PowerDB drafting

The difference from /reports/generate is the input. There you supply structured readings; here you supply what the technician actually wrote down, and the endpoint does the extraction.

GET/ai/powerdb/types

POST/ai/powerdb

POST/ai/powerdb/reportreturns a file

Field	Type	Default
`equipment_notes`	string	Required. Free text.
`device_type`	string	`"lv_breaker"`, one of the fifteen PowerDB slugs
`output_format`	string	`"pdf"` · also `"excel"` report only
`include_summary`	boolean	`false`. Off here, so the download is immediate
`job_name`, `client`, `technician`, `date`	string	`""`

cURL

curl

curl -X POST "$BASE/ai/powerdb" \
  -H "Authorization: Bearer $TOKEN" -H 'Content-Type: application/json' \
  --max-time 120 \
  -d '{
        "device_type": "mv_breaker",
        "equipment_notes": "15kV vacuum breaker, Cutler-Hammer VCP-W,
          megger at 5kV: A-B 8200, B-C 8050, C-A 8310 megohm.
          Ductor 100A: 42, 40, 44 microhm. Timing within spec.
          Contacts clean, no erosion. Passed."
      }'

TypeScript

client.ts

const form = await api('/ai/powerdb', {
  method: 'POST',
  body:   JSON.stringify({
    device_type:     'mv_breaker',
    equipment_notes: notes,
  }),
  // Node's fetch has no default timeout; browsers cap at ~300s.
  // Either way, be explicit. This call takes tens of seconds.
  signal: AbortSignal.timeout(120_000),
})

Python

client.py

form = s.post(f"{BASE}/ai/powerdb", json={
    "device_type":     "mv_breaker",
    "equipment_notes": notes,
}, timeout=120).json()   # not the default 30

Extraction is a draft, not a reading

A value the model could not find is left empty rather than guessed, but a value it misreads will look exactly like one it read correctly. Every generated form is a drafting aid that a qualified person signs. That is not a disclaimer, it is how the endpoint is meant to be used. Do not build a pipeline that files these unreviewed.

Getting started

Core concepts

Database API

Documents API

Operations API

Resources