mcts-agent
Discriminative Monte Carlo Tree Search using TypeSafe Jev System One Primitives and Gemini
Discriminative Monte Carlo Tree Search using System One and Harnesses
How it uses Jev
An autonomous reasoning and execution agent powered by Discriminative Monte Carlo Tree Search (MCTS) . The system couples TypeSafe Jev System One Primitives ( Noul , Choice , Score ) for lightning-fast, structured discriminative evaluations with Gemini 3.8 Flash for action generation in a closed-loop execution environment.
Highlights
- Fast Discriminative Scoring : Choice assigns action priors and Score evaluates hypothetical states. Low early scores do not remove proposed paths.
- Low Tree Reuse Threshold : After execution, grounded surviving paths are rescored. The tree is rebuilt only when no actionable path reaches the reuse threshold.
- Dynamic Prime Branching ( Choice ) : Dynamically samples expansion widths from prime numbers ($2, 3, 5, 7, 11, 13$) based on state uncertainty and goal complexity.
- Closed-Loop Execution & Grounding : Follows a strict cycle of Plan & Choose - Execute - Review - Adapt - Assess. Only the immediate winning action is executed; git diffs and termi…
- Agent Web UI : Prompt the agent, follow its live text output, preview changed image files, and browse or import run logs.
Getting started
- Python $\ge$ 3.10 - Git - TypeSafe API Key - Antigravity CLI ( agy ) or Gemini API key
Source: community catalog. Details and metrics reflect the published community snapshot and are not a performance endorsement. Documentation details extracted from GitHub · the README.