SmartDoc Finder

Inspiration

Every organization has tons of documents policies, manuals, PDFs, regulations but finding specific information inside them is often slow and frustrating. We wanted to build a tool that turns this static knowledge into something interactive and intelligent.

What it does

SmartDoc Finder is an AI-powered chatbot that lets users ask questions in natural language and instantly get accurate answers from company documents. It uses Elastic Search to find the most relevant content and Google Vertex AI (Gemini) to generate clear, human-like responses, including links to the original source.

How we built it

We set up an Elastic Search index and uploaded sample documents (PDF and text). We built a Python backend (FastAPI) to query Elastic and send retrieved passages to Gemini for answer generation. We created a simple web chat interface to provide an intuitive user experience. We deployed the solution using Google Cloud and integrated both Elastic and Vertex AI through their APIs.

Challenges we ran into

Configuring Elastic for multilingual (English & Polish) document search. Ensuring that Gemini generated answers were accurate and grounded in the retrieved sources. Managing API authentication between Elastic and Google Cloud securely. Optimizing query performance for larger document sets.

Accomplishments that we're proud of

Successfully integrated Elastic Search with Gemini, creating a smooth hybrid RAG pipeline. Built a working end-to-end chatbot within the hackathon timeframe. Implemented multilingual support and source linking to boost reliability. Delivered a clean, simple UI that anyone can use without training.

What we learned

How to use Elastic Search as a powerful retrieval engine for AI assistants. How to build a Retrieval-Augmented Generation (RAG) workflow with Google Vertex AI. Best practices for prompt engineering and grounding responses in real data. The importance of clear UX even for technical tools.

What's next for SmartDoc Finder

Adding user authentication and personalized document access. Supporting more file formats (Word, Confluence, Google Drive). Improving scalability to handle thousands of documents. Building analytics to track popular queries and knowledge gaps. Deploying as a plug-and-play tool for real companies.

Built With

• cloud-run)
• elastic-api
• elastic-search
• fastapi
• gemini-api
• google-cloud-(vertex-ai
• javascript
• python
• streamlit