Inquira

Inquira is an AI-powered assistant that uses Retrieval Augmented Generation (RAG) to answer questions about Indium’s services. It leverages semantic search and generative models to return intelligent, context-aware responses from a curated knowledge base.

Features

• Retrieval-Augmented Generation (RAG) architecture.
• Semantic document search using Sentence Transformers.
• Vector similarity matching with FAISS.
• Language generation using Hugging Face Transformers (e.g., Mistral-7B).
• FastAPI-powered REST API for backend processing.
• Chat-ready frontend with real-time interaction.

Usage

Basic setup and dependency management:

1. Install pipenv if you don't have it:

   pip install pipenv

2. Activate the virtual environment:

   pipenv shell

3. Install required libraries from the Pipfile:

   pipenv install

   or

   pip install -r requirements.txt

4. Ensure that the .env file contains your OpenRouter API key.

While to run the code:

cd app

Local FastAPI server (uses static html based UI):

uvicorn main:app --reload

With Streamlit-UI (in separate terminal):

streamlit run st.py

On browser:

Visit http://localhost:8000

Description about various files:

• app/static/chat.html: HTML file for the frontend chat interface.
• app/static/chatbot.js: JavaScript file handling frontend logic.
• app/.env: Contains Hugging Face API key used for generation.
• app/config.py: Sets API keys for callback.
• app/embedding.py: Embeds input text into vector form using all-MiniLM-L6-v2.
• app/main.py: Entry point of the FastAPI app; handles routing and endpoints.
• app/rag_chain.py: Combines the retriever and generator into a complete RAG chain.
• app/retriever.py: Handles document retrieval using semantic similarity.
• app/st.py: Streamlit based UI.
• app/vector_store.py: Initializes and manages the FAISS vector index.
• data/documents.json: Contains the documents about Indium’s services that are used to answer user queries.
• Pipfile: Specifies the dependencies for the project.
• Pipfile.lock: Ensures consistent dependency versions.
• requirements.txt: Lists all top-level Python dependencies.