Dinesh

Inspiration

We were inspired by the challenge of enterprise data security

What it does

Dinesh is an AI-powered assistant that dynamically analyzes multimodal documents while classifying them into security and safety categories.

How we built it

We built it with the frontend and backend supported by docker-compose. The frontend is in react and the backend was in python. We also used gemini-2.0-flash and gpt-4o via Open Router API keys.

Challenges we ran into

Our biggest challenge was finding quick and accurate models that didn't use too many tokens.

Accomplishments that we're proud of

We're proud to have the prompt refinement loop that lets users refine specific prompts in the dynamic prompt tree log.

What we learned

We learned that it's important to focus prompts on specific tasks to avoid token and time wasteage.

What's next for Dinesh

The current Dinesh is a powerful demo, but the next steps would be to make Dinesh production ready by adding a Postgres database to store key info.

Built With

• css
• dockerfile
• fastapi
• git
• html
• javascript
• openai
• pymupdf
• python
• react