A sourced fix-playbook workspace for AI and agent engineers. Find the root cause, the affected version, the rollback-safe next step, and edge-case audits.
AsyncBaseTracer in langchain-core fails to forward the `name` argument to `_create_tool_run` and the `response` argument to `_errored_llm_run`, causing async tracers (including `with_alisteners`) to record incorrect run names and lose partial LLM outputs on streaming errors, unlike their sync counterparts.
SearchParams aren't updated on client-side navigation when experimental cacheComponents and Draft Mode are both enabled; the page always shows the initial category until a hard reload. This is a caching key omission in Next.js App Router.
When using cacheComponents and partialPrefetching with a custom incremental cache that starts empty, the first request to a statically generated dynamic route (e.g., /en) results in HTTP 500 due to a NEXT_STATIC_GEN_BAILOUT error when Next.js attempts to render the fallback shell dynamically.
langchain-core's finalize_tool_call_chunk maps missing tool_call id (None) to "" in the tool_call and server_tool_call branches, causing multiple id-less parallel tool calls to share the same empty id and collide downstream (e.g., deepagents FilesystemMiddleware offloads all to /large_tool_results/unknown). The proposed fix replaces the empty-string fallback with uuid4().hex to guarantee distinct ids.
Next.js App Router with Edge SSR streaming defers metadata into Suspense chunks. Bots like TelegramBot, ChatGPT-User, GPTBot, and ClaudeBot terminate the response after the initial HTML chunk, causing empty social cards and missing SEO metadata.
When POSTGRES_DB is set to a custom database name, the initialization script docker/volumes/db/jwt.sql hardcodes 'postgres' as the database name for ALTER DATABASE SET, causing JWT settings to be applied to the default database instead of the intended custom database. This leads to silent auth and RLS failures.
In Next.js 16 with Turbopack, when a `next/dynamic` component is nested inside another `next/dynamic` component (both with `ssr: true` and CSS imports), the server-rendered HTML omits the `<link rel="stylesheet">` for the deeper component's CSS. This causes a flash of unstyled content (FOUC) and significant layout shift (CLS 0.4157) in production on slow connections. The issue is a regression introduced in 16.1.0 and persists through 16.4.0-canary.34. Webpack builds do not exhibit the problem.
Next.js 16.3.5 with reactCompiler and experimental.turbopackRustReactCompiler throws a runtime ReferenceError (Cannot access 'parseItem' before initialization) because React Compiler lowers hoisted function declarations to const bindings while leaving lazy useState initializers above them, breaking valid JavaScript hoisting.
ChatOpenAI's message conversion discards unknown fields on tool calls (e.g., Gemini's thought_signature) in both directions, causing later provider-side 400 errors in multi-turn tool calling loops.
When an AIMessage has tool_calls explicitly set to an empty list while additional_kwargs still contains the original OpenAI tool call payload, the _backwards_compat_tool_calls validator treats the empty list as missing and re-parses the additional_kwargs, causing the cleared tool calls to reappear after deserialization or model reconstruction.
In langchain_core, concatenating message chunks with '+' and the merge_message_runs function silently discard the 'name' attribute and fail to fallback to the trailing chunk's 'id' when the leading chunk's id is None, causing loss of metadata during streaming and in multi-agent scenarios.
When upgrading a Next.js TypeScript app inside a pnpm v11 monorepo, the codemod creates a nested `pnpm-workspace.yaml` in the app directory instead of updating the root workspace file, causing `pnpm install` to fail with workspace dependency resolution errors.
AutoModeMiddleware evaluates the original ToolCallRequest, but an inner middleware (e.g., HumanInTheLoopMiddleware) can replace it with a protected tool call before execution, leading to policy bypass. Ordering [auto_mode, hitl] is vulnerable while [hitl, auto_mode] blocks correctly.
When ModelRouterMiddleware is composed inside ModelFallbackMiddleware, the router unconditionally replaces the fallback model with the originally selected routed model, causing the fallback model to be skipped and the same failing model to be called twice.
The LLMToolSelectorMiddleware only passes the last user message to the selector model, discarding prior conversation context. This causes vague follow-up queries (e.g., 'do it') to be misinterpreted, selecting incorrect tools.
Next.js build-time Instant Validation for Cache Components does not populate staged fallback params in the synthetic RequestStore, causing false negatives for invalid param access outside Suspense boundaries in dynamic routes with PPR.
ChatAnthropic in langchain-anthropic 1.6.1 sends top_k=10 to claude-opus-5, which Anthropic has deprecated for Opus 4.7 and later, resulting in a 400 BadRequestError. The client should validate locally and reject unsupported parameters before making a network request.
The `TypeSafeClassifier` input schema generation fails with `PydanticUserError` because the private recursive `_StateValue` type alias is not resolvable in the dynamically created `TypeSafeClassifierInput` model namespace. This breaks `get_input_jsonschema()`, `as_tool()`, schemaful graph serialization, and Runnable composition schema introspection.
Turbopack rejects SVG images that have a width attribute and a viewBox but no height, causing dev and build failures, while webpack handles the same SVG correctly by deriving height from the viewBox aspect ratio.
When a model returns invalid structured output, ToolStrategy's default handle_errors=True causes infinite retries without any attempt limit, leading to context window exhaustion or recursion limit errors. A retry budget should be introduced to fail gracefully after a maximum number of attempts.
langchain-openai 1.6.4 rejects shell, programmatic_tool_calling, and local_shell tool types in bind_tools, and the Responses API conversion layer does not handle shell_call, shell_call_output, local_shell_call, local_shell_call_output, program, program_output items, nor preserves allowed_callers and caller fields, breaking GPT-6 programmatic tool calling end-to-end.
When using Router.replace or Router.push with a href that includes query parameters followed by a hash (e.g. /page?test=1#hash) and an as path that also includes the same hash, the getInitialProps method is not re-invoked on subsequent query changes. This breaks data fetching for masked-query patterns where query is hidden from the browser URL but hash must remain visible.
When a request receives a stale cached entry from a "use cache" function, the background revalidation incorrectly replaces the saved entry in the Resume Data Cache with the pending result of the regeneration. This leads to dev-time "uncached data" errors on validation renders and in production to runtime prefetch/RSC responses missing or mixing cache versions when using cacheComponents and partialPrefetching.
With App Router, dynamic pages using revalidate=0 or no-store fetch and a global loading.tsx get stuck on the loading fallback when JavaScript is disabled, making content invisible to crawlers and harming SEO.
Clearing a Storage policy's USING or WITH CHECK expression in the Studio dashboard shows a success toast but does not persist the change because the payload omits undefined fields, leaving the old expression active and potentially causing security misconfigurations.
ChatAnthropic allows forced tool_choice and thinking={"type":"disabled"} to reach the Anthropic API for claude-opus-5-5, causing 400 invalid-request errors. bind_tools(), call-time kwargs, model_kwargs, and create_agent() with structured-output tools are affected.
TypeSafeClassifier from langchain-typesafe declares is_lc_serializable() True and dumpd() succeeds, but load() fails with both allowed_objects='all' and allowed_objects=[TypeSafeClassifier]. The root cause is that the serialized ID ('langchain', 'classifiers', 'typesafe', 'TypeSafeClassifier') is not present in langchain-core's deserialization mapping and the 'langchain_typesafe' namespace is not in the default valid namespaces.
ChatAnthropic incorrectly routes the documented browser_toolset_20260801 client toolset through JSON-schema conversion, raising ValueError. Non-null toolset_name on tool_use blocks is lost in multi-turn requests, and ordinary tool calls expose toolset_name=None.
Importing a CSV file with a geometry column where individual rows have a large number of points (over approximately 1000) causes the Supabase dashboard import process to hang. This prevents data import for such rows and requires manual splitting or exclusion.
langchain-typesafe's TypeSafeClassifier does not validate that the 'questions' mapping is non-empty before making an HTTP request, causing invalid API calls and a regression from previous Pydantic-based validation.
LangChain, LlamaIndex, and Pydantic
Pinecone, Chroma, Weaviate, Qdrant, and neighboring stores
Next.js, Supabase, and the surrounding app stack