Dependency Risk Auditor

• 

  Live Reasoning

• 

  Output Table

• 

  Output Visuals

• 

  Output Recommendations

Inspiration

I have been stuck with catching up to repos that have severely outdated dependencies making it hard to onboard developers and maintain the security and integrity of programs. So something that audits all the repos of a company and suggests actions to keep dependencies updated would take a lot of the pain away

What it does

It looks at the repos database and derives the dependencies that the targeted repos are using and then compares them to the dev deps dataset for advisories and suggest action plans to update outdated dependencies

How we built it

we use nebius token factory to amke an agent that use Craft mcp to interact with the 2 database and does its magic. it is deployed on streamlit.

Challenges we ran into

A challenge we ran into was the speed of the ask_data/generate_sql functions of Craft MCP as we didn't know how to use it properly and that ended up considerably slowing down our application. We overcame this by sidestepping the MCP's generate_sql function for certain data questions for now to increase the speed and demonstrate the use case better. After understanding Craft MCP more we could implement a better version that uses craft MCP

Accomplishments that we're proud of

Building our first agent that uses the craft MCP to an extent and using the nebius token factory for model access.

What we learned

We learned more about interacting with databases using the craft mcp, and how llm agents can be used to understand github repos better and recommend actionable steps to improve repo health

What's next for Dependency Risk Auditor

We could expand this to consider more factors and instead of just dependcy health it could check repo health for other stuff like code quality, test coverage, etc.

Built With

• craft
• nebius
• python
• streamlit