Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/PySpur-Dev/pyspur
Graph-Based Editor for LLM Workflows
https://github.com/PySpur-Dev/pyspur
agent agents ai javascript llm llm-inference openai python react workflow
Last synced: 5 days ago
JSON representation
Graph-Based Editor for LLM Workflows
- Host: GitHub
- URL: https://github.com/PySpur-Dev/pyspur
- Owner: PySpur-Dev
- License: apache-2.0
- Created: 2024-09-23T17:24:52.000Z (4 months ago)
- Default Branch: main
- Last Pushed: 2024-12-29T10:54:31.000Z (10 days ago)
- Last Synced: 2024-12-29T11:27:59.962Z (10 days ago)
- Topics: agent, agents, ai, javascript, llm, llm-inference, openai, python, react, workflow
- Language: TypeScript
- Homepage: https://pyspur.dev
- Size: 4.28 MB
- Stars: 1,014
- Watchers: 15
- Forks: 69
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- awesome-generative-ai-data-scientist - Pyspur - Based Editor for LLM Workflows (Miscellaneous)
- awesome_ai_agents - Pyspur - Graph-Based Editor for LLM Workflows (Building / Workflows)
- awesome_ai_agents - Pyspur - Graph-Based Editor for LLM Workflows (Building / Workflows)
README
# PySpur - Graph-Based Editor for LLM Workflows
https://github.com/user-attachments/assets/9128885b-47ba-4fc6-ab6b-d567f52e332c
# ✨ Core Benefits
## Modular Building Blocks
https://github.com/user-attachments/assets/6442f0ad-86d8-43d9-aa70-e5c01e55e876
## Debug at Node Level:
https://github.com/user-attachments/assets/6e82ad25-2a46-4c50-b030-415ea9994690
## Evaluate Final Performance
https://github.com/user-attachments/assets/4dc2abc3-c6e6-4d6d-a5c3-787d518de7ae
## Coming soon: Self-improvement
https://github.com/user-attachments/assets/5bef7a16-ef9f-4650-b385-4ea70fa54c8a
# 🕸️ Why PySpur?
* **Easy-to-hack**, eg., one can add new workflow nodes by simply creating a single Python file.
* **JSON configs** of workflow graphs, enabling easy sharing and version control.
* **Lightweight** via minimal dependencies, avoiding bloated LLM frameworks.
# ⚡ Quick start
#### [ Instructions for Unix-like systems. Support for Windows PC coming soon ]
You can get PySpur up and running in three quick steps.
1. **Clone the repository:**
```sh
git clone https://github.com/PySpur-com/pyspur.git
cd pyspur
```
2. **Create a .env file:**
Create a `.env` file at the root of the project. You may use `.env.example` as a starting point.
**If you plan to use third party model providers, please add their API keys in the .env file in this step**.
3. **Start the docker services:**
```docker compose up --build -d```
This will start a local instance of PySpur that will store spurs and their runs in a local SQLite file.
Note: For some environments you may want to try: ```sudo docker compose up --build -d```.
4. **Access the portal:**
Go to `http://localhost:6080/` in your browser.
Enter `pyspur`/`canaryhattan` as username/password.
Set up is completed. Click on "New Spur" to create a workflow, or start with one of the stock templates.
5. **[Optional] Manage your LLM provider keys from the app:**
Once PySpur app is running you can manage your LLM provider keys through the portal:
Select API keys tab
Enter your provider's key and click save (save button will appear after you add/modify a key)
# Using PySpur with Ollama (Local Models)
PySpur can work with local models served using Ollama.
Steps to configure PySpur to work with Ollama running on the same host.
### 1. Configure Ollama
To ensure Ollama API is reachable from PySpur, we need to start the Ollama service with environment variable `OLLAMA_HOST=0.0.0.0` . This allows requests coming from PySpur docker's bridge network to get through to Ollama.
An easy way to do this is to launch the ollama service with the following command:
```
OLLAMA_HOST="0.0.0.0" ollama serve
```
### 2. Update the PySpur .env file
Next up we need to update the `OLLAMA_BASE_URL` environment value in the `.env` file.
If your Ollama port is 11434 (the default port), then the entry in `.env` file should look like this:
```
OLLAMA_BASE_URL=http://host.docker.internal:11434
```
(Please make sure that there is no trailing slash in the end!)
In PySpur's set up, `host.docker.internal` refers to the host machine where both PySpur and Ollama are running.
### 3. Launch the PySpur app
Follow the usual steps to launch the PySpur app, starting with the command:
```
docker compose up --build -d
```
### 4. Using Ollama models in the app
You will be able to select Ollama models [`ollama/llama3.2`, `ollama/llama3`, ...] from the sidebar for LLM nodes.
Please make sure the model you select is explicitly downloaded in ollama. That is, you will need to manually manage these models via ollama. To download a model you can simply run `ollama pull `.
## Note on supported models
PySpur only works with models that support structured-output and json mode. Most newer models should be good, but it would still be good to confirm this from Ollama documentation for the model you wish to use.
# 🗺️ Roadmap
- [X] Canvas
- [X] Async/Batch Execution
- [X] Evals
- [X] Spur API
- [x] Support Ollama
- [ ] New Nodes
- [X] LLM Nodes
- [X] If-Else
- [X] Merge Branches
- [ ] Tools
- [ ] Loops
- [ ] Pipeline optimization via DSPy and related methods
- [ ] Templates
- [ ] Compile Spurs to Code
- [ ] Multimodal support
- [ ] Containerization of Code Verifiers
- [ ] Leaderboard
- [ ] Generate Spurs via AI
Your feedback will be massively appreciated.
Please [tell us](mailto:[email protected]?subject=Feature%20Request&body=I%20want%20this%20feature%3Ai) which features on that list you like to see next or request entirely new ones.