Repository Issue Activity (beta)

#9099 Custom Tools with a zod schema do not send field descriptions to OpenAI API

#11444 @langchain/google: ToolMessage followed by HumanMessage merges into one Gemini `user` content, and Vertex rejects the request

#8454 @langchain/google-webauth: Fails to Parse Parallel Tool Calls from Gemini 2.5 Flash in Streaming Mode

#8585 Google Gemini not supporting native structured output - instead always using tool calling method

#9167 [langchain]: ChatGoogleGenerativeAI does not support transport option in js? eventhoug python based langchain has that option

#7640 Bug [@langchain/google-vertexai-web] 403 Forbidden

#7741 Bad error message from Google GenAI when no HumanMessage present in request

#7758 ChatGoogleGenerativeAI withStructuredOutputs does not bind the correct fields

#8548 google-genai: Context is not getting cached while using Gemini 2.5

#8488 Intermittent undefined is not an object (evaluating 'chatGeneration.message') with gemini flash 2.5

#7434 google-common: Support Gemini 2.0 Thinking model

#11438 @langchain/openai: a caller-supplied User-Agent is emitted as a second header, so the transport sends a comma-joined value

#11125 humanInTheLoopMiddleware behaves incorrectly on rejection with human message provided

#11435 `ProviderStrategy` parses Gemini thought summaries as structured output

#11432 `@langchain/google` uses an invalid endpoint for Vertex multi-regions

#11221 Provider HTTP spans emit as disconnected root traces

#11416 @langchain/openai: Chat Completions converter emits empty content array for v1 tool-call-only AIMessage

#11197 Surface Vertex trafficType and modelVersion on response_metadata

#9366 Adding image to request adds up >10secs for Langchain `ChatVertexAI`vs. google-genai's `GoogleGenAI`

#11424 `llmOutput.tokenUsage` is wrong on both streaming paths, so callback-based token tracking under-reports Gemini usage by orders of magnitude

#11422 `@langchain/google-genai`: `thoughtsTokenCount` is dropped, so reasoning tokens are missing and `input_tokens + output_tokens` no longer equals `total_tokens`

#11417 @langchain/openrouter: ChatOpenRouter._generate throws bare TypeError when provider returns 200 without `choices`

#11409 ChatOpenRouter streaming responses do not preserve routed model metadata

#11404 [perf] LangChainTracer includes already-posted inputs in the closing run update

#9936 ChatGoogleGenerativeAI does not support `ContentBlock.Multimodal.*`