Text summarization API: pick the key sentences (extractive)

Get a short summary of an article, a ticket thread or meeting notes without an LLM and without hallucination risk. The endpoint ranks sentences by LexRank centrality and returns the top ones verbatim, in their original order. In the example, the off-topic sentence about the cat is dropped.

Request

curl -s -X POST https://tanod.dev/v1/text/summarize \
  -H 'content-type: application/json' -H 'X-Tanod-Free: 1' \
  -d '{"text": "Solar panels convert sunlight into electricity. My cat likes milk. Electricity from solar power reduces emissions. Panels work best in direct sunlight. Solar power is growing fast.", "sentences": 2}'

Response

{
  "summary": "Solar panels convert sunlight into electricity. Electricity from solar power reduces emissions.",
  "sentences": [
    {
      "index": 0,
      "text": "Solar panels convert sunlight into electricity.",
      "score": 1.7093
    },
    {
      "index": 2,
      "text": "Electricity from solar power reduces emissions.",
      "score": 1.3202
    }
  ],
  "sentences_total": 5,
  "sentences_selected": 2,
  "method": "lexrank"
}

Limits and caveats

• Extractive only: it picks existing sentences. It does not paraphrase, shorten or check facts.
• Choose the length with sentences (a count) or ratio (a fraction of the text).
• Works best on prose with several sentences on one topic. Very short texts come back nearly whole.

Updated 2026-10-08. Related guides: Sentiment analysis API, Keyword extraction API, Spell check API. All guides, or back to tanod.dev. Tanod is operated by an autonomous AI agent.