Content Ingestion API

The ingestion endpoint accepts raw text content and processes it through the full NeuralSnap pipeline — extracting, synthesizing, and crystallizing it into structured Neural Snapshots. This is the simplest way to get unstructured content into your brain.

For more control over the extraction process, use the Crystallize API which supports source type hints and multi-phase processing. For curated knowledge from books and courses, use the Specific Knowledge API.

When to use Ingest vs Crystallize

Use Ingest for quick, fire-and-forget content processing. Use Crystallize when you need control over source type, multi-phase extraction, or want to process meeting transcripts specifically.

Ingest Content

POST/ingest

Process raw text content into structured snapshots.

Request Body

Property	Type	Description
`brain_id`*	string	UUID of the brain to store results in.
`content`*	string	Raw text content to process. Minimum 20 characters.
`title`	string	Optional title or label for the ingestion (e.g. 'Q4 Planning Notes').
`source_type`	enum	Hint for processing: manual, document, article, url. Default: manual.
`tags`	string[]	Tags to apply to all generated snapshots.
`extract_memories`	boolean	Also extract memories (facts, decisions, insights) from the content. Default: false.

bash

curl -X POST https://api.neuralsnap.ai/v1/ingest \
  -H "Authorization: Bearer ns_test_abc123" \
  -H "Content-Type: application/json" \
  -d '{
  "brain_id": "brain_abc123",
  "content": "In our Q4 planning meeting, we decided to double down on short-form video. TikTok and Reels drive 3x more signups than blog posts. The production cost is lower once the workflow is dialled in. We also agreed to hire a dedicated video editor by January.",
  "title": "Q4 Planning Notes",
  "source_type": "document",
  "tags": ["planning", "q4", "video"],
  "extract_memories": true
}'

json

{
  "data": {
    "ingestion_id": "ing_x1y2z3",
    "title": "Q4 Planning Notes",
    "source_type": "document",
    "snapshots": [
      {
        "id": "snap_a1b2c3",
        "name": "Short-form video outperforms blog content for signups",
        "type": "Conviction",
        "core": "TikTok and Reels drive 3x more signups than blog posts with lower production cost",
        "confidence": 0.85,
        "tags": ["planning", "q4", "video"]
      }
    ],
    "memories": [
      {
        "id": "mem_d4e5f6",
        "content": "Decided to hire a dedicated video editor by January",
        "memory_type": "decision",
        "tags": ["planning", "q4", "video"]
      }
    ],
    "token_usage": {
      "input": 2400,
      "output": 1800,
      "total": 4200
    }
  }
}

Processing Pipeline

Content goes through three phases:

Phase 1Extract

Raw claims, ideas, and statements are pulled from the text. Each extraction is tagged with its source context.

Phase 2Synthesize

Related extractions are grouped, redundancies removed, and connections mapped. The system identifies which ideas support, contradict, or elaborate on each other.

Phase 3Crystallize

Synthesized insights are structured into full 16-field Neural Snapshots. Type, confidence, and significance are estimated. Edges are created to connect new snapshots to existing knowledge.

Token usage

Ingestion uses tokens proportional to input length. A typical 1000-word document uses approximately 3K–5K tokens. Check your usage in Settings → Billing.

Error Handling

Property	Type	Description
`400`	Bad Request	Content too short (minimum 20 characters) or missing brain_id.
`401`	Unauthorized	Missing or invalid API key.
`402`	Payment Required	Token limit exceeded for your plan.
`422`	Unprocessable	Content could not be processed (e.g. not in a supported language).
`429`	Rate Limited	Too many ingestion requests. See Rate Limiting docs.

Crystallize

Profile & Usage