feat(llm): add MiniMax as an authenticated cloud LLM provider

[octo-patch]

Reason: Add MiniMax as an authenticated cloud LLM provider (MiniMax-M3 / MiniMax-M2.7) with global and CN OpenAI-compatible endpoints.

What

Adds MiniMax as a first-class authenticated cloud LLM provider, wired into the same selection, key-resolution, model-default, and adapter seams as the existing providers.

• Provider type (types.ts): LLMProvider now includes minimax.
• Selection (select.ts): minimax added to the auto-detect order (lowest priority) with MINIMAX_API_KEY as its env var; WIGOLO_LLM_PROVIDER=minimax forces it explicitly.
• Model defaults (model-select.ts): default model MiniMax-M3, overridable via WIGOLO_LLM_MODEL_MINIMAX. MiniMax-M2.7 is selectable via that var or a per-call override.
• Adapters: new minimax.ts provides the structured (extract) adapter and the OpenAI-compatible endpoint resolution; a matching free-form text adapter is added in text-adapters.ts and registered in the text-adapter map, and the structured adapter is registered in the extract fallback map.
• Endpoints: two OpenAI-compatible regional base URLs are supported — global (https://api.minimax.io/v1, default) and mainland China (https://api.minimaxi.com/v1). Region is chosen with WIGOLO_MINIMAX_REGION (global_en | cn_zh) or the base URL is overridden wholesale with WIGOLO_MINIMAX_BASE_URL.
• Key resolution (key-store.ts): minimax is added to the keystore providers so its key resolves through the keychain → file → env chain. Like the other env-supported-but-not-wizard-surfaced provider, it is intentionally not added to the TUI picker.

MiniMax is authenticated: the SDK attaches an Authorization: Bearer header from the resolved key. This is distinct from the existing keyless custom-URL backend, which sends no Authorization header and remains unchanged.

The structured adapter uses response_format: { type: 'json_object' } plus local schema validation with one retry, because the endpoint does not implement strict server-side JSON-schema enforcement.

Tests

• New tests/unit/extraction/llm/minimax.test.ts: regional base-URL resolution (default/global/CN/override), model metadata (default + both models), structured call parsing, model override, json_object usage, validation retry, and error paths.
• tests/unit/extraction/llm/select.test.ts: added auto-detect and explicit-override cases for the new provider.
• tests/unit/extraction/llm/types.test.ts: updated the provider-union assertion.

Checks run

• npm run lint (tsc --noEmit) — passed
• npm run build (tsup + tsc) — passed
• npx vitest run tests/unit/extraction/llm tests/unit/security/key-store.test.ts tests/unit/cli/tui/actions/provider-actions.test.ts — passed (99 tests)

Summary by CodeRabbit

• New Features

  • Added MiniMax as a supported LLM provider for structured and free-text extraction.
  • Supports MiniMax-M3 and MiniMax-M2.7 models.
  • Added automatic provider detection and configurable regional endpoints.
  • Added schema validation with an automatic retry for invalid responses.

• Tests

  • Added coverage for MiniMax configuration, model selection, extraction, validation, and provider prioritization.

[coderabbitai]

📝 Walkthrough

Walkthrough

Adds MiniMax as an LLM provider with region-aware endpoints, model and API-key selection, JSON extraction with schema-validation retry, free-text completion support, fallback dispatch, and unit tests.

Changes

MiniMax provider integration

Layer / File(s)	Summary
Provider contracts and selection src/integrations/cloud/llm/types.ts, src/integrations/cloud/llm/model-select.ts, src/integrations/cloud/llm/select.ts, src/security/key-store.ts, tests/unit/extraction/llm/select.test.ts, tests/unit/extraction/llm/types.test.ts	Adds minimax to provider types, model and API-key configuration, persisted-key discovery, and provider selection coverage.
MiniMax extraction adapter src/integrations/cloud/llm/minimax.ts, src/extraction/llm-fallback.ts, tests/unit/extraction/llm/minimax.test.ts	Implements region-aware OpenAI-compatible JSON extraction, schema validation with one retry, standardized results, fallback dispatch, and adapter tests.
MiniMax text completion adapter src/integrations/cloud/llm/text-adapters.ts	Adds and registers callMiniMaxText for free-text completions using the resolved MiniMax endpoint.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Suggested reviewers: knockoutez

Sequence Diagram(s)

sequenceDiagram
  participant extractWithLLM
  participant callMiniMax
  participant MiniMaxAPI
  participant SchemaValidator
  extractWithLLM->>callMiniMax: extraction options and API key
  callMiniMax->>MiniMaxAPI: JSON-object completion request
  MiniMaxAPI-->>callMiniMax: assistant JSON content
  callMiniMax->>SchemaValidator: validate parsed values
  SchemaValidator-->>callMiniMax: validation result
  callMiniMax->>MiniMaxAPI: retry prompt with validation errors
  MiniMaxAPI-->>callMiniMax: corrected assistant JSON content
  callMiniMax-->>extractWithLLM: standardized extraction result

Loading

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and accurately summarizes the main change: adding MiniMax as an authenticated cloud LLM provider.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

🧹 Nitpick comments (1)

src/integrations/cloud/llm/minimax.ts (1)

96-121: 🚀 Performance & Scalability | 🔵 Trivial | ⚡ Quick win

Consider capping max_tokens in the extraction call.

runOnce never sets max_tokens/max_completion_tokens, unlike callMiniMaxText in text-adapters.ts which defaults to DEFAULT_MAX_TOKENS. Without a cap, a runaway or verbose model response (across the initial call and the retry) could inflate latency/cost unexpectedly.

♻️ Proposed fix

   const response = await client.chat.completions.create(
     {
       model,
       messages,
       response_format: { type: 'json_object' },
+      max_tokens: 2000,
     },
     { signal },
   );

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/integrations/cloud/llm/minimax.ts` around lines 96 - 121, Update the
request parameters in runOnce to set the same bounded token limit used by
callMiniMaxText, reusing DEFAULT_MAX_TOKENS or the established equivalent.
Ensure the cap applies to both the initial extraction call and its retry without
changing the existing response parsing behavior.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@src/integrations/cloud/llm/minimax.ts`:
- Around line 96-121: Update the request parameters in runOnce to set the same
bounded token limit used by callMiniMaxText, reusing DEFAULT_MAX_TOKENS or the
established equivalent. Ensure the cap applies to both the initial extraction
call and its retry without changing the existing response parsing behavior.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 6a69c558-bf65-40c5-afea-5fe25d248bd8

📥 Commits

Reviewing files that changed from the base of the PR and between 56da0c8 and b05db8b.

📒 Files selected for processing (10)

• src/extraction/llm-fallback.ts
• src/integrations/cloud/llm/minimax.ts
• src/integrations/cloud/llm/model-select.ts
• src/integrations/cloud/llm/select.ts
• src/integrations/cloud/llm/text-adapters.ts
• src/integrations/cloud/llm/types.ts
• src/security/key-store.ts
• tests/unit/extraction/llm/minimax.test.ts
• tests/unit/extraction/llm/select.test.ts
• tests/unit/extraction/llm/types.test.ts

[KnockOutEZ]

Hi @octo-patch — thank you for this PR, and I'm sorry it's been sitting here without a proper review.

I've been heads-down building some pretty big features and improvements for wigolo, and as a solo dev, shipping that while also reviewing and maintaining every issue and PR has been genuinely hard. This one isn't forgotten — it just hasn't had the attention it deserves yet.

I'd really appreciate your patience here. I'm going to work through the open PRs and issues properly over the next few weeks (sooner if I can free up), and I'll follow up right here.

In the meantime, the wigolo Discord is open if you'd like to follow what's being built, ask questions, or nudge me directly: https://discord.gg/BkUUgz2bNF

Thanks again for contributing, and for understanding — it genuinely means a lot. 🙏