# The Universal Summarizer: Cutting Through the Noise
## How a single API endpoint turned 50-page PDFs and 1-hour videos into actionable insights.

The true eureka moment came when I pointed Kagi's **Universal Summarizer API** at a 48-page technical specification PDF and a 55-minute keynote presentation on WebAssembly component models.

Traditional LLM summarization often requires downloading the file, extracting raw text, chunking it, handling token limits, and babysitting rate limits. Kagi's summarizer does all of this server-side via a single clean request:

```bash
# Summarize any public URL, technical PDF, or YouTube video
curl -X POST "https://kagi.com/api/v0/summarize" \
  -H "Authorization: Bot YOUR_KAGI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
    "summary_type": "takeaway",
    "engine": "cecil"
  }'
```

### Key Capabilities:
- **Multimodal ingestion**: Works transparently on articles, research PDFs, YouTube videos, and podcasts.
- **Engine flexibility**: Choose between ultra-low latency (`cecil`) and high analytical depth (`agnes` or `muriel`).
- **Targeted formats**: Output executive summaries, bulleted takeaways, or full chapter breakdowns.

🔗 **Official Resources:**
- [Universal Summarizer API Reference](https://help.kagi.com/kagi/api/summarizer.html)
- [Kagi Engine & Model Overview](https://help.kagi.com/kagi/ai/summarizer.html)