Graphify-Labs/graphify
Graphify turns any codebase, including documentation and other files, into a queryable knowledge graph using local AST parsing, without relying on vector stores or embeddings.
Awesome AI Agents › Knowledge Graph Orchestration
mingyin1/Agents_Failure_Attribution
The project "Agents_Failure_Attribution" provides the implementation of the ICML 2025 spotlight paper titled "Which Agent Causes Task Failures and When?". This work introduces the task of automated failure attribution in large language model (LLM) based multi-agent systems. The main goal is to automatically identify the specific agent and the exact step responsible for a task failure within these complex systems. This automation significantly reduces the manual effort required for debugging by inspecting failure logs and tracing errors, thereby accelerating the development cycle of multi-agent systems. Additionally, it enables intermediate feedback for agent self-improvement by pinpointing decisive errors, which can be used for self-correction or as rewards in reinforcement learning frameworks. The project includes the "Who&When" benchmark dataset, which is the first benchmark for automated failure attribution in multi-agent systems. This dataset contains 184 annotated failure tasks collected from both algorithm-generated agentic systems and hand-crafted systems. Each failure is finely annotated with the responsible agent, the critical error step, and a natural language explanation of the failure. The dataset covers a wide range of realistic multi-agent scenarios based on queries from GAIA and AssistantBench, making it a foundational resource for developing and evaluating methods to automatically identify failure causes in complex agentic systems. The repository supports various models for failure attribution, including GPT-4o, GPT-4, Llama-3.1, and Qwen2.5 models. It provides scripts for inference and evaluation, allowing users to run different failure attribution methods such as all-at-once judging, step-by-step judging, and binary search judging on both hand-crafted and algorithm-generated datasets. The project also includes experimental results and ablation studies to demonstrate the effectiveness of the proposed methods. Overall, this project is a significant contribution to the field of multi-agent systems and automated debugging, providing tools, datasets, and benchmarks to advance research and practical applications in failure attribution for LLM-based multi-agent systems.
https://github.com/mingyin1/Agents_Failure_Attribution
Also in Knowledge Graph Orchestration
microsoft/autogen
AutoGen is a Microsoft-developed framework for building multi-agent AI applications that operate autonomously or with human collaboration, featuring layered APIs, developer tools, and community support.
siyuan-note/siyuan
SiYuan is a privacy-first, self-hosted, open-source personal knowledge management system with block-level referencing and Markdown WYSIWYG editing capabilities.
Hawksight-AI/semantica
Self-hostable Python platform that builds context and knowledge graphs from enterprise data with deterministic reasoning, ontology management and W3C PROV-O provenance for auditable agent decisions.
camel-ai/oasis
OASIS is an open-source, scalable social media simulation platform that uses large language models and rule-based agents to realistically simulate the behavior of up to one million users on platforms like Twitter and Reddit for studying complex social phenomena.
liveblocks/liveblocks
Liveblocks is a platform providing customizable pre-built features to easily add multiplayer, engaging, and AI-ready collaborative experiences to applications without disrupting development roadmaps.
Paper2Poster/Paper2Poster
Paper2Poster is an open-source multi-agent system that automates the generation of editable scientific posters from research papers in PDF format, featuring a comprehensive evaluation suite to ensure visual and textual quality.
kingjulio8238/Memary
Memary is an open-source memory layer that emulates human memory to enhance autonomous agents' reasoning and learning capabilities, supporting local LLMs and vision models with multi-agent and multi-graph database integration.