Skip to main content

​
YouTube videos

Drop a YouTube URL. ChatGrid extracts the transcript, chunks it, and vectorizes it. 
curl -X POST https://api.chatgrid.ai/v1/boards/{boardId}/documents/vectorize \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://youtube.com/watch?v=your-video-id"}'
Long videos (1+ hours) may take 30-60 seconds to process. Use the Jobs endpoint to check status.

​
PDFs

Upload a PDF, then vectorize it. 
# Step 1: Upload the file
curl -X POST https://api.chatgrid.ai/v1/boards/{boardId}/assets \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "file=@architecture.pdf"

# Step 2: Vectorize it (use the asset URL from step 1)
curl -X POST https://api.chatgrid.ai/v1/boards/{boardId}/documents/vectorize \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url": "{assetUrl}", "metadata": {"source": "architecture.pdf"}}'
Text is extracted from the PDF (including scanned pages via OCR), then vectorized.

​
Web pages

Any public URL works. ChatGrid fetches and vectorizes the content. 
curl -X POST https://api.chatgrid.ai/v1/boards/{boardId}/documents/vectorize \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://blog.example.com/postgres-scaling-tips"}'

​
Plain text

Paste in notes, decisions, meeting summaries — anything. 
curl -X POST https://api.chatgrid.ai/v1/boards/{boardId}/documents/vectorize \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "content": "Decision: We chose Inngest over Temporal for job orchestration. Reasons: managed service, no infra to operate, built-in rate limiting. Temporal is the upgrade path if we outgrow Inngest.",
    "metadata": {"source": "decision-log", "date": "2026-03-04"}
  }'

​
Search your content

Once content is vectorized, search it semantically. 
curl -X POST https://api.chatgrid.ai/v1/boards/{boardId}/documents/search \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query": "why did we choose Inngest?", "limit": 3}'
Returns the most relevant chunks with their source metadata: 
{
  "object": "list",
  "data": [
    {
      "content": "Decision: We chose Inngest over Temporal for job orchestration...",
      "metadata": {"source": "decision-log", "date": "2026-03-04"},
      "similarity": 0.92
    }
  ]
}

​
In Claude Code

All of this works via natural language with MCP: 
> Add this YouTube video to the board: https://youtube.com/watch?v=...
> Upload and vectorize architecture.pdf to the backend board
> Search the board for "why did we choose Inngest?"

​
Next steps

Team Knowledge

Share a board across your whole team

Streaming

Stream AI responses in real time