Draft: Retrieval Augmented Generation (RAG) for GitLab Duo
What does this MR do and why?
This MR adds a blueprint for Retrieval Augmented Generation (RAG) for GitLab Duo.
Visit https://gitlab.com/gitlab-org/gitlab/-/blob/rag-for-duo-chat/doc/architecture/blueprints/gitlab_duo_rag/index.md to review the blueprint. Here is the documentation structure:
-
index.md
... Overview of the process-
postgresql.md
... Retrieve from PostgreSQL -
elasticsearch.md
... Retrieve from Elasticsearch -
vertex_ai_search.md
... Retrieve from Google Vertex AI Search
-
A few notes:
- We happily accept co-authors or reviwers. Feel free to add your name to the
authors
orapprovers
column if you want to actively participate the discussions. - We have a weekly sync meeting at
Tuesday 9:00 AM UTCMonday 14:00 UTC.
Proposed timeline:
- ~ 19th February: Conduct thorough research on the proposed solutions and put as many data points as possible in this blueprint.
- 19th February: Undraft this MR to ask reviews from wider audiences and a coach. See Architecture Design Workflow for more info.
- 22nd February: Conclude the research and review by merging this MR. Remained open discussions should be explicitly documented and followed-up in a separate MR.
Related to
- Duo Chat:
- Code Suggestions and X-Ray:
- Vector stores and Semantic search
- PoC: Embedding storage and AI in Elastic (#439358 - closed)
- Handling Chat embeddings for SM/Dedicated (&11600 - closed)
- GitLab Embeddings Support (#438176 - closed)
- Store embeddings (#404396 - closed)
- https://gitlab.com/gitlab-org/gitlab/-/issues/438178+
- https://docs.gitlab.com/ee/development/advanced_search.html (Current Elastic Search integration)
- Iteration plan: RAG- Ben Venker, Pini Wietchner (#441110)
- Repository X-Ray build vector storage prototype... (#440822 - closed)
- https://gitlab.com/gitlab-com/gl-infra/production-engineering/-/issues/25052+
- Allow Elasticsearch framework to index any data (#442197 - closed)
MR acceptance checklist
Please evaluate this MR against the MR acceptance checklist. It helps you analyze changes to reduce risks in quality, performance, reliability, security, and maintainability.
Screenshots or screen recordings
Screenshots are required for UI changes, and strongly recommended for all other merge requests.
Before | After |
---|---|
How to set up and validate locally
Numbered steps to set up and validate the change are strongly suggested.
Edited by Shinya Maeda