Skip to content

v0.6.4 release

Choose a tag to compare

View all tags
@wxai-space wxai-space released this 27 May 05:57
· 73 commits to main since this release
v0.6.4
ca4ab4b

LightAgent v0.6.4 Release Notes

We are pleased to announce the release of LightAgent v0.6.4. This release focuses on reliability, clearer troubleshooting, safer memory guidance, and better OpenAI-compatible provider documentation.

Upgrade Guide

pip install --upgrade lightagent>=0.6.4

New Features

  • Structured Error Codes: Adds a LightAgent error taxonomy for common model-provider, tool execution, JSON parsing, rate limit, authentication, and oversized request failures.
  • Provider Configuration Guidance: Adds documentation for OpenRouter, vLLM, llama.cpp, Ollama OpenAI-compatible endpoints, and self-hosted gateways.
  • browser-use Integration Guide: Adds guidance and an updated example for browser-use 0.11+ compatibility with langchain_openai.ChatOpenAI.

Improvements

  • Runtime Tool Dispatch Reliability: Runtime tools passed through run(..., tools=[...]) are registered into the active dispatcher before the model request, so tools advertised for a run can also be executed when called.
  • Memory Security Guidance: Adds practical guidance for shared long-term memory and graph-backed memory deployments, including tenant isolation, provenance, trust levels, and retrieval checks.
  • Documentation Refresh: Updates README and FAQ links for model providers, error handling, browser-use integration, and memory security guidance.

Bug Fixes

  • Fixes runtime tool dispatch when constructor-level tools are omitted.
  • Prevents duplicate OpenAI tool schemas when tools are re-registered.
  • Improves user-facing messages for model request failures and tool execution errors.

What's Changed

  • fix: handle P0 issues
  • Delete legacy top-level tools directory from main
  • feat: handle P1 issues
  • docs: add v0.6.4 release news

Full Changelog: v0.6.3...v0.6.4

Assets 2
Loading