๐ The First Self-Improving agents with RL / Prompting Optimization
Python agentaigptmulti-agentpythonreinforcement-learningself-improvementsociety-of-mindswarm-intelligence
Upload REAMDE.md & .gitignore
Mingchen Zhuge35edfb94d803
diff --git a/.gitignore b/.gitignoreindex 68bc17f..0bcd49f 100644--- a/.gitignore+++ b/.gitignore@@ -158,3 +158,40 @@ cython_debug/# and can be added to the global gitignore or merged into this file. For a more nuclear# option (not recommended) you can uncomment the following to ignore the entire idea folder.#.idea/++# Others+*.DS_Store++# log+result/+logs/+lib/++# VSCODE+vscode/+.vscode/++#.pytest_cache+.pytest_cache/++# others+.coverage+.coveragec++# vis+result/example.html+example.html+datasets/gaia/test_files/+datasets/gaia/val_files+datasets/gaia/*.json+datasets/gaia/utils/*.jsonl++/runs/++.$*.drawio.bkp+/diagrams/*.pdf+/diagrams/*.png+/figures/*.pdf+/figures/*.png++.pypirc\ No newline at end of filediff --git a/README.md b/README.mdindex 4b84550..e51486f 100644--- a/README.md+++ b/README.md@@ -1,2 +1,139 @@-# GPTSwarm-๐ GPTSwarm: LLM agents as Graphs+[](https://gptswarm.org)+[](https://arxiv.org/pdf/2305.17066.pdf)+[](https://github.com/mczhuge/GPTSwarm/blob/main/LICENSE)+[](https://github.com/metauto-ai/gptswarm/issues)+[](https://twitter.com/AI_KAUST)+[](https://metauto.ai/images/wechat.jpeg)++<p align="left">+<a href=""><img src="swarm/utils/assets/logo.png" alt="GPTSwarm" width="430px"></a>+</p>++๐ **GPTSwarm is a graph-based framework for LLM-based agents, providing two high-level features:**++* It lets you build LLM-based agents from graphs.+* It enables the customized and automatic self-organization of agent swarms with self-improvement capabilities.+++## Edge optimization example++Here is the edge optimization process that updates edge probabilities toward improvement of the benchmark score. Notice that within an agent, the edges are fixed, whereas the inter-agent connections are getting optimized towards either edge pruning (value 0, blue) or creation (value 1, red).++<img src="assets/edge_opt.gif" alt="Edge optimization" width="300">++## About GPTSwarm++<img src="assets/gpt_swarm.png" alt="Framework" width="800">++At a granular level, GPTSwarm is a library that includes the following components:+++| Module | Description |+| ---- | --- |+| [**swarm.environment**](swarm/environment) | Domain-specific operations, agents, tools, and tasks |+| [**swarm.graph**](swarm/graph) | Graph-related functions for creating and executing agent graphs and swarm composite graphs |+| [**swarm.llm**](swarm/llm) | Interface for selecting LLM backends and calculating their operational costs |+| [**swarm.memory**](swarm/memory) | Index-based memory |+| [**swarm.optimizer**](swarm/optimizer) | Optimization algorithms designed to enhance agent performance and overall swarm efficiency |+++## Quickstart++**Clone the repo**++```bash+git clone --recurse-submodules https://github.com/mczhuge/GPTSwarm.git+cd GPTSwarm/+```++**Install packages**+```+conda create -n swarm python=3.10+conda activate swarm+pip install -r requirements_py310_<linux|macos>.txt+```++**You should add API keys in `.env.template` and change its name to `.env`**++```python+OPENAI_API_KEY="" # for OpenAI LLM backend+SEARCHAPI_API_KEY="" # for Web Search+```++**Getting started with GPTSwarm is easy. Quickly run a predefined swarm**++```python+from swarm.graph.swarm import Swarm++swarm = Swarm(["IO", "IO", "IO"], "gaia")+task = "What is the capital of Jordan?"+inputs = {"task": task}+answer = await swarm.arun(inputs)+```++**or make use of tools, such as the file analyzer**++```python+from swarm.graph.swarm import Swarm+swarm = Swarm(["IO", "TOT"], "gaia")+task = "Tell me more about this image and summarize it in 3 sentences."+files = ["./datasets/demos/js.png"]+inputs = {"task": task, "files": files}+danswer = swarm.run(inputs)+```++Check out the minimal Swarm example in Colab here: [](https://colab.research.google.com/github/mczhuge/GPTSwarm/blob/main/notebooks/demo_swarm.ipynb).++See how to create a custom Agent and run a Swarm with it here: [](https://colab.research.google.com/github/mczhuge/GPTSwarm/blob/main/notebooks/demo_custom_agent.ipynb).++Watch the following for an example of how to run the notebooks.++https://github.com/mczhuge/GPTSwarm/assets/100462607/a05f78b7-5f9d-4762-8337-973702ce4493+++**๐ See [our experiments](https://github.com/mczhuge/GPTSwarm/tree/main/experiments) for more advanced use of our framework.**++## Class diagram++<img src="assets/class_diagram.png" alt="Edge optimization" width="700">++## Example of the Swarm++<img src="assets/swarm_v3.png" alt="Edge optimization" width="500">++## More Visualizations++<img src="assets/swarm_vis.png" alt="Edge optimization" width="800">++## Running with a local LLM++We support local LM inference via [LM Studio](https://lmstudio.ai). Download their desktop app for Mac or Windows, choose a model from the Huggingface repository and start the server. Use `model_name='lmstudio'` in GPTSwarm code to run with the local LLM.++<img src="assets/lm_studio.png" alt="Edge optimization" width="800">++## Initial Contributors++* [Mingchen Zhuge](http://metauto.ai) (PhD Student, KAUST; Project Initiator)+* [Wenyi Wang](https://scholar.google.com/citations?user=79ODhuQAAAAJ&hl=en&oi=ao) (PhD Student, KAUST; Initial Participant)+* [Dmitrii Khizbullin](http://www.khizbullin.tech) (Research Engineer Lead, KAUST; Project Engineer Lead)+* [Louis Kirsch](http://louiskirsch.com) (PhD Student, IDSIA)+* [Francesco Faccio](https://scholar.google.com/citations?user=0z3DkrkAAAAJ&hl=en&oi=ao) (PostDoc, IDSIA; Visiting Researcher, KAUST)+* [Jรผrgen Schmidhuber](http://www.idsia.ch/~juergen/) (Director, KAUST AI Initiative; Scientific Director, IDSIA)++Please read our [developer document](DEVELOPMENT.md) if you are interested in contributing.+++## Citation+Please cite our paper if you find the library useful or interesting.+```+@article{zhuge2024gptswarm,+ title={Language Agents as Optimizable Graphs},+ author={Zhuge, Mingchen and Wang, Wenyi and Kirsch, Louis and Faccio, Francesco and Khizbulin, Dmitrii and Schmidhuber, Juergen},+ journal={Technical Report},+ year={2024}+}+```++++