Bluesky Labeler Collective Development Tool - Prototype

Prototype agent for the Collaborative Engineering project.

Overview

The project contains logic for the Bluesky Labeler developer agent workflow prototype that should:

• Answer general questions about Bluesky, the AT protocol, and Labelers using web search or RAG
• Evaluate the semantic relevance of retrieved documents based on user queries
• Reformulate user queries for better document retrieval performance
• Generate a proposal for a label based on user input and context
• Enable user approval or editing of the proposal
• Generate a sample labeler TypeScript project based on the approved proposals

Tip

Detailed Requirements and Specifications are on a live Google doc.

Directory Structure

The project is a Langgraph Server application and follows the required application structure configuration for deployment.

└── 📁bsky-collective-eng
    └── 📁brainstorming_agent
        └── 📁constants
            ├── output_samples.py
            ├── prompt_templates.py
            ├── retrieval_sources.json
        └── 📁utils
            ├── __init__.py
            ├── nodes.py
            ├── rag_utils.py
            ├── state.py
            ├── test_func.ipynb
            ├── tools.py
        ├── __init__.py
        ├── agent.py
        ├── requirements.txt
    └── 📁data
        └── 📁bsky-docs
    ├── .gitignore
    ├── langgraph.json
    └── README.md

Important

The project needs a .env file with the OPENAI_API_KEY to use the GPT-4.1 model and TAVILY_API_KEY for the web search tool. Be sure to add that file after cloning this repo.

Installation & Usage

Using Langgraph Studio

Note

You'll need to add the LANGSMITH_API_KEY to your .env file. You can create a LangSmith account if needed.

To install Langgraph Studio

Installation

Clone this repository

Add the .env file in the root of the project

Initialize and activate a virtual environment

conda create -n my_virtual_env python=3.12
conda activate my_virtual_env

Install dependencies (from the root of the project folder)

pip install -r requirements.txt

and run

langgraph dev

Note

You'll need to run langgraph dev --tunnel if you are running on browsers that block http connections by default (e.g., Safari, Brave Browser).

More Resources

• Multi-agent systems
• Langchain tools. The protoype uses the Tavily Web Search tool
• LangSmith - Get Started
• Document loaders. The prototype uses the WebBaseLoader document loader for the retrieval tool.
• ChatOllama