feat(llm): add built-in Mistral AI provider preset

[xavierpestel-ai]

Note

Draft companion to #780. Not requesting review until maintainers confirm the direction on the discovery issue. Opened as a draft so the concrete diff is available for inspection alongside the proposal.

Description

Adds Mistral AI as a built-in LLM provider preset, so that users would be able to select it with:

ocr config set provider mistral
ocr config set model codestral-latest
export MISTRAL_API_KEY=...
ocr review

Design

Mistral's Chat Completions endpoint at https://api.mistral.ai/v1 is OpenAI-compatible (verified — see the testing section below), so this is a registry-only addition using the existing ProtocolOpenAIChatCompletions. It mirrors the pattern of the recently merged provider presets: Eden AI (#346), Ollama Cloud (#375), and LiteLLM (#385).

Scope guardrails

No changes to any of:

• internal/llm/client.go
• internal/llm/protocol.go
• internal/llm/resolver.go
• cmd/opencodereview/config_cmd.go
• cmd/opencodereview/provider_cmd.go
• action.yml
• go.mod / go.sum

The diff is one new entry in internal/llm/providers.go and one new test in internal/llm/providers_test.go. Total: 52 additions, 1 deletion.

Model list rationale

Kept deliberately minimal — the preset only seeds the interactive picker; users can select any Mistral model via ocr config set model <name>.

• codestral-latest
• mistral-large-latest
• mistral-small-latest

Disclosure

I work at Mistral AI. This is a personal contribution submitted under the same preset pattern as PRs #346, #375, #385.

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Refactoring (no functional changes)
• Documentation update
• CI / Build / Tooling

How Has This Been Tested?

• make test passes locally
• Manual testing (described below)

Local verification

Check	Result
make check (license, gofmt, go vet, go mod tidy)	passes
go test -race ./internal/llm/	passes
TestLookupProvider_MistralDetails (new)	passes
TestListProviders_Order (updated to include mistral in sorted position)	passes
TestProviders_AllProtocolsCanonical	passes
make build	passes
internal/llm coverage	94.3 % overall, 100 % on providers.go

Live smoke test against api.mistral.ai

POST https://api.mistral.ai/v1/chat/completions with an OpenAI-shape request body returned an OpenAI-shape response containing id, object, created, model, choices[0].message.{role,content}, finish_reason, usage.{prompt_tokens,completion_tokens,total_tokens}, and tool_calls. The existing OpenAIClient parses this without modification.

End-to-end ocr review against Mistral

Built the ocr binary with this change, set provider=mistral, model=codestral-latest, ran ocr review on a scratch git repo whose staged diff introduced a SQL injection (db.Exec("DELETE FROM users WHERE id = " + id)). The review produced a single finding at the correct line with severity security · high and a parameterized-query suggested fix. Command exited cleanly.

Paths not exercised

To be transparent about what was and was not verified:

• Streaming responses — the smoke-test diff was below the threshold that triggers streaming. Mistral's SSE format is documented as OpenAI-compatible, and other OpenAI-compatible presets in this repo already exercise the streaming path, but I did not personally observe a streamed response from Mistral in this run.
• Only codestral-latest was exercised end-to-end. mistral-large-latest and mistral-small-latest were only exercised at the registry-lookup level.

Checklist

• My code follows the project's coding style (go fmt, go vet)
• I have performed a self-review of my code
• I have added tests that prove the feature works
• New and existing unit tests pass locally with my changes
• I have updated the documentation accordingly (if applicable) — no README changes needed; README does not enumerate providers (matches PRs feat(llm): add built-in Eden AI provider preset #346, feat(llm): add built-in Ollama Cloud provider preset #375, feat: add LiteLLM AI gateway provider preset #385)
• I have signed the CLA — will sign on bot prompt

Related Issues

Companion to #780 (direction check). Do not auto-close on merge — #780 is for direction confirmation, not a bug to close.

[CLAassistant]

All committers have signed the CLA.

[github-actions]

✅ OpenCodeReview: Review complete: 0 finding(s) across 1 selected item(s).

[lizhengfeng101]

LGTM