Python orchestration framework for building agentic AI and LLM applications, with support for RAG pipelines, agents, and multi-step workflows.
1.1K+1 this week154PythonApache-2.0Well maintainedUpdated yesterday
om-ai-lab/OmAgent
A Python library for building multimodal language agents that can reason over text, images, video and audio inputs.
- Stars
- 2,671
- Forks
- 289
- Open issues
- 23
- Last commit
- 22 days ago
What is it?
OmAgent is a Python library for building multimodal language agents. It wraps worker orchestration, task queues and node execution behind a simple interface, so developers can define an agent and compose reusable components into more complex agents.
The library provides a graph-based workflow engine and several memory types for contextual reasoning. It includes native support for vision-language models, video processing and mobile device connection, and it implements agent algorithms such as ReAct, CoT and SC-CoT. Models can be run locally through Ollama or LocalAI, and the architecture can be deployed distributed or in a Lite mode without middleware.
The project ships example applications, including a video question-answering agent with a Gradio interface and a mobile personal assistant, along with a comparison of agent operators on reasoning datasets.
Key features
- 01Graph-based workflow orchestration engine with multiple memory types
- 02Native support for vision-language models, video processing and mobile device connection
- 03Built-in agent algorithms including ReAct, CoT and SC-CoT
- 04Local model deployment through Ollama or LocalAI
- 05Distributed deployment with scaling, plus a Lite mode without middleware
Use cases
- Building video question-answering agents that reason over uploaded footage
- Prototyping image or video understanding pipelines with VLMs
- Creating a mobile personal assistant that handles multimodal input
- Comparing reasoning strategies such as CoT and ReAct on the same LLM
In collections
Topics
FAQ
- What is OmAgent?
- A Python library for building multimodal language agents that can reason over text, images, video and audio inputs. OmAgent is a Python library for building multimodal language agents. It wraps worker orchestration, task queues and node execution behind a simple interface, so developers can define an agent and compose reusable components into more complex agents.
- Is OmAgent free to use? What license does it have?
- OmAgent is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
- Is OmAgent still maintained?
- Actively maintained. The last commit to OmAgent was on Oct 16, 2026.
- What can I use OmAgent for?
- Building video question-answering agents that reason over uploaded footage Prototyping image or video understanding pipelines with VLMs Creating a mobile personal assistant that handles multimodal input Comparing reasoning strategies such as CoT and ReAct on the same LLM
- Where can I get OmAgent?
- The source code and installation instructions are on GitHub: https://github.com/om-ai-lab/OmAgent
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Open-source RAG backend platform with a spreadsheet-like UI, combining SQLite, LanceDB, and managed LLM orchestration behind a REST API.
1.1K48PythonApache-2.0Well maintainedUpdated 17 days ago
A Go framework for building LLM applications and AI agents, with composable components, graph orchestration, and human-in-the-loop interrupts.
13.6K+71 this week1.2KGoApache-2.0Updated 3 days ago
A C#/.NET library for running LLaMA-family and other GGUF language models locally on CPU or GPU, built on llama.cpp.
3.8K+5 this week516C#MITUpdated yesterday
Swap arxiv.org for talk2arxiv.org in any paper link to read the paper and chat with an AI that has read its full text.
53131TypeScriptApache-2.0Updated last month
Open-source multimodal AI agent stack from ByteDance, with a CLI/Web UI (Agent TARS) and a desktop GUI agent app (UI-TARS Desktop).
39.6K+83 this week4KTypeScriptApache-2.0Updated 30 days ago