feat: add MiniMax as AI provider for completion and embedding

[octo-patch]

Summary

Add MiniMax as a first-class AI provider in seekdb's AI service framework, enabling users to use MiniMax's LLM and embedding models for in-database AI workflows.

Changes

• Provider registration: Add MINIMAX to VALID_PROVIDERS[] in ob_ai_service_struct.cpp
• Provider constant: Add MINIMAX to ObAIFuncProviderUtils in ob_ai_func_utils.h
• LLM completion: Route MiniMax completion to OpenAI-compatible handler (MiniMax /v1/chat/completions API is fully OpenAI-compatible)
• Embedding: Add ObMiniMaxUtils::ObMiniMaxEmbed custom handler for MiniMax native embedding API (embo-01, 1536 dimensions), which uses texts instead of input and vectors instead of data in request/response
• Unit tests: 12 tests in test_ob_minimax_utils.cpp covering header, body construction, response parsing, edge cases, and provider routing
• Integration test: MiniMax endpoint CRUD test in test_ai_service.cpp
• Documentation: Update README.md and README_CN.md to list MiniMax as a supported AI provider

MiniMax Models

Type	Model	Description
LLM	MiniMax-M2.7	Latest model, 1M context
LLM	MiniMax-M2.5	Previous generation
LLM	MiniMax-M2.5-highspeed	Fast inference, 204K context
Embedding	embo-01	1536 dimensions

Usage Example

-- Register MiniMax LLM endpoint
CALL DBMS_AI_SERVICE.CREATE_AI_MODEL_ENDPOINT('minimax_llm', '{
  "url": "https://api.minimax.io/v1/chat/completions",
  "access_key": "your-minimax-api-key",
  "ai_model_name": "minimax_model",
  "provider": "MINIMAX",
  "request_model_name": "MiniMax-M2.7"
}');

-- Register MiniMax embedding endpoint
CALL DBMS_AI_SERVICE.CREATE_AI_MODEL_ENDPOINT('minimax_embed', '{
  "url": "https://api.minimax.io/v1/embeddings",
  "access_key": "your-minimax-api-key",
  "ai_model_name": "minimax_embed_model",
  "provider": "MINIMAX",
  "request_model_name": "embo-01"
}');

Test Plan

• Unit tests pass: test_ob_minimax_utils (12 tests covering header/body/parse/routing)
• Integration test: test_minimax_ai_model_endpoint in test_ai_service (endpoint CRUD)
• Existing tests unaffected (no changes to other provider logic)
• MiniMax completion works with MiniMax-M2.7 model
• MiniMax embedding works with embo-01 model

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

octo-patch seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[hnwyllmm]

Hi @octo-patch , thanks for your contribution!
I checked minimaxi document and it says minimaxi provides an openai compatibility interface. So we can use openai based function to connect to minimaxi?

[octo-patch]

Thanks for the review, @hnwyllmm! Yes, you're absolutely right — MiniMax provides an OpenAI-compatible interface at https://api.minimax.io/v1. I can refactor the implementation to use the existing OpenAI-based function with MiniMax's base URL as a custom endpoint, which would be cleaner and reduce code duplication.

I'll also look into signing the CLA. Would you like me to push the refactored version first?

[hnwyllmm]

Sorry for late response. I think it's better to use openai based implementation, such as HUNYUAN and ALIYUN. Below are the example code for HUNYUAN and ALIYUN:

The code is located in src/sql/engine/expr/ob_expr_ai/ob_ai_func_utils.cpp ObAIFuncUtils::get_complete_provider.