Scribd
Upload
7 views4 pages

AIChatbot Assignment

The document outlines the requirements for building a 'Chat with your files' application that allows users to upload documents and images, store them in a bucket, and interact with an AI assistant using RAG. It specifies mandatory tech stack components, core features such as authentication, file upload, RAG pipeline, and chat experience, as well as non-functional requirements and evaluation criteria. The project aims to provide a seamless chat experience grounded in user-uploaded content while ensuring security and proper access control.

Uploaded by

Saksham Jain
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
7 views4 pages

AIChatbot Assignment

The document outlines the requirements for building a 'Chat with your files' application that allows users to upload documents and images, store them in a bucket, and interact with an AI assistant using RAG. It specifies mandatory tech stack components, core features such as authentication, file upload, RAG pipeline, and chat experience, as well as non-functional requirements and evaluation criteria. The project aims to provide a seamless chat experience grounded in user-uploaded content while ensuring security and proper access control.

Uploaded by

Saksham Jain
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd

Assignment: File-Chat RAG Chatbot

Goal

Build a “Chat with your files” application where users upload documents and images, store
them in a bucket,

and chat with an AI assistant that answers questions grounded in the uploaded content
using RAG.

Tech Stack (Mandatory)

Frontend: React (TypeScript)

Backend: Django + Django REST Framework

Optional: Django Channels for realtime or streaming chat

Core Features

1. Authentication

- Basic email/password or token-based authentication

- Users must only access their own files and chats

2. File & Image Upload + Bucket Storage

Users should be able to:

- Upload files (PDF, TXT, DOCX) and images (PNG, JPG, JPEG)

- View a list of uploaded assets with metadata (type, size, uploaded_at, etc.)

- Rename/update metadata

- Delete assets (and associated indexed data)


All raw files and images must be stored in object storage (not local disk).

Bucket options (choose one and justify):

- AWS S3 / S3-compatible storage

- Google Cloud Storage

- Azure Blob Storage

- Cloudflare R2

- DigitalOcean Spaces

- Supabase Storage

- Cloudinary (especially suitable for images)

3. RAG Pipeline (Text + Images)

For documents:

- Extract text from files

- Chunk text using a reasonable strategy

- Generate embeddings for chunks

- Store embeddings with sufficient metadata for retrieval

For images:

- Extract textual meaning using an AI vision/OCR approach

(e.g., OCR, image captioning, or multimodal models)

- Convert extracted text or descriptions into embeddings

- Store image-derived embeddings with metadata

During chat:

- Retrieve relevant text and/or image-derived chunks


- Generate grounded answers using retrieved context

- Clearly respond when answers are not present in the uploaded content

4. Chat Experience

- Chat UI where users can select one or more assets (or all assets)

- Persistent conversation history

- Each assistant response should surface the retrieved context

(snippets, citations, or references to documents/images)

AI Resources (choose and justify)

LLMs:

- OpenAI (GPT-4.1 / GPT-4o-mini with vision support)

- Anthropic Claude (vision models)

- Google Gemini (multimodal)

- Open-source multimodal models (with justification)

Embeddings:

- OpenAI embeddings

- Cohere embeddings

- Gemini embeddings

- Open-source embeddings (bge, instructor, etc.)

Vector Stores:

- Postgres + pgvector (preferred)

- Pinecone
- Qdrant

- Weaviate

- Chroma

Non-functional Requirements

- Clean README with setup instructions

- Environment variable management

- Proper access control and user isolation

- Logging for ingestion, indexing, and retrieval

- Basic error handling (invalid file type, extraction failure, missing keys)

Deliverables

- GitHub repository or zip archive

- README and .[Link]

- 2–5 minute demo video showing upload → chat → delete flow

Evaluation Criteria

- Correctness of the end-to-end RAG flow (text + images)

- Completeness of asset CRUD functionality

- Quality of retrieval and grounding in responses

- Image understanding approach and justification

- Security and access control

- Code quality, structure, and maintainability

- Documentation clarity

- Bonus features (streaming, abstractions, multi-document/multi-image chat)

You might also like