RAG Implementation for Market Health Reporter

[evgenydmitriev]

We have a tool called Market Health Reporter for creating automated reports with metrics spikes and their interpretation using our free Market Health API.

To participate in the challenge, submit a pull request that adds Retrieval Augmented Generation (RAG) functionality to the Market Health Reporter to obtain additional article context from external resources and request a review from this issue assignees. The ultimate goal is to improve the model output to better match our contribution guidelines and the overall structure of this example article. You can find additional information about the reporter in the old challenges. Expanding the pull request description with your methodology can help us better understand your reasoning and evaluate your submission faster.

Anyone can participate in this challenge. You don't need anyone's approval to try! All submissions will be reviewed by the maintainers, and additional changes to your pull request may be asked of you to bring your submission to improve code quality, security, and/or efficiency. See more details on the Challenge Program and check out the success stories of the challenge winners.

Looking forward to your submissions!

How to Claim Bounty

Email your bitcoin or stablecoin payment address to [email protected]. Expect all payouts to be completed by the end of the month.

[winterForestStump]

Hello,

The #480 PR introduces a Retrieval-Augmented Generation (RAG) approach to the Market Health Reporter.

• The script fetches additional context from a local database of HTML files to enhance the original report generated by the Market Health Reporter.
• Uses Cohere's API for topic generation, document embedding, and final report generation.
• Implements HNSWLIB for similarity search, creating a lightweight, serverless solution.
• Via a /rag:: comment on an issue, generates a new report and opens a pull request with the result.

More detailed information in the readme file.

This is my first contribution to a public project, and I am still learning best practices in coding and project collaboration. I appreciate any feedback you can provide, especially regarding code efficiency, style, or improvements to align better with your standards.

@Kseymur , I would be grateful if you could kindly review this PR when you have time.