Vercel AI Chatbot

This is a clone of Vercel's AI Chatbot at commit 7d8e713.

Running locally

Requirements

• Docker
• A Sindri API Key: this can be generated by creating a Sindri account then visiting the API Keys page.

1. Clone the Repo

In a terminal,

git clone https://github.com/Sindri-Labs/vercel-ai-chatbot.git
cd vercel-ai-chatbot

2. Run (with Sindri and evllmp)

Modify the below command to include your Sindri API Key in place of YOUR_SINDRI_API_KEY then run it in the terminal (building may take several minutes)

OPENAI_API_KEY=YOUR_SINDRI_API_KEY docker compose up --build

You will know it is ready when you see logs similar to

vercel-ai-chatbot-frontend-1  |  ✓ Starting...
vercel-ai-chatbot-frontend-1  |  ✓ Ready in 542ms

3. Visit the Site and Start Chatting!

Visit http://localhost:3000 and start chatting as a guest user. Chat history will not be saved between sessions.

Notes:

• All chats are automatically encrypted with Sindri's Trusted Execution Environment (TEE)! The console logs for the evllmp container show encrypted/decrypted chats.
• Sindri is still in development. File uploads and web searches are still in development and currently do not work.

4 (Optional). Create an Account and Log in to Save your Chat History

Visit http://localhost:3000/register and create an account.

You can also visit the Account menu in the bottom of the left sidebar. Here you can create accounts and log into existing accounts.

5. Stop

To bring it all down, run:

docker compose down

You can bring it up again with docker compose up. This will skip the building phase and existing accounts and their chat histories will be restored.

Other

Environment Variables

Variable	Description	Default
AUTH_SECRET	JWT encryption secret	Required (create with openssl rand -base64 32)
BLOB_READ_WRITE_TOKEN	Vercel Blob Storage key	Required
MAX_COMPLETION_TOKENS	Maximum tokens for responses	1000
OPENAI_API_KEY	Your OpenAI API key	Required
OPENAI_BASE_URL	Custom OpenAI-compatible endpoint	https://api.openai.com/v1
OPENAI_MODEL	Model to use for chat completions	gpt-4o-mini
PORT	Port the server is exposed	3000
SINDRI_BASE_URL	Sindri OpenAI-compatible endpoint	https://sindri.app/api/ai/v1/openai
USE_TOOLS	Enable AI tools (weather, documents, etc.)	false

Key Modifications

• Docker Deployment: Complete containerized setup with PostgreSQL and Redis
• OpenAI Integration: Switched from xAI to OpenAI with configurable base URL
• PDF Support: Added PDF and JSON file upload support (up to 100MB)
• Authentication: Fixed auth flow and added guest mode
• Environment Variables: All configuration via Docker environment variables
• This Vercel version still uses the OpenAI /v1/chat/completions instead of the newer /v1/responses endpoint. Sindri does not yet support /v1/responses.