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
FoundationAgents/AFlow
AFlow is a research framework, published at arXiv 2410.10762, that treats the design of an agentic workflow as a search problem instead of a manual engineering task. A workflow is a sequence of LLM-invoking nodes connected by edges and represented as executable code, which lets it express graph, network and control-flow structures rather than a fixed template. The framework is built from four components. A node is the basic unit of LLM invocation and exposes control over model, temperature, output format and prompt. An operator is a predefined combination of nodes that narrows the search space, covering common patterns such as generate, format, review, revise, ensemble, test and programmer, and users can define their own operators by following the supplied implementations. The optimiser applies a variant of Monte Carlo tree search in which an LLM proposes modifications, iteratively selecting, expanding, evaluating and updating candidate workflows according to measured performance. The evaluator scores a workflow on a target task and feeds the result back into the search, so improvement is driven by benchmark outcomes rather than by intuition. Experiments cover six datasets spanning code generation, mathematics and multi-hop question answering: HumanEval, MBPP, GSM8K, MATH, HotpotQA and DROP, with downloadable data and evaluation code included. Custom tasks are supported by subclassing a base benchmark and implementing problem evaluation, score calculation and result columns, then registering the benchmark name with the evaluator and optimiser. Running an optimisation means configuring an LLM in a YAML file, choosing a dataset, the number of resampled workflows per round, the maximum number of rounds, validation rounds and optional early stopping via convergence checking. The code originates from the MetaGPT project and is aimed at researchers and practitioners studying automated workflow discovery.
https://github.com/FoundationAgents/AFlow
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.
mll-lab-nu/RAGEN
A reinforcement learning framework for training multi-turn reasoning LLM agents on the StarPO algorithm, shipping ten Gym-compatible environments plus diagnostics for echo-trap and reasoning-collapse failure modes.