feat: upgrade MiniMax default model to M2.7#1291
Conversation
Add MiniMax LLM support via the OpenAI-compatible API, following the same pattern as the existing Qwen and DeepSeek providers. Changes: - Add MinimaxLLMConfig with api_key, api_base, extra_body fields - Add MinimaxLLM class inheriting from OpenAILLM - Register minimax backend in LLMFactory and LLMConfigFactory - Add minimax_config() to APIConfig with env var support (MINIMAX_API_KEY, MINIMAX_API_BASE) - Add minimax to backend_model dicts in product/user config - Add MiniMax example scenario in examples/basic_modules/llm.py - Add unit tests for config and LLM (generate, stream, think prefix) - Update .env.example and README with MiniMax provider info MiniMax API: https://api.minimax.io/v1 (OpenAI-compatible) Models: MiniMax-M2.5, MiniMax-M2.5-highspeed (204K context)
- Update default model from MiniMax-M2.5 to MiniMax-M2.7 in API config - Update example code to use MiniMax-M2.7 as default with M2.7-highspeed listed - Update unit tests to reference M2.7 and M2.7-highspeed models - Keep all previous models (M2.5, M2.5-highspeed) as available alternatives
There was a problem hiding this comment.
Thanks for adding MiniMax as a first-class LLM provider! The implementation is clean — inheriting from OpenAILLM is the right call, and the config/factory/test coverage all follow existing patterns nicely.
I made a small fix to keep the backup_* fields in OpenAILLMConfig — they were unintentionally removed during the dev branch merge, which broke OpenAILLM.init() for MinimaxLLM and the related tests. All 3 failing tests pass now.
Great contribution!
Enhanced MemReader configuration to support backup client and general model.
Restored backup configuration options in test_llm.py.
|
Hey @octo-patch! Thanks for the contribution — already merged your PR, thanks for putting this together! 🚀 Btw if you enjoyed contributing and want to stay connected with the project, we run a Discord server for contributors, a chill space to discuss features, ask questions, or pick up new issues 👉🏻 MemOS Discord Server (Contributor Exclusive) Thanks again for shipping this, and happy MemOSing 🧠✨ |
3 participants
Summary
Changes
minimax_config()) to MiniMax-M2.7Why
MiniMax-M2.7 is the latest flagship model with enhanced reasoning and coding capabilities, superseding M2.5 as the recommended default.
Testing