mirror of
https://github.com/danny-avila/LibreChat.git
synced 2026-02-11 12:04:24 +01:00
🚀 feat: Agent Cache Tokens & Anthropic Reasoning Support (#6098)
* fix: handling of top_k and top_p parameters for Claude-3.7 models (allowed without reasoning) * feat: bump @librechat/agents for Anthropic Reasoning support * fix: update reasoning handling for OpenRouter integration * fix: enhance agent token spending logic to include cache creation and read details * fix: update logic for thinking status in ContentParts component * refactor: improve agent title handling * chore: bump @librechat/agents to version 2.1.7 for parallel tool calling for Google models
This commit is contained in:
parent
34f967eff8
commit
9802629848
11 changed files with 187 additions and 40 deletions
|
|
@ -1,6 +1,6 @@
|
|||
const { HttpsProxyAgent } = require('https-proxy-agent');
|
||||
const { anthropicSettings, removeNullishValues } = require('librechat-data-provider');
|
||||
const { checkPromptCacheSupport, getClaudeHeaders } = require('./helpers');
|
||||
const { checkPromptCacheSupport, getClaudeHeaders, configureReasoning } = require('./helpers');
|
||||
|
||||
/**
|
||||
* Generates configuration options for creating an Anthropic language model (LLM) instance.
|
||||
|
|
@ -49,13 +49,14 @@ function getLLMConfig(apiKey, options = {}) {
|
|||
clientOptions: {},
|
||||
};
|
||||
|
||||
requestOptions = configureReasoning(requestOptions, systemOptions);
|
||||
|
||||
if (!/claude-3[-.]7/.test(mergedOptions.model)) {
|
||||
if (mergedOptions.topP !== undefined) {
|
||||
requestOptions.topP = mergedOptions.topP;
|
||||
}
|
||||
if (mergedOptions.topK !== undefined) {
|
||||
requestOptions.topK = mergedOptions.topK;
|
||||
}
|
||||
requestOptions.topP = mergedOptions.topP;
|
||||
requestOptions.topK = mergedOptions.topK;
|
||||
} else if (requestOptions.thinking == null) {
|
||||
requestOptions.topP = mergedOptions.topP;
|
||||
requestOptions.topK = mergedOptions.topK;
|
||||
}
|
||||
|
||||
const supportsCacheControl =
|
||||
|
|
|
|||
|
|
@ -109,4 +109,45 @@ describe('getLLMConfig', () => {
|
|||
// Just verifying that the promptCache setting is processed
|
||||
expect(result.llmConfig).toBeDefined();
|
||||
});
|
||||
|
||||
it('should include topK and topP for Claude-3.7 models when thinking is not enabled', () => {
|
||||
// Test with thinking explicitly set to null/undefined
|
||||
const result = getLLMConfig('test-api-key', {
|
||||
modelOptions: {
|
||||
model: 'claude-3-7-sonnet',
|
||||
topK: 10,
|
||||
topP: 0.9,
|
||||
thinking: false,
|
||||
},
|
||||
});
|
||||
|
||||
expect(result.llmConfig).toHaveProperty('topK', 10);
|
||||
expect(result.llmConfig).toHaveProperty('topP', 0.9);
|
||||
|
||||
// Test with thinking explicitly set to false
|
||||
const result2 = getLLMConfig('test-api-key', {
|
||||
modelOptions: {
|
||||
model: 'claude-3-7-sonnet',
|
||||
topK: 10,
|
||||
topP: 0.9,
|
||||
thinking: false,
|
||||
},
|
||||
});
|
||||
|
||||
expect(result2.llmConfig).toHaveProperty('topK', 10);
|
||||
expect(result2.llmConfig).toHaveProperty('topP', 0.9);
|
||||
|
||||
// Test with decimal notation as well
|
||||
const result3 = getLLMConfig('test-api-key', {
|
||||
modelOptions: {
|
||||
model: 'claude-3.7-sonnet',
|
||||
topK: 10,
|
||||
topP: 0.9,
|
||||
thinking: false,
|
||||
},
|
||||
});
|
||||
|
||||
expect(result3.llmConfig).toHaveProperty('topK', 10);
|
||||
expect(result3.llmConfig).toHaveProperty('topP', 0.9);
|
||||
});
|
||||
});
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue