langchainhub

Fetches pre-built prompt templates from the LangChain Hub repository using a client-side API that resolves template identifiers (e.g., 'rlm/rag-prompt') to structured prompt objects. The client handles authentication via API keys, manages HTTP requests to the Hub's REST endpoints, and deserializes YAML/JSON template definitions into LangChain-compatible PromptTemplate instances with variable interpolation support.

Loads serialized LangChain Runnable chains (agents, sequential pipelines, branching logic) from the Hub and instantiates them as executable Python/JavaScript objects. The client deserializes the chain's DAG structure, resolves LLM provider references (OpenAI, Anthropic, etc.), injects API keys from environment variables, and returns a ready-to-invoke Runnable that can be called with input data without additional configuration.

Manages API authentication to the LangChain Hub using API keys stored in environment variables or passed explicitly to the client constructor. The client maintains a session with automatic header injection (Authorization: Bearer token), handles token refresh if needed, and provides error handling for 401/403 responses with clear messaging about missing or invalid credentials.

Queries the LangChain Hub API to list available prompts, chains, and other resources with optional filtering by owner, type, or tags. The client returns paginated results as structured objects with metadata (name, description, owner, creation date), enabling programmatic discovery of Hub resources without manual browsing of the web interface.

Validates locally-defined prompts and chains against the LangChain Hub's serialization schema before pushing them to the Hub. The client checks for required fields, correct data types, valid variable references, and compatibility with supported LLM providers, returning detailed validation errors that guide users to fix issues before publication.

Publishes locally-created prompts, chains, and other artifacts to the LangChain Hub with metadata (name, description, tags, visibility). The client serializes LangChain objects to the Hub's format, handles multipart uploads if needed, manages versioning (creating new versions or updating existing ones), and returns a shareable Hub URL for the published resource.

Automatically resolves environment variable references in retrieved prompts and chains (e.g., ${OPENAI_API_KEY}, ${MODEL_NAME}) by substituting values from the runtime environment. The client supports both simple variable substitution and nested references, with fallback defaults and validation that required variables are present before chain execution.

Automatically enables tracing and observability for chains loaded from the Hub by injecting LangSmith callbacks into the Runnable execution context. The client detects when LangSmith credentials are available (LANGSMITH_API_KEY, LANGSMITH_PROJECT) and transparently logs all chain steps, LLM calls, and tool invocations to LangSmith for debugging and monitoring without requiring explicit callback configuration.