Grant Writing Assistant

RAG

LLM

Pinecone

VectorStore

Docker

CI/CD

LLMOps

Langsmith

Railway

FASTAPI

Streamlit

Author

Prasad Bhoite

Typing Animation

Tip

Grant Writing Assistant is an AI-powered RAG application designed to help users query and synthesize information from a custom knowledge base about the NeighborhoodHELP (NHELP) program.
It leverages Retrieval-Augmented Generation (RAG) to provide accurate, context-aware answers for grant writing and program documentation.

🚀 Features

💬 Conversational AI: Remembers chat history to handle follow-up questions.
📚 Dynamic Knowledge Base: Ingests information from PDFs and live websites.
🔎 Source Transparency: Displays references for every answer.
⚡ Streaming Responses: Token-by-token output for responsiveness.
🛠️ Developer Dashboard: Manage prompts and knowledge base via a Streamlit admin app.

🧱 MLOps & Professional Practices

🔄 CI/CD: Automated linting with Ruff using GitHub Actions.
📊 Evaluation: Built-in pipeline with Ragas to measure faithfulness, relevancy, and answer quality.
👁️ Observability: End-to-end tracing with LangSmith.
📦 Containerization: Fully reproducible deployments using Docker.
📝 Prompt Management: Externalized into version-controlled prompts.yml.

🧬 Tech Stack

Framework: Streamlit
LLM Orchestration: LangChain
LLMs: OpenAI GPT Models
Vector Store: Pinecone
Observability: LangSmith
Evaluation: Ragas
Deployment: Docker + Railway
CI/CD: GitHub Actions

🏗️ System Architecture

This project uses a monolithic architecture where a single Streamlit application handles both the user interface and the RAG logic.
Data ingestion is managed by an offline script (ingest.py).

RAG Pipeline Diagram

⚡ Getting Started

▶️ Running the Application

Access in Browser Open: https://grantwritingassistantnhelp-production-c6ef.up.railway.app/