Document Chat

Chat with your documents using AI. Upload PDFs, paste URLs, import YouTube videos, or paste text — then ask questions and get intelligent answers with citations.

Overview

Document Chat uses Retrieval-Augmented Generation (RAG) to provide accurate, grounded answers from your content. Every response includes citations linking back to the exact source passages.

Supported Sources

Getting Started

Source Types

File Upload

Upload documents directly from your computer. Supported formats:

URL Import

Import content from any public web page:

# Supported URLs
https://example.com/article
https://docs.example.com/guide
https://blog.example.com/post

The system will:

• Fetch the page content
• Extract the main article/content
• Remove navigation, ads, and boilerplate
• Preserve headings, lists, and formatting

YouTube Import

Import YouTube video transcripts:

# Supported URL formats
https://www.youtube.com/watch?v=VIDEO_ID
https://youtu.be/VIDEO_ID

Requirements:

• Video must have captions (auto-generated or manual)
• Videos without captions cannot be imported

The system imports the full transcript with timestamps, allowing you to ask questions about any part of the video.

Paste Text

Paste any text content directly:

• Meeting notes
• Email threads
• Code snippets
• Research notes

Chat Interface

Asking Questions

Ask questions in natural language:

• "What are the key findings in this report?"
• "Summarize the main arguments"
• "What does it say about X?"
• "Compare the approaches mentioned in section 2 and 3"

Citations

Every AI response includes numbered citations:

The report indicates revenue grew 35% year-over-year [1],
driven primarily by enterprise expansion [2].

Click any citation to:

• View the original passage
• See the page number or location
• Navigate to that section in the document

Chat Settings

Customize chat behavior:

Collections

Organize documents into collections for:

• Project Organization: Group related documents
• Multi-Document Chat: Chat across an entire collection
• Team Collaboration: Share collections with team members

Creating Collections

1. Click New Collection in the sidebar
2. Enter a name and optional description
3. Choose a color for visual organization
4. Click Create

Moving Documents

• Drag documents into collections
• Use the document menu → Move to Collection
• Documents can belong to one collection at a time

Collection Chat

Chat with all documents in a collection at once:

1. Select a collection in the sidebar
2. Click Chat with Collection
3. Ask questions that span multiple documents

API Access

Access Document Chat via the REST API.

Upload Document

curl -X POST https://api.novakit.ai/v1/documents/upload \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "file=@document.pdf" \
  -F "collection_id=optional-collection-id"

Import from URL

curl -X POST https://api.novakit.ai/v1/documents/import \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "source_type": "url",
    "url": "https://example.com/article",
    "name": "Example Article"
  }'

Import YouTube

curl -X POST https://api.novakit.ai/v1/documents/import \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "source_type": "youtube",
    "url": "https://www.youtube.com/watch?v=VIDEO_ID"
  }'

Chat with Document

curl -X POST https://api.novakit.ai/v1/documents/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "document_ids": ["doc_123", "doc_456"],
    "message": "What are the key findings?",
    "model": "gpt-4o"
  }'

Response Format

{
  "content": "The key findings include...[1]...[2]",
  "citations": [
    {
      "index": 1,
      "chunk_id": "chunk_abc",
      "document_id": "doc_123",
      "text": "Original passage...",
      "page": 5
    }
  ],
  "usage": {
    "input_tokens": 1234,
    "output_tokens": 567
  }
}

Billing

Document Chat uses credits from your Chat Tokens quota:

Best Practices

Document Preparation

• Clean formatting: Remove unnecessary headers/footers
• Text-based PDFs: Ensure PDFs contain selectable text
• Reasonable length: Very long documents may have lower retrieval accuracy

Effective Questions

• Be specific: "What does section 3 say about pricing?" vs "Tell me about pricing"
• Reference context: "Based on the Q4 report, what were the growth metrics?"
• Ask follow-ups: Build on previous answers for deeper exploration

Collection Strategy

• Group by project or topic
• Keep collections focused (5-20 documents)
• Use descriptive names

Troubleshooting

Document Processing Failed

Chat Issues

Retry Failed Documents

For URL and YouTube imports that failed:

1. Go to AI → Documents
2. Find the failed document
3. Click Retry to attempt processing again

Limits