Recommended action: Fix/upgrade the server or proxy to support secure renegotiation. If you must connect, enable legacy connect narrowly for that client (e.g. Python 3.12+ ctx.options |= ssl.OP_LEGACY_SERVER_CONNECT, or an OpenSSL config with Options = UnsafeLegacyServerConnect s
Recommended action: Surface the verification URL and code to the human through a channel they actually watch, stop polling at expires_in, and only restart on explicit request; for unattended agents use a pre-provisioned token (GH_TOKEN) or GitHub App installation token
Recommended action: pip install -U "langchain-aws[nova-sonic]"; use amazon.nova-2-sonic-v1:0 for Nova 2 voices and endpointing_sensitivity. Option: Install the nova-sonic extra and match the model [evidence: official_recommended_action] Steps: 1. pip install -U "langchain-aws[nov
Recommended action: GET existing capability hosts at the scope and reuse the same name; wait for in-flight operations (operationResults) and retry with exponential backoff. Option: Reuse the existing name / wait and retry [evidence: official_recommended_action] Steps: 1. GET .../
Recommended action: Fix host time (NTP/chrony; in WSL/containers resync after sleep). In Python there is no automatic correction — resync the clock rather than retrying. Option: Resync the host clock [evidence: official_recommended_action] Applies when: See trigger Steps: 1. Linu
Recommended action: Poll no faster than interval; on slow_down set interval = response.interval (or +5 s) and continue; stop at expires_in (default 900 s). Option: Honor the dynamic polling interval [evidence: official_recommended_action] Applies when: See trigger Steps: 1. inter
Recommended action: Check SELECT pg_is_in_recovery() and SHOW default_transaction_read_only; point writes to the primary; replace session-level read-only settings with BEGIN TRANSACTION READ ONLY; run migrations/admin tasks over direct connections; RESET/DISCARD ALL before return
Recommended action: Add a safety margin for tool schema tokens (or count with a provider that includes tools) and treat CountTokens as an estimate. Option: Budget extra tokens for tools [evidence: documented_workaround] Steps: 1. Estimate tool schema tokens separately and
Recommended action: Pass both handlers; return true only for a bounded number of retries (e.g. retryCount < 2) and log the retryAfter. Option: Provide both handlers [evidence: official_recommended_action] Applies when: See trigger Steps: 1. throttle: { onRateLimit: (retryAfter, o
Recommended action: Remove the unsupported parameter (set it per-transaction or via ALTER ROLE/DATABASE ... SET) or use the unpooled connection string. SET search_path over a transaction-mode pool also fails later ('Relation not found'). Option: Move settings off the startup pack
Recommended action: Always set a sanitized, unique 'name' on file blocks (strip extension and disallowed characters). Option: Name every document block [evidence: official_recommended_action] Steps: 1. Add 'name': sanitized_stem (unique per request). Expected: Documents accepted.
Recommended action: Use turn_detection={'type':'none'} with keepalive_silence=True and commit manually, or append audio continuously; set base_url to an API root like https://api.together.ai/v1. Option: Manual commit + keepalive [evidence: official_recommended_action] Steps: 1. t
Recommended action: Pass leeway (seconds or timedelta) to jwt.decode and fix NTP on both sides. Option: Use leeway [evidence: official_recommended_action] Applies when: See trigger Steps: 1. jwt.decode(token, key, algorithms=[...], leeway=30) Expected: Error no longer occurs Evid
Recommended action: Always authenticate (GH_TOKEN/GITHUB_TOKEN or a GitHub App installation token) and read x-ratelimit-remaining/x-ratelimit-reset; use a dedicated GitHub App installation for automation so its budget is separate from users. Option: Authenticate every call and is
Recommended action: Increase connect_timeout (Prisma: connect_timeout parameter per Neon Prisma guide), add retry on first connect, validate/recycle idle pool connections, or disable scale to zero on paid plans. Option: Tolerate activation latency [evidence: official_recommended_
Recommended action: Switch to ChatBedrockConverse (preferred) or ChatBedrock. Option: Use the chat model classes [evidence: official_recommended_action] Steps: 1. from langchain_aws import ChatBedrockConverse Expected: Works with Claude 3+. Evidence basis (self-declared by the co
Recommended action: Disable thinking for structured-output calls, or use native structured outputs / tool_choice auto with a validation-and-retry loop; watch for the downgrade warning. Option: Separate thinking and forced-tool calls [evidence: official_recommended_action] Steps: