Add Graph Recommendation Engine #791
Conversation
…per repo graphs as of now.
…from their respective folders
feat: Added a Chatbot Interface
For some reason the reasoning model isn't great at parsing out indicies from a JSON.
feat: Used Qwen3 reasoning model instead of llama3.2
chore: Added graphs that don't require per repository settings.
yyualvin
changed the title
feat: Use embeddings model instead of reasoning model
yyualvin
changed the title
Allows hot reloading of the app server so that a rebuild is not necessary for small changes in the Python Dash Application.
…ents in PR assignment visualization
Styling Contributors Section based on Figma Template
Revert "Styling Contributors Section based on Figma Template"
cdolfi
moved this from Backlog
to Testing | Waiting on PR response | Blocked
in Aspen Project Board
cdolfi
left a comment
cdolfi
left a comment
There was a problem hiding this comment.
please squash commits down, I have a few questions, and work with @EngCaioFonseca on if these changes would be compatible with a production environment and if not what needs to be iterated on
8Knot/prompt.md
Outdated
| @@ -0,0 +1,13 @@ | |||
| ## Background | |||
There was a problem hiding this comment.
what is the purpose of this file and why is it where it is?
There was a problem hiding this comment.
That was the prompt file for when I was using an LLM to select the graphs. I'll remove this since now I'm using an embeddings model.
| - ./env.list | ||
| environment: | ||
| - EIGHTKNOT_SEARCHBAR_OPTS_SORT=shortest # Try different options: shortest, longest, alphabetical | ||
| - EIGHTKNOT_SEARCHBAR_OPTS_MAX_RESULTS=5500 # Maximum number of results to return (org and repo) |
There was a problem hiding this comment.
Can you provide context to these changes?
| @@ -1,4 +1,4 @@ | |||
| FROM registry.access.redhat.com/ubi9/python-39:latest | |||
| FROM registry.access.redhat.com/ubi9/python-311:latest | |||
There was a problem hiding this comment.
Does it need to be at python 11? If so we will need to check to make sure everything else is compatible with this change
There was a problem hiding this comment.
I believe llamastack requires a Python version higher than 3.9, however I can revert it back as llamastack isn't being used currently.
| # top-level required modules w/o pinned dependencies | ||
| # DON'T ADD NEW DEPENDENCIES TO REQUIREMENTS.TXT MANUALLY | ||
| # ADD THEM TO REQUIREMENTS-BASE.TXT, THEN RUN | ||
| # pip3 freeze -r requirement-base.txt > requirements.txt |
There was a problem hiding this comment.
@EngCaioFonseca Is this still accurate post moving over to uv?
There was a problem hiding this comment.
@cdolfi no, you're right. It should be moved from pip3 to uv pip
|
@yyualvin Please add instructions to this on how to generate an API key and further description on how this works |
|
I feel like it would also be super nice to have this just fail gracefully too - i.e. if no API key is specified, the "Chat" tab just doesnt appear in the running 8Knot (i.e. without preventing it running) |
4 participants
Demo Video
Screen.Recording.2025-06-13.at.3.16.17.PM.mov
How this Works
Graph names and their descriptions are encoded as vectors in a Milvus vector database. When a user enters a query, a vector search will be initiated on the database, and the top 5 results are returned.
Graphs must be specified in graphs.json in this exact format.
{ "name" : "File Language By File", "about" : "Visualizes the percent of files or lines of code by language.", "id" : "gc_code_language" },Deployment
You must specify these environmental variables in the env.list file (the embeddings model endpoint, and the credentials). Please do not use a different endpoint.