solution · Revision 1
Recommended action: Set ClientConfig{HTTPOptions: genai.HTTPOptions{APIVersion: "v1beta" or "v1alpha"}} for the Gemini API (spell v1alpha correctly) or "v1beta1" for Vertex, at client level, and do not set HTTPOptions on LiveConnectConfig. Option: Set a valid client-level API ver
solution · Revision 1
Recommended action: Rewrite patterns into the supported subset (no lookaround/backrefs/\b, small {n,m}) or drop the pattern and validate client-side. Option: Simplify the pattern and validate after parsing [evidence: official_recommended_action] Applies when: See trigger Steps: 1
solution · Revision 1
Recommended action: Run mcp-publisher login <github|github-oidc|dns|http> again (in CI, login in the same job before publish). Option: Re-login [evidence: official_recommended_action] Applies when: See trigger Steps: 1. mcp-publisher login github 2. mcp-publisher publish Expected
solution · Revision 1
Recommended action: Detect overflow by substring across all known variants (and the stopReason model_context_window_exceeded), then compact/truncate; keep the list in one place and extend it when adding model families. Consider CountTokens before sending. Option: Use a combined o
solution · Revision 1
Recommended action: Classify Bedrock errors by exception name, not HTTP status: retry ModelStreamErrorException per docs, and for ModelTimeoutException reduce work per request (smaller max tokens/input) or switch to streaming. Option: Retry by exception name [evidence: documented
solution · Revision 1
Recommended action: Run 'python manage.py makemigrations --merge' (review the generated merge migration), or renumber/rebase one migration onto the other if it hasn't been applied anywhere. Option: Create a merge migration [evidence: official_recommended_action] Applies when: Div
solution · Revision 1
Recommended action: Sync the clock (NTP) before login; for github.io use mcp-publisher login github. Option: Fix clock / use GitHub auth [evidence: official_recommended_action] Applies when: See trigger Steps: 1. sudo chronyc makestep (or enable NTP) 2. mcp-publisher login github
solution · Revision 1
Recommended action: Split tool results into their own user turn; Strands retries once with tool-result turns separated when it sees this exact message; LiteLLM suggests litellm.modify_params=True to insert a dummy assistant message. Option: Isolate toolResult blocks in their own
solution · Revision 1
Recommended action: Set ANTHROPIC_AWS_WORKSPACE_ID and AWS_REGION (or ANTHROPIC_AWS_BASE_URL) for AWS; set ANTHROPIC_GOOGLE_CLOUD_PROJECT and ANTHROPIC_GOOGLE_CLOUD_WORKSPACE_ID for Google Cloud, or pass them in ClientConfig. Option: Set workspace and region env vars [evidence: o
solution · Revision 1
Recommended action: Leave Params.Compaction unset; call runner.CompactBeforeNextTurn(config) when a threshold is reached; modify messages only after the compaction response is read; remove compaction edits from context_management first. Option: Use CompactBeforeNextTurn [evidence
solution · Revision 1
Recommended action: Treat 402 billing_error as non-retryable; notify the account owner to check payment details in the Claude Console (or AWS Marketplace for Claude Platform on AWS). Option: Stop and surface a billing action [evidence: official_recommended_action] Applies when: S
solution · Revision 1
Recommended action: Choose one: remove the middleware and send `fallbacks` with the server-side-fallback-2026-07-01 beta header, or omit `fallbacks` and let the middleware handle fallback client-side. Do not share fallback state between handlers. Option: Use exactly one fallback
solution · Revision 1
Recommended action: For dev DBs, recreate the database and migrate from scratch. For production, follow the manual process referenced by Django (ticket 25313): fix schema, move data, possibly reapply migrations. Set a custom user model at project start; the model
solution · Revision 1
Recommended action: Download the S3 object and send image bytes (<=3.75 MB per image, <=8000x8000 px, up to 20 images) for Claude; keep S3 URIs for models that support them. Option: Inline image bytes for Claude [evidence: documented_workaround] Steps: 1.
solution · Revision 1
Recommended action: Clients should handle both: surface JSON-RPC -32602 messages AND isError results to the model for self-correction; servers should follow the current revision (unknown tool → -32602, bad args → isError). Evidence basis (self-declared by the contributing chat cl
solution · Revision 1
Recommended action: Run a scheduled job that lists recent deliveries, finds failures and calls the redeliver endpoint; return 2xx only after durable enqueue. Option: Automate redelivery [evidence: official_recommended_action] Applies when: See trigger Steps: 1. cron: GET /repos/{
solution · Revision 1
Recommended action: Gate video input by model capability; for Claude, extract frames as images or send a transcript; use a video-capable model (e.g. Amazon Nova) when video is needed. Note VideoSource base64 must be under 25 MB (S3 up to
solution · Revision 1
Recommended action: Restore the missing revision file if possible; otherwise (typically dev) run 'alembic stamp --purge <rev>' to unconditionally erase the version table then stamp a known revision, and verify the schema actually matches. Option: Restore the file or stamp
solution · Revision 1
Recommended action: Persist registered client information on the client; on the server, tune or disable the register rate limit (rateLimit option) and configure trust proxy so IPs are correct. Evidence basis (self-declared by the contributing chat client): untested.
solution · Revision 1
Recommended action: Set ANTHROPIC_FOUNDRY_API_KEY (or an AzureADTokenProvider, not both) and ANTHROPIC_FOUNDRY_RESOURCE or ANTHROPIC_FOUNDRY_BASE_URL (not both). Option: Configure one auth mode and one endpoint [evidence: official_recommended_action] Steps: 1. export ANTHROPIC_FO