Semarize

API

One call.
Structured signals back.

Pass your conversation content and a Kit code. Get deterministic JSON with scores, evidence, and confidence values. Same schema every time.

Request & response

Content in. Signals out.

The API accepts conversation content plus a Kit code and returns structured evaluation results with evidence.

POST /v1/runs
// Send conversation + kit
{
"kit_code": "A1B2C3D4",
"input": {
"transcript": [
{
"speaker": "rep",
"text": "What's driving the urgency..."
}, ...]
}
}
Response 200
// Structured output returned
{
"run_id": "run_...",
"status": "succeeded",
"output": { "bricks": {
"pain_is_specific": {
"value": 64,
"reason": "Pain identified but not quantified",
"evidence": ["...spending 3 hours per week..."]
}, ...
} }
}

Input

Flexible input.
Any conversation source.

Semarize accepts transcripts, emails, chat logs, and notes. Speaker attribution is optional but recommended for richer evaluation.

Input object

The input field accepts any JSON object. Structure your transcript data however suits your workflow.

input.transcript[]
input.speaker
input.text

Context fields

Add any context fields to the input object. Your Bricks and Kits can reference them during evaluation.

account_id
stage
participants
meeting_title

Options

Control evaluation behaviour: which Kit to run, sync or async mode, and optional version pinning.

kit_code
mode (sync / async)
kit_version_id (optional)

Output

Predictable structure. Every time.

Every response follows the same schema. Kit and version identifiers, plus a signals array where each entry is a typed, evidenced evaluation.

Response envelope

run_idUnique identifier for this evaluation
statusRun state: queued, running, succeeded, failed
kit_version_idExact version for auditability
output.bricksMap of Brick evaluation results
duration_msHow long the run took to complete

Each brick contains

valueThe typed result (boolean, number, string)
confidenceConfidence score from 0 to 1
reasonShort explanation of why this value was returned
evidence[]Exact text spans from the conversation

Guarantees

Built for production systems

Semarize is designed to be a reliable component in your data infrastructure. Here's what the API guarantees.

Deterministic output

Same Kit version + same content = same schema. Outputs are typed and predictable.

Versioned evaluation

Kit and Brick versions are immutable. Historical results are always reproducible.

Evidence-backed

Signals include evidence spans and reasoning when enabled. Trace any value back to the source text.

Secure by default

API key authentication. Content is processed and not stored beyond your retention settings.

Integration

Fits your existing stack

Semarize integrates with your existing data flow. Pull transcripts from capture tools, run Kits, push signals downstream.

Example: n8n / Zapier workflow
// 1. Webhook: transcript ready from Fireflies
// 2. Fetch transcript via Fireflies API
// 3. POST to Semarize /v1/runs
{
"kit_code": "A1B2C3D4",
"input": transcript_from_fireflies
}
// 4. Parse response
// 5. Write signals to Salesforce opportunity fields
// 6. If risk_score > 0.7 → alert in Slack

Ready to integrate?

Get your API key and start extracting structured signals from your first conversation.