feat(provider): add AnonRouter

[anonrouterai]

I maintain AnonRouter and added it as a disabled built-in provider following DeepChat’s existing OpenAI-compatible provider pattern.

What changed:

• registers the AnonRouter API base URL and account/documentation links
• uses the shared API-key, OpenAI-compatible runtime for authenticated model discovery, chat, streaming, tools, and embeddings
• preserves creator/model IDs returned by the catalog
• adds the official monochrome provider icon and focused registry/discovery tests

Users need an AnonRouter key with OpenAI-compatible access enabled. No AnonRouter-specific transport was added.

Checks:

• pnpm run format
• pnpm run lint
• pnpm run i18n
• pnpm run typecheck
• pnpm exec vitest run test/main/provider/defaultProviders.test.ts test/main/provider/basicApiKeyProviders.test.ts (35 passed)

Summary by CodeRabbit

• New Features

  • Added AnonRouter as a supported provider for OpenAI-compatible model access.
  • Added provider setup links, including API key, documentation, and model pages.
  • Added AnonRouter branding in the model provider interface.
  • Added authenticated model discovery through AnonRouter’s models endpoint.

• Tests

  • Added coverage for provider configuration, model discovery, authentication, and model metadata.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: 9e5f5035-dba8-4db2-a622-039beaf86e99

📥 Commits

Reviewing files that changed from the base of the PR and between ebfe83a and 293bd6e.

⛔ Files ignored due to path filters (1)

• src/renderer/src/assets/llm-icons/anonrouter.svg is excluded by !**/*.svg

📒 Files selected for processing (5)

• src/main/provider/defaults.ts
• src/main/provider/providerRegistry.ts
• src/renderer/src/components/icons/modelIconRegistry.ts
• test/main/provider/basicApiKeyProviders.test.ts
• test/main/provider/defaultProviders.test.ts

Included review availability: Your plan provides up to 8 included reviews per hour; 7 remain after this review.

---

📝 Walkthrough

Walkthrough

The change adds AnonRouter as a disabled OpenAI-compatible provider with API-key authentication, model discovery, default metadata, renderer icon support, and test coverage.

Changes

AnonRouter provider integration

Layer / File(s)	Summary
Provider registration src/main/provider/defaults.ts, src/main/provider/providerRegistry.ts, test/main/provider/defaultProviders.test.ts	Adds AnonRouter defaults, registry settings, endpoint metadata, and coverage for the disabled provider configuration.
Authenticated model discovery test/main/provider/basicApiKeyProviders.test.ts	Verifies Bearer authentication, model discovery, nested model identifiers, ownership values, provider identity, and the models endpoint.
Provider icon registration src/renderer/src/components/icons/modelIconRegistry.ts	Registers the AnonRouter icon and includes it in monochrome icon handling.

Priority: ⬇️ Low

Estimated code review effort: 2 (Simple) | ~10 minutes

Change: Feature

Suggested reviewers: zerob13

Merge Risk: ⚪ Minimal · up to 293bd

The AnonRouter provider integration and icon registration are consistent with their runtime and renderer contracts, with no established merge-blocking risk.

🚥 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 concisely describes the main change: adding AnonRouter as a provider.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 5…
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 a new PR

---

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.}

[zerob13]

Review: Approve ✅

This PR adds AnonRouter as a disabled-by-default built-in OpenAI-compatible provider. It follows the exact same pattern as recent provider additions (cheaper-inference, RunInfra, Synthorai): a DEFAULT_PROVIDERS entry, an OPENAI_BASE + api-key registry definition, a monochrome icon wired into modelIconRegistry, and two test blocks mirroring the existing ones. 6 files, +118/−1. No over-engineering, no breaking changes, no scope creep.

Verified against the live AnonRouter API

The registry definition matches what AnonRouter's compatibility mode actually exposes:

Capability (per docs.anonrouter.ai/compatibility)	PR config	OK?
POST /v1/chat/completions (stream + non-stream)	openai-completions + OPENAI_BASE	✅
GET /v1/models (bearer call)	modelSource: 'openai', checkStrategy: 'fetch-models'	✅
POST /v1/embeddings supported	embeddingStrategy: 'openai' (default)	✅ — correct, unlike cheaper-inference/api-route which disable it

Live endpoint checks: POST /v1/chat/completions returns 401 without a key (route exists); GET /v1/models returns 401 authentication_error with an ar_-prefixed key (route exists, auth-gated — an unauthenticated GET falls through to a generic 404, so don't be fooled when testing without a key). All four website URLs (official / api-keys / docs / models) return 200.

Validation run

• vitest run test/main/provider/basicApiKeyProviders.test.ts test/main/provider/defaultProviders.test.ts → 35/35 pass
• pnpm run typecheck (node + web) → clean
• oxfmt --check on the 5 changed TS files → clean
• anonrouter.svg → valid XML, fill="currentColor" only, correctly added to monoIconUrls

Non-blocking notes

1. README provider table not updated. Precedent is mixed — Synthorai and OrcaRouter are listed in the table, RunInfra and cheaper-inference are not. If the table is meant to track notable providers, consider adding AnonRouter to the three READMEs; if it's best-effort, ignore this.
2. Inherent provider limitations worth knowing (no action needed in this PR): AnonRouter rejects array/multimodal message content and unknown fields (n > 1, logprobs, logit_bias), so sending images to an AnonRouter model will fail with a 400 from their side. This is a server-side contract, not something the client wiring can fix.
3. websites.models points to the marketing page anonrouter.ai/models; the browsable catalog with privacy labels lives at docs.anonrouter.ai/models. Either is fine — just flagging the difference.

---

Detailed verification log

• Review basis: PR head 293bd6e2c, merge-base diff against dev (6 files, +118/−1)
• Registry parity: anonrouter definition is ...OPENAI_BASE, credentialStrategy: 'api-key' — identical shape to cheaper-inference/api-route minus the embeddingStrategy: 'none' override, which is correct here because AnonRouter documents POST /v1/embeddings as supported
• Endpoint probes (2026-09-19): GET https://anonrouter.ai/ → 200; /models → 200; https://docs.anonrouter.ai/quickstart → 200; https://anonrouter.ai/home/api-keys → 200; GET api.anonrouter.ai/v1/models (no auth) → 404 catch-all, (with Bearer ar_test) → 401 authentication_error; POST api.anonrouter.ai/v1/chat/completions (no auth) → 401
• Compat-mode gating: GET /v1/auth/capabilities requires auth, so operator-level compat.enabled could not be verified without a key; docs state new keys are compatibility-enabled by default, which this integration relies on (plain bearer, no ticket flow — DeepChat cannot do the ticket dance)
• anonrouter is not in the PublicProviderConf DB (checked local + remote all.json), so modelSource: 'openai' is the right choice
• Tests: the basicApiKeyProviders block mirrors the RunInfra test (fetch mock, definition match, discovery assertions, check(), no generate-text); the defaultProviders block mirrors the cheaper-inference one. Proportionate to existing coverage — no over-testing.