init_chat_model

This release introduces significant improvements to the MCP (Multi-modal Conversational Protocol) handling, including new elicitation features and tool conversion. It also brings general bug fixes and refactors across various components.

This release introduces new features like exposing `trace_policy` on `AgentMiddleware` and adding `reasoning_effort` as a chat model parameter. It also includes several bug fixes, such as preserving history on summarization failures and handling malformed structured-output responses.

This release introduces new features for prompt caching in OpenAI and enhanced initialization capabilities for chat models with the `meta` extra. It also includes dependency updates.

This release introduces significant updates to streaming protocols (v2 and v3), enhances agent capabilities with new middleware features, and includes numerous dependency bumps and bug fixes across various components.

This release includes numerous performance improvements, dependency bumps, and feature additions like content-block-centric streaming (v2) and integration of stream_events(version='v3'). Several bug fixes address token counting, model recognition, and middleware configuration.

This release introduces content-block-centric streaming (v2) and support for the v3 stream protocol, alongside numerous bug fixes related to tracing, validation, and dependency updates. Path saving/loading functions for prompts have been deprecated.

This release introduces significant feature updates, including content-block-centric streaming (v2) and support for the v3 event protocol. Numerous bug fixes address issues related to tracing, serialization, SSRF hardening, and model wrapper compatibility.

This release focuses heavily on stability, security hardening (anti-SSRF), and improved token counting, especially for multimodal inputs and tool schemas. Several prompt loading/saving methods have been deprecated due to path validation improvements.

This release focuses heavily on stability, security hardening (anti-SSRF, pygments update), and feature parity, including adding Bedrock and OpenRouter support, and enhancing token counting for multimodal inputs and tool schemas. Several internal cleanups and deprecations related to prompt saving paths were also introduced.

This release includes minor dependency bumps, CI improvements, and adds LangSmith integration metadata to agent and chat model initialization functions. A bug fix was also applied to the OpenAI Responses API input typing.

Version 1.2.20 of langchain-core includes security hardening, improved LangSmith integration metadata, and a fix for tracing invocation parameters.

This release introduces support for OpenRouter and automatic server-side compaction in OpenAI, alongside several dependency updates and bug fixes related to model initialization and detector output.

This release bumps the minimum required version for `langchain-core` to 0.3.73 and patches a critical ReDoS vulnerability in MRKL/ReAct regexes. It also includes numerous internal cleanups, dependency bumps (like Ruff), and updates to model handling, particularly for Anthropic and Google models.

This release focuses on bug fixes for message summarization and tool schemas, adds Google GenAI support to embeddings, and introduces several linting and type-checking improvements.

This release updates langchain-classic to 1.0.1, featuring enhanced model initialization for Anthropic and Google GenAI, a transition to uuid7 for run IDs, and the removal of the Tigris provider.

This release introduces a strict mode for structured outputs in ProviderStrategy and adds an extras field to BaseTool, alongside fixes for multithreading race conditions and HuggingFace model initialization.

Release 1.2.0 focuses on fixing and improving the init_chat_model integration for HuggingFace backends and updating core dependencies.

This release introduces support for Upstage models in the chat model initializer and adds an agent name field to AIMessage, alongside typing improvements in core message utilities.

LangChain v1.0.0 release introduces Python 3.14 support, a suite of new tool and model middleware, and significant refactoring of the agents and tools namespaces for production stability.

Initial release of LangChain v1.0.0 and Core v1.0.0rc, featuring a major refactor of middleware hooks (wrap_model_call/wrap_tool_call), async support for agents, and the removal of global variables.

LangChain v1.0.0 release featuring a major refactor of the agents API, introduction of a robust middleware system (PII, retries, limits), and namespace consolidation. It transitions the project to a stable versioning scheme with enhanced async support and improved Human-In-The-Loop patterns.

Initial release of LangChain v1.0.0 series, featuring a major refactor of middleware hooks (wrap_model_call/wrap_tool_call), improved HITL capabilities, and namespace consolidation.

LangChain v1.0.0 introduces a revamped agent API, a new middleware system for model and tool calls, and significant namespace consolidation. It transitions the codebase to Python 3.10+ and standardizes asynchronous support across core components.

Initial release of LangChain v1.0.0 (alpha stages), featuring a major refactor of the agent API, the introduction of a middleware system for LLM calls, and the removal of legacy globals and Python 3.9 support.

LangChain v1.0.0 introduces a major architectural shift focusing on agent middleware, improved HITL patterns, and a streamlined core namespace. It officially drops support for Python 3.9 and renames several key agent APIs for better consistency.

LangChain v1.0.0 introduces a major architectural shift focusing on a new middleware system, improved agent hooks, and the removal of legacy global variables and Python 3.9 support. It standardizes the agent API and enhances production features like HITL and PII filtering.

Official release of the langchain-deepseek partner package (v1.0.0), featuring integration with init_chat_model, reasoning content fixes for R1 models, and a migration to uv for dependency management.

This release introduces support for xAI and IBM WatsonX AI models in init_chat_model, improves JSON parsing for null action inputs, and enhances async testing with Blockbuster.

This release marks the initial launch of the langchain-deepseek package, renaming it from its official placeholder and integrating ChatDeepSeek into the LangChain ecosystem.

LangChain 0.3.16 introduces DeepSeek support for chat models, Ollama support for embeddings, and updates the ruff linting version to 0.9.