>> Purpose: Split documents into chunks suitable for RAG pipelines and vector databases. LLMs have context limits; large documents need to be chunked before embedding or retrieval.

>> How it works: Upload PDF, DOCX, TXT, or HTML. Select a tier. Lower tiers use rule-based splitting. Higher tiers use OpenAI embeddings or Claude for boundary detection. AI Pro adds per-chunk summaries and entity extraction.

>> Payment: Free tier available (100 pages/day). Paid tiers require USDC on Solana via x402 protocol. Payment included in request header after wallet signature.

ENDPOINTS

GET / - API info and config
GET /health - Health check
POST /estimate - Get price estimate for file
POST /chunk/demo - Free tier (100 pages/day)
POST /chunk/structure - Structure tier ($0.002/page)
POST /chunk/semantic - Semantic tier ($0.004/page)
POST /chunk/ai - AI tier ($0.01/page)
POST /chunk/ai-pro - AI Pro tier ($0.025/page) [ASYNC]
GET /job/{id}/status - Poll async job status
GET /job/{id}/result - Get async job result

AI PRO ASYNC PROCESSING

AI Pro tier uses async processing for long-running LLM analysis. Instead of waiting, you receive a job_id immediately:

1. POST to /chunk/ai-pro → returns job_id
2. Poll /job/{id}/status for progress
3. Fetch /job/{id}/result when completed

Jobs expire after 1 hour. See API docs for details.

RESPONSE FORMAT

{
  "success": true,
  "tier": "structure",
  "total_chunks": 4,
  "chunks": [
    {
      "text": "chunk content...",
      "metadata": {
        "index": 0,
        "quality_score": "A",
        "has_code": false,
        "has_table": false
      }
    }
  ],
  "metadata": {
    "filename": "doc.pdf",
    "estimated_pages": 2
  }
}