🛠 Open Source Repo Agent (AgentStack + Multi-Retrieval Query)

This project is an AgentStack-powered AI pipeline that assists in managing GitHub issues efficiently. It uses Retrieval-Augmented Generation (RAG) with Multi-Retrieval Queries to analyze GitHub issues, extract relevant information, and update issues with related discussions.

🚀 Features

• Retrieves and processes GitHub issues from a repository.
• Uses Multi-Retrieval Queries to improve similarity search by reformulating queries from different perspectives.
• Stores issues in a vector database (Chroma) for efficient retrieval.
• Extracts issue URLs and adds comments to GitHub issues with related discussions.
• Runs as a structured pipeline using agentstack.

🛠 How It Works

The agent follows a structured pipeline:

1️⃣ create_vector_db (Task)

• Retrieves all issues from the GitHub repository.
• Stores them in ChromaDB for efficient retrieval.

2️⃣ full_rag_chain (Agent)

• Uses Multi-Retrieval Query Generation to generate alternative versions of the user query.
• Reformulates queries from different perspectives to enhance similarity-based retrieval.
• Retrieves the most relevant GitHub issues by overcoming distance-based limitations.
• Generates an AI-assisted response.

3️⃣ extract_issue_urls (Task)

• Extracts relevant issue links from the generated response.
• Adds a comment to the corresponding GitHub issue.

4️⃣ Pipeline Execution (run())

• Defines and executes the entire workflow using agentstack.
• Ensures that each step updates the state correctly.

---

🔍 Multi-Retrieval Query for Improved Similarity Search

What is Multi-Retrieval Query?

Instead of using a single search query, the Multi-Retrieval Query method generates five alternative perspectives of the user's query to:

• Improve similarity-based search results.
• Reduce distance-based search errors.
• Ensure better document retrieval.

Example of Multi-Retrieval Query Reformulation

Input Query:

"How do I extract a datetime from a timestamp?"

Generated Alternative Queries:

1. "Convert timestamp to datetime format in Python."
2. "How can I parse timestamps into datetime objects?"
3. "Best ways to retrieve datetime from a timestamp."
4. "Extract date and time from a given timestamp."
5. "Methods to transform a timestamp into a readable date."

These alternative queries increase retrieval accuracy, helping the AI find more relevant GitHub issues.

---

📌 Example Output

When the agent runs, it automatically adds a comment to the GitHub issue:

✅ Successfully added a comment to Issue #123:
The issue titled "How do I extract a datetime from a timestamp?" can be found at:
https://github.com/YOUR_REPO/issues/456

---

🔍 Troubleshooting

1️⃣ GitHub Authentication Issues?

Make sure your GITHUB_TOKEN is correct and has repo permissions.

2️⃣ LangChain Deprecation Warnings?

Use updated imports:

from langchain_community.vectorstores import Chroma
from langchain_community.embeddings import OpenAIEmbeddings

3️⃣ State Update Errors?

Ensure that each task/agent returns an updated state:

return {**state, "retrievelink": response_text}

---

🛠 Technologies Used

• Python 3.9+
• AgentStack
• LangChain
• ChromaDB
• GitHub API (PyGitHub)
• Multi-Retrieval Query Generation

---

📬 Contact

For questions or contributions, feel free to reach out:

• GitHub: Neel Harsola
• Email: [email protected]

---