Enhancement: Support o1 tool-calling

[weipienlee]

See update message below.

What happened?

The o1 releases (non-preview) now support function calling but does not work with the LibreChat "plugin" mode.

• o1 chat model works in (non-plugin) chat mode.
• o1 plugin mode works with "openai service" (not Azure) with same steps described below, EXCEPT if an o1 model is selected for the agent e.g. o1-mini

Steps to Reproduce

1. select plugin from dropdown
2. choose "o1-2024-12-17" model (for some reason it doesn't list "o1" model, it does list "o1-mini" without date)
3. set temperature to 1.0
4. choose agent llm gpt-4o-mini
5. choose tavily search plugin (or google)
6. ask avatar "whats the weather in london now"

What browsers are you seeing the problem on?

No response

Relevant log output

LibreChat         | 2025-01-15 22:12:17 error: [RunManager] handleLLMError: {"context":"plugins","conversationId":"45e5cbce-14ac-4a84-98c2-ac0b583cebf4","initialMessageCount":8} 400 Unsupported value: 'messages[0].role' does not support 'system' with this model.
LibreChat         | Error in handler Handler, handleLLMError: Error: Error: 400 Unsupported value: 'messages[0].role' does not support 'system' with this model.
LibreChat         | [chain/error] [1:chain:AgentExecutor] [1.86s] Chain run errored with error: "400 Unsupported value: 'messages[0].role' does not support 'system' with this model.\n\nError: 400 Unsupported value: 'messages[0].role' does not support 'system' with this model.\n    at APIError.generate (/app/node_modules/@langchain/openai/node_modules/openai/error.js:45:20)\n    at OpenAI.makeStatusError (/app/node_modules/@langchain/openai/node_modules/openai/core.js:293:33)\n    at OpenAI.makeRequest (/app/node_modules/@langchain/openai/node_modules/openai/core.js:337:30)\n    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)\n    at async /app/node_modules/@langchain/openai/dist/chat_models.cjs:1548:29\n    at async RetryOperation._fn (/app/node_modules/p-retry/index.js:50:12)"
LibreChat         | 2025-01-15 22:12:17 error: [PluginsClient] executorCall error: 400 Unsupported value: 'messages[0].role' does not support 'system' with this model.

Screenshots

Code of Conduct

• I agree to follow this project's Code of Conduct

[weipienlee]

UPDATE, I just read in documentation that plugins is deprecated. Unfortunately, o1 models yield errors with agents despite it support function calling (unless openai really distinguish between function- and tool-calling).

LibreChat         | 2025-01-15 22:44:35 error: [api/server/controllers/agents/client.js #sendCompletion] Unhandled error type 404 tools is not supported in this model. For a list of supported models, refer to https://platform.openai.com/docs/guides/function-calling#models-supporting-function-calling.
LibreChat         | 
LibreChat         | Troubleshooting URL: https://js.langchain.com/docs/troubleshooting/errors/MODEL_NOT_FOUND/
LibreChat         | 
LibreChat         | 2025-01-15 22:44:35 error: [handleAbortError] AI response error; aborting request: 404 tools is not supported in this model. For a list of supported models, refer to https://platform.openai.com/docs/guides/function-calling#models-supporting-function-calling.
LibreChat         | 
LibreChat         | Troubleshooting URL: https://js.langchain.com/docs/troubleshooting/errors/MODEL_NOT_FOUND/

while Azure o1 models (@temp 1.0) yield:

LibreChat         | 2025-01-15 22:50:11 error: [api/server/controllers/agents/client.js #sendCompletion] Unhandled error type 400 Unsupported value: 'stream' does not support true with this model. Supported values are: false.
LibreChat         | 2025-01-15 22:50:11 error: [handleAbortError] AI response error; aborting request: 400 Unsupported value: 'stream' does not support true with this model. Supported values are: false.

seems same as previously resolved issue with normal Azure o1 completion end point.

[danny-avila]

o1 tool calling is still new and has to be implemented still, it's handled differently and the fact that streaming isn't supported deviates from the current system.

[weipienlee]

Ah tnx @danny-avila, i hoped it was just the same problem as the completion end-point, just removing the parameter and such. Looking forward for the support.