Add MiniMax (https://platform.minimaxi.com/) as a built-in chat model
provider with M2.7, M2.7-highspeed, M2.5, and M2.5-highspeed models.
MiniMax provides OpenAI-compatible API, so it works seamlessly with
the existing OpenAIBase and ChatOpenAI integration paths. Users only
need to set MINIMAX_API_KEY to start using MiniMax models.
Changes:
- Add MiniMax to DEFAULT_CHAT_MODEL_PROVIDERS (models.py)
- Add MINIMAX_API_KEY to .env.template
- Add MiniMax provider icon and frontend registration
- Update model-config docs with MiniMax entry
- Add 22 unit tests + 3 integration tests
Root cause: DashScope embeddings accepts at most 10 inputs per request, but code used fixed batch_size=40, causing 400 Bad Request during indexing.
Changes:
- add batch_size to EmbedModelInfo with default 40
- set dashscope/text-embedding-v4 batch_size to 10
- persist batch_size in BaseEmbeddingModel and use it as default in batch/abatch encode
- update Milvus, evaluation, and upload-graph embedding paths to read model-configured batch_size
