[Zhangbo-cn]

What

Adds the 月之暗面 Kimi / Moonshot Kimi chat preset requested in #240 (mainland chat row).

• src/canvas/providerPresets.ts: new kimi preset — kinds: ["agent"], url: https://api.moonshot.cn/v1, openai + api + bearer, logoOrganization: "moonshotai", English name Moonshot Kimi. Placed after bailian-beijing, before zhipu.
• No new adapter or endpoint code is needed: Moonshot is OpenAI-compatible, so the existing agent transport drives it unchanged.
• modelsByKind is deliberately omitted, so Fetch models stays a live /models call and the picker lists whatever the entered key can actually reach (same approach as deepseek). With a mainland key that is kimi-k3 and kimi-k3.1-pro.
• The international endpoint (api.moonshot.ai) is a separate row of the [tracking] providers: API integrations #240 matrix and is intentionally not matched — it stays a custom OpenAI-compatible service. The docs paragraph says so.
• Logo: nothing to add, moonshotai already resolves through modelBrands.json → kimi.svg.
• docs/guides/user-guide.md + user-guide.zh-CN.md: one paragraph per locale (preset, live discovery, international endpoint → custom service).
• tests/providerPresets.test.ts: 2 tests (preset shape + discovery-stays-on + localized labels + kind filtering; and providerPresetFor() re-selecting kimi for saved mainland connections while .ai, deeper paths and other kinds fall back to custom).
• tests/browser/kimi_real.mjs + manifest.json: an opt-in real-endpoint browser check (helper, needs MOONSHOT_API_KEY, incurs charges, no lane runs it).

How tested

Gates on the rebased branch:

./node_modules/.bin/tsc --noEmit      # clean
node scripts/check-i18n.mjs           # 3011 keys, 0 problems
node scripts/check-browser-manifest.mjs  # 75 entries OK
npm test                              # 810 pass, 0 fail
scripts/generate_contracts.py --check # verified

Both new tests were mutation-checked: changing the preset URL to .ai fails 2 assertions, adding modelsByKind fails 1, restoring passes 20/20.

Real endpoint against a locally run app (isolated creation server + vite dev server), with a real mainland key:

MOONSHOT_API_KEY=… OPENDIRECTOR_URL=http://127.0.0.1:5190 \
OPENDIRECTOR_API=http://127.0.0.1:8040 node tests/browser/kimi_real.mjs

The script asserts, not just screenshots: /api/model-configs/discover returns kimi-k3, the streamed reply arrives in the UI, the agent-written asset comes back with provenance.source == "agent_written" and modelSnapshot.modelId == "kimi-k3", it appears in the asset list, adds a canvas node, and survives a reload. Browser requests to moonshot.(cn|ai) are asserted never to happen — the key stays server side.

Evidence

Preset (zh-CN)
Preset (en-US)
Streamed reply
Asset list
Canvas node
After reload

Notes

• Prompt used in the run: write a text asset named kimi-note.txt inside the project. The agent completed it through opendirector asset write-text, so the whole tool-calling loop works over Kimi.
• kimi-k3 is slow-ish for a full tool round trip (~2 min for 4 calls in my run), which is why the browser check allows long waits; that is the model, not the integration.
• Thanks @FayeSpica for the assignment. Happy to adjust the row's wording in [tracking] providers: API integrations #240, or add the international endpoint here if you'd rather keep both Kimi rows in one preset.