feat(providers): add MiniMax AI provider support

[ximiximi423]

Add MiniMax as a new LLM provider, enabling users to use MiniMax models through the OpenAI-compatible API endpoint.

---

Summary by cubic

Added MiniMax as a new chat provider via an OpenAI-compatible endpoint. Streaming now strips tags across chunk boundaries, flushes buffered output on stream end, and preserves tool call updates; added vitest tests with coverage.

• New Features

  • Provider key minimax; chat models MiniMax-M2.5 and MiniMax-M2.5-highspeed.
  • Config: API Key (MINIMAX_API_KEY) and Base URL (MINIMAX_BASE_URL, default https://api.minimax.io/v1).
  • Streams and object generation strip … and repair JSON when needed.
  • Added vitest and @vitest/coverage-v8 with test scripts (test, test:run, test:coverage) and 18 unit tests for streaming logic.

• Bug Fixes

  • Streaming: chunk-safe stripping, flush remaining buffer at end, discard incomplete think content.
  • Preserve toolCallChunk updates even when no visible text is emitted.

^{Written for commit 960b426. Summary will update on new commits.}

[cubic-dev-ai]

2 issues found across 3 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="src/lib/models/providers/minimax/minimaxLLM.ts">

<violation number="1" location="src/lib/models/providers/minimax/minimaxLLM.ts:75">
P2: Streaming think-tag stripping is not chunk-boundary safe: when a `</think>` is split across chunks, the partial delimiter is discarded and `insideThinkTag` stays true, suppressing all subsequent output.</violation>

<violation number="2" location="src/lib/models/providers/minimax/minimaxLLM.ts:86">
P2: streamText drops toolCallChunk updates when no visible output is emitted, so function-call deltas can be lost if a chunk contains only tool call data.</violation>
</file>

---

Since this is your first cubic review, here's how it works:

• cubic automatically reviews your code and comments on bugs and improvements
• Teach cubic by replying to its comments. cubic learns from your replies and gets better over time
• Add one-off context when rerunning by tagging @cubic-dev-ai with guidance or docs links (including llms.txt)
• Ask questions if you need clarification on any suggestion

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

[cubic-dev-ai]

1 issue found across 1 file (changes from recent commits).

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="src/lib/models/providers/minimax/minimaxLLM.ts">

<violation number="1" location="src/lib/models/providers/minimax/minimaxLLM.ts:89">
P2: Trailing content can be dropped because the buffered tail is never flushed when the stream ends; on `chunk.done`, the remaining buffer is discarded and the final characters are lost.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

[xkonjin]

Quick review pass:

• Main risk area here is auth/session state and stale credential handling.
• I didn’t see targeted regression coverage in the diff; please add or point CI at a focused test for the changed path in index.ts, index.ts, minimaxLLM.ts.
• Before merge, I’d smoke-test the behavior touched by index.ts, index.ts, minimaxLLM.ts with malformed input / retry / rollback cases, since that’s where this class of change usually breaks.

[ximiximi423]

Thanks for the review @xkonjin ! I have added the fix for buffer flushing on stream end and unit tests covering think-tag stripping, chunk boundary handling, and toolCallChunk preservation. Please review it when you have a moment.