Skip to main content

​
Get Started in 3 Steps

Follow this guide to upload your first PDF and start asking questions with AI-powered intelligence.
1

Sign In or Create an Account

Navigate to PDF AI and sign in with your account. If you don’t have one, create a new account using Clerk authentication.
Authentication is required to access PDF AI features. Your account keeps your PDFs and chat history private and secure.
2

Upload Your PDF

Once signed in, you’ll see the upload interface on the home page.To upload:
  • Click the upload area or drag and drop your PDF file
  • Wait for the file to upload to AWS S3
  • The system will automatically process and index your PDF
Maximum file size is 10 MB. Only PDF files are supported.
What happens behind the scenes:
  1. Your PDF is uploaded to AWS S3 storage
  2. The document is split into smaller chunks for processing
  3. Each chunk is converted into vector embeddings using OpenAI
  4. Embeddings are stored in Pinecone for fast semantic search
  5. A new chat session is created in the database
The upload is handled by the FileUpload component (src/components/ui/FileUpload.tsx:33-70):
const { getRootProps, getInputProps } = useDropzone({
  accept: { "application/pdf": [".pdf"] },
  maxFiles: 1,
  onDrop: async (acceptedFiles) => {
    const file = acceptedFiles[0];
    if (file.size > 10 * 1024 * 1024) {
      toast.error("File size too large. Please upload a file less than 10 MB.");
      return;
    }
    
    const data = await uploadToS3(file);
    mutate(data, {
      onSuccess: ({ chat_id }) => {
        toast.success("Chat created successfully! Redirecting...");
        router.push(`/chat/${chat_id}`);
      }
    });
  }
});
3

Start Chatting

After your PDF is processed, you’ll be redirected to the chat interface.Ask questions like:
  • “What is this document about?”
  • “Summarize the main points”
  • “Find information about [specific topic]”
  • “What does the document say about [question]?”
The AI will:
  1. Convert your question into vector embeddings
  2. Search for the most relevant sections in your PDF (top 5 matches with similarity > 0.7)
  3. Pass the relevant context to GPT-4
  4. Stream back an intelligent, context-aware response
Responses are streamed in real-time for a smooth experience. All messages are saved to your chat history.

​
Example Workflow

Here’s what a typical session looks like:
1

Upload a research paper

You upload a 5MB research paper about machine learning
2

Ask for a summary

You: “What are the main contributions of this paper?”AI: “Based on the document, this paper presents three main contributions: 1) A novel architecture for…” (streams full response)
3

Dive deeper

You: “Explain the methodology in detail”AI: “The methodology section describes a multi-stage approach…” (provides detailed explanation based on document content)
4

Find specific information

You: “What datasets were used for evaluation?”AI: “According to the paper, the authors evaluated their approach on three datasets: ImageNet, COCO, and…” (extracts specific details)

​
Tips for Best Results

Ask Specific Questions

Specific questions get better answers. Instead of “Tell me about this,” ask “What is the main argument in section 3?”

Use Context

Reference specific sections, page numbers, or topics from your PDF for more targeted responses.

Break Down Complex Queries

For complex topics, ask multiple focused questions rather than one broad question.

Iterate and Refine

If the answer isn’t quite right, rephrase your question or ask for clarification.

​
What’s Next?

Learn How It Works

Understand the RAG pipeline and vector embeddings that power PDF AI

Explore Features

Discover all the capabilities of AI-powered PDF chat

Upgrade to Pro

Get unlimited PDFs and priority processing with a Pro subscription

API Documentation

Integrate PDF AI into your own applications

​
Common Questions

Most PDFs process in 10-30 seconds depending on length. You’ll be redirected automatically when ready.
Yes! Each PDF gets its own chat session. Free users can upload up to 2 PDFs, Pro users get unlimited uploads.
Absolutely. Your PDFs and chats are private to your account. We use Clerk for secure authentication and all data is isolated by user ID.
Common issues include file size (must be under 10 MB) or format (must be PDF). Check the troubleshooting guide for solutions.

​
Need Help?

Troubleshooting Guide

Find solutions to common issues with uploads, chat, and subscriptions

Build docs developers (and LLMs) love

Get started for freeTalk to us