problem · Revision 1
## Observed symptom [Bug] Bedrock: Claude Opus 5.5 structured output (generateObject) fails with 400 "tools.0.custom.strict: Extra inputs are not permitted" Thanks for shipping the Opus 5.5 rules in #19867 so soon after launch. canary `d322793` (2026-09-23) ## Environment GitHub
problem · Revision 1
## Observed symptom [BUG] >=2.1.154 breaks `--resume` with `ANTHROPIC_CUSTOM_HEADERS` set in settings. - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This
problem · Revision 1
## Observed symptom Bedrock: every docs example model ID fails (400 invalid / 404 end-of-life) — use the Claude Opus 5.5 inference profile Thanks for the `[1m]` docs and for surfacing `contextWindow` in the result output (#1608). I used that
problem · Revision 1
## Observed symptom Bug: the server crashed with a segfault inside nvngx_dlss.dll during an SR video job (27th job of the process, no cancel involved) During a series of back-to-back server jobs, the server process crashed with a segmentation fault
problem · Revision 1
## Observed symptom setup (list_environments, doctor) On a corporate network that performs TLS inspection (Zscaler), every FlowAgent tool that makes an outbound HTTPS request fails with an unhelpful `{"code":"tool-error","message":"fetch failed","errorName":"TypeError"}`. ## Lite
problem · Revision 1
## Observed symptom Hangs between stages on docker-windows on Windows 11 with KB5124008 <!-- Briefly summarize the bug --> So after KB5124008 (and now, KB5129195), our Windows docker jobs got a lot slower. It seems that for some reason there's
problem · Revision 1
## Observed symptom docker-autoscaler prunes active instances mid-job on v19.3.1, independent of token rotation A user on GitLab Runner v19.3.1 with the AWS fleeting plugin reports docker-autoscaler pruning active instances while jobs are still running, producing SSH tunnel EOF e
problem · Revision 1
## Observed symptom Regression in 12.3.0: bin shim baked as 'exec node <target>' before native binary replaces the target, breaking packageManager delegation 12.3.0 and 12.3.1 (regression from 12.2.1) Link to the code that reproduces this issue ## Environment GitHub pnpm/pnpm
problem · Revision 1
## Observed symptom prisma db seed (ts-node) fails on the generated Prisma 7 client `npx prisma db seed` (and therefore `npm run db:reset`) fails: `Running seed command `ts-node prisma/seed.ts` ...` ## Literal error text ``` Error: Cannot find module './internal/class.js'
problem · Revision 1
## Observed symptom Prisma Ui related crashes After playing for a while, the game will CTD when opening either the prisma chat ui or the SN prisma ui opening either the prisma chat ui or the SN prisma ui, after
problem · Revision 1
## Observed symptom supabase db reset intermittently fails: temporary Realtime init job races with the running Realtime service on "schema_migrations_pkey" **CLI version:** 2.118.0 (Linux, Docker Engine, self-hosted GitHub Actions runner), Postgres 17 local stack. `supabase db re
problem · Revision 1
## Observed symptom Functions set_auth leaves stale credentials when Authorization uses different casing `SyncFunctionsClient.set_auth()` and `AsyncFunctionsClient.set_auth()` leave an existing lowercase or mixed-case Authorization header untouched and add a second credential und
problem · Revision 1
## Observed symptom [bug] All Edge Functions Redeploy Instead of Just Modified Ones <!-- Please fill in all fields in English --> What happened (required) There seems to be a regression where all of the edge functions are being re-deployed
problem · Revision 1
## Observed symptom google_model_profile sends invalid MINIMAL thinking_level for gemini-3.5-pro / gemini-3.1-pro and fails to strip Vertex regional prefixes `google_model_profile` in `pydantic_ai/profiles/google.py` currently has two issues with Gemini 3 / 3.5 and Vertex AI regi
problem · Revision 1
## Observed symptom [Bug] Gemini 3.x: ToolResult media still causes 400 "Requests ending with a model turn are not supported" — #9647's nesting never runs in the agent loop When a tool returns media (`ToolResult(images=...)`, `files=`, `audios=`, `videos=`), the next
problem · Revision 1
## Observed symptom [Bug]: Gemini /v1/responses previous_response_id replays malformed thought_signature and returns TYPE_BYTES Base64 400 without tools When LiteLLM Proxy bridges OpenAI-compatible `/v1/responses` requests to Vertex AI Gemini, a later request in a conversation ch
problem · Revision 1
## Observed symptom [BUG] Gemini can't perform tools Gemini rejected OpenAI-ish tools calling. It return: `400 Bad Request: vertex-ai error: * GenerateContentRequest.tools[0].function_declarations[18].parameters.properties[sheets].items: missing field.` Desired method: Convert Ve
problem · Revision 1
## Observed symptom Invalid API hostname generated when using GCP multi-region (e.g. region: "eu") Before submitting your bug report - [x] I've tried finding an answer on the [Continue docs site](https://docs.continue.dev/) - [x] I'm not able to find an [open
problem · Revision 1
## Observed symptom [BUG] Claude for M365 (Bedrock direct): 400 "thinking.adaptive.block_binding: Extra inputs are not permitted" with Opus 4.6 Environment - Path: Bedrock direct (entra_sso=1), manifest generated with build-manifest.mjs - Add-in version: [ip], build 5f2d8b3776a49
problem · Revision 1
## Observed symptom [BUG]: phoenix-evals AnthropicAdapter structured output fails on Claude Opus 5.5 (Bedrock): forced tool_choice rejected with 400 Thanks for the evals adapters, and for #16391, which adds Claude Opus 5.5 pricing and playground support. Where do you use