This document outlines how your code is used and processed when using Recurse ML products
Last Updated: 16th December, 2024
When utilising RecurseML, your code is connected to our Webserver endpoint hosted on CloudRun via a webhook from your GitHub or GitLab account. This webhook is triggered upon a Pull Request (PR) or Merge Request (MR) on a code repository connected to RecurseML by a user.
Our Webserver reads the connected codebase and performs preprocessing to dispatch a bug analysis report. Relevant subportions of code are sent to LLM APIs and our internal models for inference. The Webserver combines these results, compiles a final bug report, and sends them to the connected GitHub or GitLab account. This report appears as a series of comments on the PR/MR for which RecurseML has been run.
Where a user submits feedback to RecurseML, we store the feedback together with the original comment (and associated code snippets) in a Postgres server. This server is hosted on Recurse ML's Google Cloud Platform Instance.