headroomlabs-ai/headroom
Headroom is a context compression layer for AI agents, reducing token usage by 60-95% across tool outputs, logs, RAG chunks, files, and conversation history, while maintaining accuracy.
Awesome AI Agents › End-to-End Optimizers
mll-lab-nu/RAGEN
RAGEN, short for Reasoning Agent, is a research framework from the MLL Lab at Northwestern University for training LLM agents with multi-turn reinforcement learning. Its core is StarPO, State-Thinking-Actions-Reward Policy Optimization, a unified formulation that optimises whole interaction trajectories rather than single responses, with configurable control over how reasoning is elicited, how rewards are assigned across turns, and how prompts and rollouts are structured. Training can be run at trajectory level or turn-wise. The repository ships ten built-in environments covering different agent skills: Sokoban, FrozenLake, WebShop, DeepCoder, SearchQA, Lean, Bandit, Countdown, MetaMathQA and Sudoku, all behind a Gym-compatible interface so that custom environments can be added with little glue code. Beyond the training loop, the project is explicitly framed around diagnosis of why agent reinforcement learning breaks. The first release documents the Echo Trap, where a policy collapses onto repeated self-reinforcing reasoning patterns. The second release studies reasoning collapse systematically and adds two lightweight interventions: SNR-adaptive filtering, which drops rollouts whose reward variance makes their gradients noisy, and mutual-information proxy metrics that detect template collapse while training is still running, so a run can be monitored rather than judged only by final reward. Both versions are backed by arXiv papers and project pages, and the older documentation is kept for users of the first version. The intended audience is researchers and practitioners doing agentic reinforcement learning who need a reproducible baseline, a set of standard multi-turn environments, and instrumentation that explains training instability instead of only reporting scores. It is a research codebase rather than a production agent runtime.
https://github.com/mll-lab-nu/RAGEN
Also in End-to-End Optimizers
topoteretes/cognee
Cognee is an open-source project that provides scalable and modular memory solutions for AI agents using ECL pipelines, enabling efficient data ingestion, knowledge graph generation, and query capabilities to enhance AI performance and reduce hallucinations.
letta-ai/letta
Letta is an open-source framework for building stateful AI agents with advanced reasoning, transparent long-term memory, and context management, supporting multiple LLM backends and providing a graphical development environment.
langchain-ai/opengpts
OpenGPTs is an open-source platform that replicates and extends OpenAI's GPTs and Assistants API, offering customizable cognitive architectures, extensive tooling, and support for multiple language models and vector databases.
aiwaves-cn/agents
Agents 2.0 is an open-source framework that enables the training and self-evolution of autonomous language agents through symbolic learning inspired by neural network methodologies.
getzep/zep
Zep is an AI memory foundation that uses a temporal Knowledge Graph to enable continuous learning and personalized experiences for AI agents by efficiently managing and retrieving contextual chat and data artifacts.
algorithmicsuperintelligence/optillm
OptiLLM is an optimizing inference proxy that significantly improves the accuracy and performance of large language models on reasoning tasks without requiring any model training, by applying over 20 advanced optimization techniques during inference.
langfengQ/verl-agent
verl-agent is an extension of veRL designed for scalable training of large language and vision-language model agents via reinforcement learning, featuring customizable input structures, diverse RL algorithms including GiGPO, and support for multi-modal environments.