runtime-protocol ping test,
|
|
* false = ping failed (listed by provider but unusable at runtime,
|
|
* e.g. DashScope compatible-mode may list models the native SDK rejects).
|
|
* null = not probed (probe disabled or still pending).
|
|
*/
|
|
private Boolean probeOk;
|
|
|
|
/** Reason text when probeOk=false (short, suitable for UI badge tooltip) */
|
|
private String probeError;
|
|
|
|
/**
|
|
* RFC-049 PR-1-UI (narrow): whether this model's {@code ModelFamily} accepts the
|
|
* OpenAI {@code reasoning_effort} parameter specifically. True <em>only</em> for
|
|
* the OpenAI reasoning family (gpt-5, o1, o3, o4 variants). Retained for callers
|
|
* that need to know parameter-level compatibility; the UI "deep thinking" toggle
|
|
* should use {@link #supportsThinking} instead.
|
|
*/
|
|
private boolean supportsReasoningEffort;
|
|
|
|
/**
|
|
* RFC-049 PR-1-UI (broad): whether this model supports <em>any</em> form of
|
|
* deep thinking — either OpenAI-style via {@code reasoning_effort}, provider-
|
|
* native (Kimi K2.x, DeepSeek-Reasoner, qwen-thinking), or Anthropic extended
|
|
* thinking (Claude family). This is what the UI "deep thinking" toggle reads.
|
|
*
|
|
* <p>Derived from the model name so every construction site stays consistent.
|
|
*/
|
|
private boolean supportsThinking;
|
|
|
|
/**
|
|
* Explicit per-model input window from {@code mate_model_config}; null when
|
|
* the operator has not set one (stored as 0). This is what the management
|
|
* UI's input binds to — an empty field means "let the server decide".
|
|
*/
|
|
private Integer maxInputTokens;
|
|
|
|
/**
|
|
* The window context budgeting would use if a turn ran right now, computed
|
|
* without any probe traffic: explicit config, else the built-in window
|
|
* table, else the global default. Display only.
|
|
*/
|
|
private Integer effectiveMaxInputTokens;
|
|
|
|
/**
|
|
* Where {@link #effectiveMaxInputTokens} came from — {@code configured},
|
|
* {@code catalog} or {@code default} — so the UI can say why a number is
|
|
* what it is instead of presenting a guess as configuration.
|
|
*/
|
|
private String maxInputTokensSource;
|
|
|
|
public ModelInfoDTO(String id, String name) {
|
|
this.id = id;
|
|
this.name = name;
|
|
this.supportsReasoningEffort = ModelFamily.detect(id).supportsReasoningEffort();
|
|
this.supportsThinking = computeSupportsThinking(id);
|
|
}
|
|
|
|
/**
|
|
* Thinking capability at the product level (what the UI toggle should reflect):
|
|
* any model family that has a thinking mode, regardless of how it is triggered
|
|
* (parameter vs. model-native vs. Anthropic extended thinking).
|
|
*/
|
|
private static boolean computeSupportsThinking(String modelName) {
|
|
if (modelName == null || modelName.isBlank()) return false;
|
|
ModelFamily family = ModelFamily.detect(modelName);
|
|
if (family.isThinking()) return true; // OPENAI_REASONING / KIMI_THINKING / DEEPSEEK_REASONER / GENERIC_THINKING
|
|
// Anthropic Claude supports extended thinking via AnthropicChatOptions.thinking;
|
|
// ModelFamily doesn't model non-OpenAI-compatible providers so match by name.
|
|
return modelName.toLowerCase().contains("claude");
|
|
}
|
|
}
|