Severity
Reports at this severity, newest stored first.
When a piece cannot be split further because no separators remain, it is appended directly to the output without applying strip_whitespace, causing inconsistent leading/trailing whitespace compared to merged chunks.
MarkdownHeaderTextSplitter incorrectly includes CommonMark optional closing hash sequences in header metadata values, e.g. '# Title ##' yields metadata 'Title ##' instead of 'Title'. This breaks filtering and grouping by header.
The parse_json_markdown function in langchain_core.utils.json uses a case-sensitive regex to identify fenced code blocks, causing parsing failures when models output fences with uppercase 'JSON' or 'Json'. This affects users of JsonOutputParser and other utilities relying on this function.
ChatDeepSeek only overrides sync `_generate` and `_stream` to translate `JSONDecodeError` from the OpenAI SDK into a DeepSeek-specific message. The inherited async `_agenerate` and `_astream` are not wrapped, so async callers still receive the opaque `json.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0)`.
In langchain-xai, ChatXAI.validate_environment creates two separate openai.OpenAI clients (and two openai.AsyncOpenAI) for each instance, rather than deriving client from root_client like sibling integrations. This causes duplicated resources, inconsistent close behavior, and potential connection pool waste. The fix is to build root clients first and set client/async_client as views.
In Next.js development mode, soft navigation to a route whose server component redirects after async work triggers an uncaught TypeError because React Server Components client code calls performance.measure with a negative end timestamp.
langchain-typesafe's ModelRouterMiddleware adds model_route to the agent input JSON schema even though the middleware unconditionally overwrites any caller-provided value in before_agent. This advertises an input field that is silently ignored, causing schema-driven clients to pass a route choice that has no effect.
On Android Chrome, the Authentication → Users filter form has no visible submit button, and pressing the keyboard's enter/go key does not trigger the filter. The form does not submit on mobile or desktop versions, but the issue is most severe on mobile where no button exists.
When using the browser back button to return to a page with a URL hash, Next.js erroneously scrolls to the anchor element, disregarding the preserved scroll position. This is caused by the Container component's componentDidUpdate triggering scrollToHash even on popstate navigations.
The @next/next/no-html-link-for-pages rule does not work when next.config.js specifies custom pageExtensions because the rule hardcodes default page extensions and fails to identify custom page files, leading to false negatives.
In Next.js dev with Turbopack, using next/dynamic to load a component that contains a plain import() causes the page HTML to reference a chunk that the dev server never emits, resulting in 404 errors on every page load. This affects both App Router (preload link) and Pages Router (script src). The actual dynamic import still works because the emitted loader chunk has a different hash.
The convert_to_openai_messages utility in langchain-core omits AIMessage.invalid_tool_calls from the converted OpenAI message, while the corresponding ToolMessage replies are retained, leading to orphaned tool_call_id references in exported conversations.
The UsageMetadataCallbackHandler in langchain_core only records token usage when the response's response_metadata contains a 'model_name' key. If model_name is absent, usage_metadata is silently discarded, defeating the purpose of the callback and causing silent data loss in usage tracking for custom or third-party chat models.
In langchain-anthropic's _format_messages, when an AIMessage contains both a tool_use content block and a matching tool_call in the structured tool_calls field, the tool_use block is rebuilt from tool_calls and any extra keys on the original block, notably cache_control, are discarded. This causes prompt-cache breakpoints placed on tool_use blocks to be silently lost.
The constructor of HTMLHeaderTextSplitter sorts headers_to_split_on using a bare int(tag[1:]) conversion, which fails for any tag outside h1-h6 (e.g., div -> 'iv'). The _generate_documents method already contains a fallback level of 9999 for such tags, causing an internal inconsistency that crashes at initialization.
In langchain-openai 1.6.2, Responses API usage metadata for 'priority' and 'flex' model providers double-counts cache_write_tokens: the non-cached input bucket only subtracts cached_tokens, not cache_write_tokens, causing inflated cost estimates.
The translated READMEs under i18n/ contain incorrect repository-relative image paths that resolve to non-existent i18n/apps/... locations instead of the actual repo-root apps/... assets, causing broken images in those documents.
Editing a server component before the page finishes hydrating triggers a React warning ('Can't perform a React state update on a component that hasn't mounted yet') and can lead to a hydration mismatch. The issue occurs only in Next.js dev mode with HMR active, both Turbopack and webpack.
The `next info` command incorrectly displays the latest stable version instead of the actual installed canary version, misleading users about their Next.js version.
Report of numerous test failures in langchain-core 1.6.5 on FreeBSD 15.1 (Python 3.12), likely due to platform-specific environmental issues such as event loop policy or locale settings.
ChatAnthropic.bind_tools() warns and removes forced tool_choice when thinking is set on the constructor, but the same thinking supplied via invoke() is not checked. This sends an invalid combination (thinking enabled + tool_choice={'type':'any'}) to Anthropic API, resulting in HTTP 400.
In App Router-only builds, requesting a dynamic route with a malformed percent-encoded segment (e.g. /items/%E0%A4) triggers a DecodeError during route matching. Base-server attempts to render the Pages Router /_error page with a 400 status, but that page does not exist in builds without any Pages Router routes, resulting in a bare 500 response.
The dashboard does not display the 'Last used' column for legacy API keys, contrary to documentation, preventing users from auditing key usage.
When converting a finalized AIMessage into the v3 protocol event stream using message_to_events(), the generated message-finish event does not include the AIMessage.additional_kwargs field. This causes provider-specific data to be lost during cache replay, making replayed execution inconsistent with the original live streaming path.
When using ChatOpenAI with use_responses_api=True and output_version='v1', only the first chunk (response.created) carries the provider ID resp_...; subsequent content chunks get id=None and are filled with lc_run--<run_id>. The final merged AIMessage prefers the provider ID, causing duplicate messages in LangGraph stream_mode='messages' and frontends that reconcile by ID.
In Next.js 16.x, the image optimizer's fetchExternalImage rethrows raw fetch errors, causing a 500 Internal Server Error for upstream DNS failures, connection refusals, and TLS handshake errors. The expected behavior is a client-visible upstream error status (502), consistent with other upstream failure handling in the optimizer.
Using router.replace inside a Link's onClick handler before navigation causes the previous page to be skipped on back navigation on iOS 17+ and Chrome 127+ due to a WebKit history API race; workaround is to prevent default and sequence replace/push with a delay.
The public ChatOpenAI.stream() API raises AttributeError when an OpenAI streaming chunk contains a tool-call delta whose `function` field is null. The parser assumes `rtc["function"]` is a dict and calls `.get` on None, causing a crash. Even though the repro uses a mocked response, the public API should handle malformed or unexpected provider payloads gracefully.
The langchain.mcp adapter drops the optional `_meta` field from MCP `CallToolResult` when converting to a LangChain `ToolMessage`, causing loss of out-of-band response metadata. The hotfix stores this metadata under a dedicated `mcp_meta` key in `response_metadata`, without conflating it with the static tool definition `_meta`.
In the SQL Editor AI, clicking 'Accept Changes' sets a loading state and then returns early if the diff editor is not yet mounted, leaving the button in a perpetual loading state until the page is refreshed.