flash-3.5

[Makisuo]

No description provided.

[pullfrog]

✅ No new issues found.

Reviewed changes — switches the apps/chat-flue default chat + triage model from native Workers AI to OpenRouter's openrouter/google/gemini-3.5-flash, and wires the OpenRouter provider + API key through env, secrets, and the dev-vars example.

• Default model swap — DEFAULT_MODEL (src/agents/maple-chat.ts) and DEFAULT_TRIAGE_MODEL (src/workflows/triage.ts) both move from cloudflare/@cf/... to openrouter/google/gemini-3.5-flash; MAPLE_CHAT_MODEL / MAPLE_TRIAGE_MODEL overrides and the cloudflare/* provider remain available.
• Provider registration — src/app.ts registers the openrouter catalog provider at module scope, guarded on env.OPENROUTER_API_KEY, mirroring the existing observe / setupTelemetry per-isolate registration pattern.
• Env + secret wiring — OPENROUTER_API_KEY?: string added to ChatFlueEnv; alchemy.run.ts binds it via the existing optionalSecret helper (bound only when set); .dev.vars.example documents the key.

The wiring is consistent with existing chat-flue patterns and there are no stale references — the guarded registerProvider cannot throw on missing key, and the optionalSecret binding matches the other optional secrets in alchemy.run.ts. The model id google/gemini-3.5-flash is a valid live OpenRouter model. Worth a conscious note (not a blocker, and already documented in the code comments): this flips the default from keyless native Workers AI to paid OpenRouter usage billed to OPENROUTER_API_KEY (~$1.50/$9 per 1M tokens), so the cost profile of every default chat + triage run changes.

  ^{｜ ⚠️ this action is pinned to a commit SHA, which freezes the cleanup step — switch to @v0 or keep the SHA fresh with Dependabot ｜ View workflow run ｜ Using Claude Opus ｜ 𝕏}