Skip to content

Release 0.6

Choose a tag to compare

View all tags
@bdambrosio bdambrosio released this 10 Jan 22:37
· 1591 commits to main since this release
v0.6
bb466c8
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Stable release, subject to further testing.
Major overhauls of many elements, including:

  1. cleaning / tightening basic incremental planner reasoning loop
  2. implementation of composite python tools with subplanner access
  3. addition of 'world' tool support. World-specific tools are loaded in addition to the basic tools/ tools when a world is specified in .yaml. the world 'init' tool is run automatically at the end of startup.
  4. goals do not need to be prefixed with goal: in the UI input area - all input is automatically a 'goal' for the agent.
  5. benchmarks - mmlu, gaia, hle, hotpotqa, are in the tests dir. Consider these as experimental, they often break with core engine zenoh api, but are mostly working in this release.
  6. initial support for several 'worlds', including Minecraft, OSWorld, ScienceWorld (AllenAI). Minecraft is the most fleshed out, as I've settled on it as a testbed to drive agent development forward, at least for the moment. The theme of the approach is to NOT ignore world idiosyncracies (as mineflayer did), but rather treat them as the core research issue for 'real' agents
  7. Pbly lots more. I'll try to add more wiki pages. Post issues or use discussions!
Assets 2
Loading