ChatPDF-GPT

Introduction

ChatPDF-GPT is an innovative project that harnesses the power of the LangChain framework, a transformative tool for developing applications powered by language models. This unique application uses LangChain to offer a chat interface that communicates with PDF documents, driven by the capabilities of OpenAI's language models.

In this project, the language model is connected to other data sources and allows interaction with its environment, thus embodying the principles of the LangChain framework. Users can upload a PDF document, which is then processed and saved in Pinecone, a vector database, and Supabase storage. Users can then chat with the uploaded PDF, with the AI utilizing the content of the document to engage in a meaningful conversation.

The project relies on the Next.js framework, a leading choice for creating robust, full-stack Web applications. The UI components are beautifully crafted using the Radix UI library and styled with Tailwind CSS, based on the elegant template provided by shadcn/ui.

Features

1. Upload a PDF: Users can upload a PDF document, which is then stored in Pinecone vector database and Supabase storage.
2. Chat with PDF: The application processes the content of the PDF and allows users to engage in a conversation with the document using OpenAI API.
3. PDF Preview: Users can preview the actual PDF document they are interacting with, using the powerful React component package, @react-pdf-viewer.
4. List PDFs: The application provides a list of all uploaded PDF documents stored in the Supabase database.
5. Delete a PDF: Users have the ability to remove a document from the database.
6. Cite Sources: The AI chat interface provides sources from the PDF for every reply, allowing users to navigate directly to the reference in the PDF.

Usage Examples

ChatPDF-GPT is equipped with examples that illustrate various operations such as:

1. Interacting with Pinecone: saving embeddings, deleting a record.
2. Uploading a file to Supabase storage and also deleting a file.
3. Listing available documents from Supabase database.
4. Previewing a PDF using the @react-pdf-viewer.
5. Navigating directly to the source of an AI reply in the PDF.

Quick Testing Using the Demo

To test the functionality of this project using the demo, you will need to provide your own credentials for OpenAI, Supabase, and Pinecone. For Supabase, you can follow the step-by-step guide provided below to setup and retrieve the necessary credentials. For acquiring credentials for OpenAI and Pinecone, please consult the corresponding documentation as a step-by-step guide may not be available. Always ensure you are following the latest instructions provided by the respective services.

1. OpenAI

OpenAI

2. Supabase

1. Creating a New Project in Supabase:

   • Open your web browser and navigate to app.supabase.io.
   • Click on "New project".
   • Enter your project details and wait for the new PostgreSQL database to launch. It may take a few minutes for the process to complete.

2. Retrieving the Database Connection URL:

   • Once your database is successfully created, navigate to your project's page.
   • Go to the "Settings" section.
   • Click the "Database" tab in the sidebar.
   • Scroll down to the bottom of the page and look for the "Connection string" section.
   • Choose "Nodejs" and copy the URL.

This connection string will be used for the DATABASE_URL environment variable in your application.

3. Retrieving the Connection Pooling Connection String:
   • Still in the "Database" tab of the "Settings" section, scroll to find the "Connection Pooling" section.
   • Choose "Nodejs" and copy the Connection Pooling URL.

This URL will be used for the DIRECT_URL environment variable in your application.

4. Retrieving Storage Keys:
   • From your project's page, go to the "Settings" section.
   • Click the "API" tab in the sidebar.
   • Here, you will find your SUPABASE_URL and SUPABASE_KEY. Copy these values.

The SUPABASE_URL is the URL for your project, while SUPABASE_KEY is the public anonymous key for your project.

5. Setting up the Supabase Bucket:

   • From your project's page, go to the "Storage" section.
   • Here, create a new bucket or use an existing one. The name of the bucket will be used as SUPABASE_BUCKET in your application.

6. Setting Up Environment Variables in Your Application:

   • Now, navigate to your project where you are using Supabase.
   • Add the following environment variables with the values you copied from the Supabase console:
     • DATABASE_URL
     • DIRECT_URL
     • SUPABASE_KEY
     • SUPABASE_URL
     • SUPABASE_BUCKET

These keys will allow your application to interact with the Supabase services.

7. Managing Storage Bucket Policies:
   • Navigate to the "Storage" section from your project's page.
   • Click on the bucket for which you want to manage policies.
   • Click on the "Policies" tab to view the existing policies for the bucket.
   • To create a new policy, click the "Create policy" button and fill in the policy details as per your requirement.
   • To edit an existing policy, click on the policy in the list and make the necessary changes.

Please note that while it's possible to set a policy that makes your storage bucket publicly accessible, you should do this with caution. Making your bucket publicly accessible means that anyone with the URL to an object can access it. This might be useful for testing, but for production applications, you should consider more restrictive policies to ensure the security of your data. Always consult the Supabase documentation or a security expert to understand the implications of different policies.

With this, you should be able to set up Supabase for your project and manage storage policies as per your requirements.

3. Pincone

Pinecone

Setup and Installation

To set up and run ChatPDF-GPT on your local machine, follow the steps below:

1. Clone the project repository:

   git clone https://github.com/anis-marrouchi/chatpdf-gpt.git
   

2. Navigate into the project directory and install the dependencies using pnpm:

cd chatpdf-gpt
pnpm install

3. Create a .env file in the root directory and fill in your credentials (OpenAI, Pinecone, Supabase) as indicated in the .env.example file.

4. Create the database schema using Prisma. You must make you have run the prisma generate command prisma generate

   npx prisma migrate dev --name init
   

5. Start the server:

   npm run dev
   

Contribution

ChatPDF-GPT is an open-source project and we warmly welcome contributions from everyone. Please read our contributing guide for more details on how to get started.

Credits

This project stands on the shoulders of giants. Our work would not be possible without the vast array of libraries, frameworks, and tools that the open source community has produced. Specifically, we would like to express our appreciation to:

1. The LangChain team for their groundbreaking framework for applications powered by language models.

2. OpenAI for their state-of-the-art language models, which make the chat functionality possible.

3. Supabase for their open-source Firebase alternative which we used to build secure and performant backends.

4. Pinecone for their vector database that allows easy and efficient storage and retrieval of vector embeddings.

5. Next.js and Vercel for their comprehensive framework which allowed us to build this full-stack Web application with ease.

6. shadcn for their elegant UI components which we built upon to create a beautiful and user-friendly interface.

7. Radix UI for their robust, accessible and customizable component library that forms the backbone of our UI.

8. @react-pdf-viewer for their powerful React component, which lets users preview the actual PDF document they are interacting with.

And all the other dependencies, both listed and not listed, that contributed to the realization of this project. Our contribution is modest in comparison to their collective effort.

License

ChatPDF-GPT is open-source software licensed under the MIT license.