Agent DailyAgent Daily
releaseintermediate

[Release] langchain-ai/langchain langchain-core==1.3.0a1: langchain-core==1.3.0a1

By github-actions[bot]github
View original on github

langchain-core version 1.3.0a1 is an alpha release featuring performance improvements, security hardening, and bug fixes. Key updates include reduced streaming metadata overhead, enhanced anti-SSRF protections, improved token counting with multimodal support, and numerous stability fixes across serialization, tool handling, and model integrations. The release represents incremental improvements building on the 1.2.x series with focus on reliability and developer experience.

Key Points

  • Performance optimization: Reduced streaming metadata overhead to improve throughput and reduce memory consumption
  • Security hardening: Enhanced anti-SSRF protections and added security warnings for deserialization best practices
  • Multimodal token counting: Added support for counting tokens from multimodal messages and tool schemas in `count_tokens_approximately`
  • Tool handling improvements: Fixed merge_lists for parallel tool calls, improved error messages for non-JSON-serializable schemas, and preserved default_factory in tool call schemas
  • Serialization robustness: Added sanitization to templates, improved path validation in prompt save/load, and fixed nested mustache variable extraction
  • Model integration updates: Added ChatBaseten to serializable mapping, improved Bedrock model support, and added ChatAnthropicBedrock wrapper
  • Dependency updates: Upgraded critical dependencies (pygments>=2.20.0 for CVE-2026-4539, urllib3 to 2.6.3, requests to 2.33.0)
  • Error handling: Added ContextOverflowError for better token limit management, improved error messages for tool schemas, and better exception handling in tracers
  • Documentation improvements: Enhanced docstrings for RunnableConfig, RunnableSerializable, and tool decorators with usage examples
  • LangSmith integration: Added metadata support to create_agent and init_chat_model for better tracing and observability

Found this useful? Add it to a playbook for a step-by-step implementation guide.

Workflow Diagram

Start Process
Step A
Step B
Step C
Complete
Quality

Concepts