# KhoomeiK/LlamaGym

Fine-tune LLM agents with online reinforcement learning

Repository: https://github.com/KhoomeiK/LlamaGym
Canonical: https://ross.abutalabs.com/products/llamagym
Language: Python
License: MIT
License Family: permissive
Last push: 2024-03-19T17:34:28+00:00

## Health v2 (maintenance only)
Score: 25/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 65
- inputs: {"age_days": 915, "days_push": 897, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1254, forks 65 (observed 2026-08-28T04:04:08.818012+00:00)

## What it is
LlamaGym is a Python library that simplifies fine-tuning LLM-based agents with online reinforcement learning in Gym-style environments. It provides a single Agent abstract class that handles conversation context, episode batching, reward assignment, and PPO setup.

## Use cases
- fine-tune an LLM agent with reinforcement learning in a Gym environment
- train an LLM to play blackjack via RL
- experiment with agent prompts and hyperparameters across RL environments
- simplify PPO setup for LLM agents
- build agents that learn online from reward signals

## When to choose
- you want to fine-tune an LLM agent with RL in a Gymnasium environment without writing boilerplate
- you want a minimal abstract class to iterate on agent prompting and hyperparameters
- you already have a Gym environment and an LLM you want to train

## When to avoid
- you need a production-ready, actively maintained RL training framework
- you want offline fine-tuning without reinforcement learning
- you need multi-agent or non-Gym training setups

## Facets
- artifact type: library
- maturity: experimental
- function: reinforcement-learning, llm-training, agent-framework, machine-learning
- domain: reinforcement-learning, large-language-models, machine-learning
- platform: python
- tags: gym-environments, fine-tuning, ppo, abstract-class, openai-gym, ai-agents

## Member repositories
- KhoomeiK/LlamaGym (main) score 25

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:08.818012+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T05:07:23.092990+00:00, confidence not recorded.
  - readme: https://github.com/KhoomeiK/LlamaGym (fetched 2026-08-28T04:04:08.818012+00:00, sha 36ebe41314d6)
- Data as of 2026-08-30T08:39:29.467469+00:00.