Offline RAG Chatbot for Windows
Budget: ₹1,500 – ₹12,500 INR
I’m looking for a lightweight Windows-based chatbot that runs 100 % locally—no cloud calls, no telemetry. The agent should combine an open-source generative model with a Retrieval-Augmented Generation (RAG) pipeline so it can:
• ingest and index my own text, image, and PDF files
• extract key data, summarise content on demand, and generate automated responses
• answer questions through a simple chat window using only the indexed material
Scope at this stage is modest. A working proof-of-concept that demonstrates end-to-end flow—from local file ingestion to accurate chat responses—is enough. Please include:
1. A straightforward installer or script that sets up the model, vector store, and any OCR you choose for images/PDFs.
2. A minimal desktop or browser-based UI to type queries and view responses.
3. Clear instructions so I can add or remove local data and completely cut network access without breaking the app.
Open-source frameworks (e.g., LangChain, llama.cpp, Haystack) are fine as long as all components stay on-device. Keep RAM/VRAM requirements reasonable for a typical consumer PC. Once the prototype works reliably offline, I’ll test it with a small document set; if it passes, we can discuss possible future refinements.
Let's dive to bigger versions once the trail version is successful
• ingest and index my own text, image, and PDF files
• extract key data, summarise content on demand, and generate automated responses
• answer questions through a simple chat window using only the indexed material
Scope at this stage is modest. A working proof-of-concept that demonstrates end-to-end flow—from local file ingestion to accurate chat responses—is enough. Please include:
1. A straightforward installer or script that sets up the model, vector store, and any OCR you choose for images/PDFs.
2. A minimal desktop or browser-based UI to type queries and view responses.
3. Clear instructions so I can add or remove local data and completely cut network access without breaking the app.
Open-source frameworks (e.g., LangChain, llama.cpp, Haystack) are fine as long as all components stay on-device. Keep RAM/VRAM requirements reasonable for a typical consumer PC. Once the prototype works reliably offline, I’ll test it with a small document set; if it passes, we can discuss possible future refinements.
Let's dive to bigger versions once the trail version is successful