[Programcode]

Every family in the catalog now uses its newest OpenRouter model, and Claude models are the defaults: Claude Haiku 5.5 for Basic and Claude Opus 5.5 for Pro. Basic accounts get exactly three models. The catalog stays at 10.

Catalog

Basic (Free):

• anthropic/claude-haiku-5.5: default, and the model Auto Fast prefers and falls back to.
• openai/gpt-6-luna: replaces GPT-5.6 Luna.
• meta/muse-spark-1.3-contributor: the contributor tier, which replaces the regular Muse Spark 1.3. It shows as "Muse Spark 1.3". The privacy copy is unchanged, by Adam's decision.

Pro:

• anthropic/claude-opus-5.5: default, and the model Auto Pro prefers and falls back to.
• anthropic/claude-sonnet-5.5: replaces Claude Sonnet 5.
• openai/gpt-6.1-sol: replaces GPT-5.6 Sol and Terra (GPT-6 has no Terra).
• google/gemini-3.8-flash: moved from Basic to Pro.
• deepseek/deepseek-v4.1-flash: replaces DeepSeek V4 Flash and moves to Pro.
• deepseek/deepseek-v4-pro-0813 and z-ai/glm-5.3-flash: unchanged; still the newest in their family.

Auto Fast now routes only among the three Basic models, so it no longer sends Basic users to GLM.

Not added: Claude Fable 5.1 and GPT-6 Astra cost $10/$50 per million tokens. New entries carry no supportsTemperature, matching #134.

Retired IDs

A saved setting or an installed PWA that still sends an old ID keeps working. The settings store and both ai and ai-stream resolve it before validation:

• GPT-5.6 Luna and DeepSeek V4 Flash (both were Basic) become Claude Haiku 5.5.
• Claude Sonnet 5 and Claude Opus 5 become Claude Opus 5.5.
• GPT-5.6 Sol and Terra become GPT-6.1 Sol.
• Muse Spark 1.3 becomes Muse Spark 1.3 contributor.

A retired Basic ID never resolves to a Pro model; a catalog test covers that. A Basic user with Gemini saved is moved to Haiku on the next load, the same as any Pro-only model.

Deploy order

ai, ai-stream and voice went to staging and then production from the PR head (920154b) before the merge. The new frontend rewrites every saved GPT-5.6 Luna setting to anthropic/claude-haiku-5.5, which the old server rejects, so shipping the frontend first would have broken AI for most users until the functions caught up. The new functions accept every old ID through retiredModelSuccessor, so the old frontend kept working in the meantime. One exception: a Basic user with Gemini 3.8 Flash saved gets a 403 until the new frontend loads and moves them to Haiku.

The PR head was main plus this PR. After the merge, the functions are redeployed from main if backend:status flags them, so the deploy record and the staging-sync timer agree. The production rollback baseline is at ~/ops/backups/debatelab-functions/prod-20261011T1207.

Verification

• GitHub Actions did not run. Every job fails in 1–3 s with zero steps, the same as on Stop sending temperature to OpenRouter (fixes Claude Opus 5 404) #134 (Actions billing). The local results below stand in for CI and are not a CI pass.

• Local npm run verify at 45c9cff: exit 0, with 275 client tests, 334 Edge tests and 30/30 desktop smoke tests. Codex review r4 on 920154b found nothing. lint, typecheck, docs:check and build are clean. The full npm run verify and recipe results are in comments below.

• Direct OpenRouter calls with the exact payload provider.ts builds (JSON format, reasoning.effort, exclude: true, require_parameters: true, max_tokens 32768, no temperature) returned HTTP 200 with JSON for every new model at low, medium and high: Haiku, Sonnet and Opus 5.5, GPT-6 Luna, GPT-6.1 Sol, DeepSeek V4.1 Flash and Muse Spark 1.3 contributor. That is 21 of 21.

• Staging, real calls as the seeded users after the deploy:

  • Pro: Opus 5.5 returns 200.
  • Trial: Haiku 5.5 and Muse Spark contributor return 200.
  • Trial asking for Opus 5.5 gets a 403.
  • Old IDs still work: gpt-5.6-luna resolves to Haiku 5.5 and claude-sonnet-5 to Opus 5.5.
  • Auto Pro and Auto Fast (streamed) return 200.

• Production: all three functions answer CORS and return 401 without auth. The deployed model-catalog.ts is identical to the PR head. No production Pro login exists, so no authenticated production call was made.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
debate-navigator	Ready	Preview	Oct 9, 2026 5:32pm UTC
debatelab-staging	Ready	Preview	Oct 9, 2026 5:32pm UTC

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[coderabbitai]

Warning

Review limit reached

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

Next included review available in 30 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available.

Learn how review limits work.

Review configuration:

⚙️ Run configuration

• Configuration used: Repository YAML (base), Organization UI (inherited)
• Review profile: CHILL
• Plan: Free
• Run ID: e6ed2ae3-4b22-465f-b189-09e358413c13

---

📥 Commits

Reviewing files that changed from the base of the PR and between 781d603 and 920154b.

---

📒 Files selected for processing (31)

• .claude/skills/verify-debatelab/features/README.md
• .claude/skills/verify-debatelab/features/case-generation.md
• .claude/skills/verify-debatelab/features/model-defaults.md
• .claude/skills/verify-debatelab/recipes/model-defaults.spec.ts
• docs/architecture.md
• docs/testing.md
• e2e/production/canary.spec.ts
• scripts/prompt-examples.ts
• src/lib/stores/settings-store.ts
• src/locales/en.json
• src/locales/es.json
• src/locales/hu.json
• src/test/ai/streaming.test.ts
• src/test/debate/debate-context.test.tsx
• src/test/security-hardening.test.ts
• src/test/settings/settings-migration.test.ts
• supabase/README.md
• supabase/functions/_shared/auto-model-router.test.ts
• supabase/functions/_shared/auto-model-router.ts
• supabase/functions/_shared/entitlements.test.ts
• supabase/functions/_shared/model-catalog.test.ts
• supabase/functions/_shared/model-catalog.ts
• supabase/functions/ai-stream/auth_test.ts
• supabase/functions/ai-stream/handler_test.ts
• supabase/functions/ai-stream/index.ts
• supabase/functions/ai-stream/index_test.ts
• supabase/functions/ai/debate-context.test.ts
• supabase/functions/ai/generate-honesty_test.ts
• supabase/functions/ai/handler-research.test.ts
• supabase/functions/ai/index.ts
• supabase/functions/ai/voice.test.ts

---

---

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

---

Note

🎁 Summarized by CodeRabbit Free

Your organization is on the Free plan. CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please upgrade your subscription to CodeRabbit Essentials by visiting https://app.coderabbit.ai/settings/billing.

---

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

[Programcode]

Codex review r1 (head 6f556bb) found one issue: the model-defaults recipe checked the upstream model, which live verify rewrites to the Muse Spark pin. Fixed in 30e9ccf: the recipe now checks requestedModel ?? model. The recipe hasn't been run yet because the shared local stack is held by another worktree. 3b04533 also moves the rest of the catalog to each family's newest model; the PR description has the details.

[Programcode]

Codex review r3 (head d0894e3) found no entitlement or retired-ID problems. It raised one low finding: docs/testing.md still said the Muse Spark contributor model is test-only and not in the app catalog. Fixed in 920154b; the doc now says it is also the Basic "Muse Spark 1.3" model, and recipes check the requested model. Since r3, the branch also merged main (#134) and switched the settings cleanup test to current models (45c9cff). r4 reviews the new head.

[Programcode]

Local npm run verify passed at 45c9cff, exit 0: docs check, lint, typecheck, npm test 275/275, collaboration tests, test:edge 334/334, build, service-worker test and desktop Chromium @smoke 30/30. The only commit since then, 920154b, changes docs. This is a local pass, not CI: GitHub Actions jobs still fail in 1–3 s with zero steps (billing). The model-defaults recipe is still waiting for the shared local stack, which another worktree holds.

[Programcode]

recipes/model-defaults passed 6/6 on the fake-mode verify stack at 920154b, desktop-chromium and mobile-chromium:

• model-pro-default: a new Pro account's settings and selector start on Claude Opus 5.5, and Generate records anthropic/claude-opus-5.5 in ai_requests.
• model-basic-default: a Basic account starts on Claude Haiku 5.5. GPT-6 Luna and Muse Spark 1.3 are enabled; Opus, Sonnet and Gemini are disabled. Generate records Haiku 5.5.
• model-retired: saved anthropic/claude-sonnet-5 configs are rewritten to Opus 5.5 on screen and in user_settings.