Configuration reference

OHM Studio's defaults are tuned so a fresh hospital deployment works with zero configuration. This page is the full menu of what's available when you want to customise — every server env var, every per-API toggle in Studio, every SDK option.

Per-API settings (Studio → API → Settings tab)

Rate-limit behaviour: Per-key per-minute. When exceeded, returns 429 Too Many Requests with retry-after header. SDK throws OHMRateLimitError carrying retryAfterSec.

Per-call SDK options

Every SDK method accepts these (all optional, all default off):

Server environment variables

These live in apps/api/.env (or your container env). All optional — sensible defaults baked in.

Database / storage

LLM / extraction

The hospital orchestrator has no LLM secrets — every AI feature (audio transcription, structured extraction, visit-notes chat with hospital-DB tools, Studio playgrounds, codes AI Assistant, all SDK endpoints) calls the OHM Engine, which holds the provider keys internally. The hospital boundary returns vendor-neutral results and error codes (see below).

Hospital-side AI config:

Async-job worker

Rate limiting

Studio UI

Branding (optional)

SDK client options

OHM is hospital-deployed (see Versions → 0.12.0). Each hospital exposes its own API URL — point baseUrl at the hospital you're integrating with. https://api.ohm.doctor is OHM's own demo hospital and works for evaluation; production integrations should target the customer hospital.

new OHM({
  apiKey: "ohms_test_xxx",                         // required
  baseUrl: process.env.OHM_API_URL!,               // e.g. https://api.<hospital>.example
  timeoutMs: 60_000,                               // default 60s — bump for large audio
  maxRetries: 2,                                   // default 2 — auto-retry on transient
  fetch: globalThis.fetch.bind(globalThis),        // override for proxies / Node 18+
  onUsage: (event) => mySentry.capture(event),     // optional telemetry hook
  acknowledgeBundledKey: false,                    // RN only — required for ohms_live_*
  mock: false,                                     // unit-test mode
});

Webhook receiver (when using async + webhooks)

Verify signatures + dedupe deliveries:

import { createHmac } from "node:crypto";

const seenDeliveries = new Set();   // replace with Redis SET in prod

app.post("/ohm-callback", express.json(), (req, res) => {
  // 1. ACK fast — Stripe-style, return 200 BEFORE doing work.
  res.status(200).end();

  // 2. Idempotency — same delivery may arrive twice on retry.
  const id = req.headers["x-ohm-delivery-id"];
  if (seenDeliveries.has(id)) return;
  seenDeliveries.add(id);

  // 3. HMAC signature verification (per-job secret stored when you
  //    created the job — fetch from your DB by jobId).
  const job = await fetchJob(req.body.jobId);
  const expected = createHmac("sha256", job.webhookSecret)
    .update(JSON.stringify(req.body))
    .digest("hex");
  if (`sha256=${expected}` !== req.headers["x-ohm-signature"]) {
    console.warn("invalid signature; dropping");
    return;
  }

  // 4. Process.
  if (req.body.event === "extraction.job.completed") {
    await persistResult(req.body);
  }
});

"Just defaults, nothing custom" deployment

If you skip this page entirely, here's what a hospital gets out of the box:

• 60 rpm per API key
• 100 MB per audio upload
• Foundation block enforced on every extraction
• No payload retention (privacy-first)
• No PHI redaction (LLM sees the transcript verbatim)
• Auto-retry on transient LLM failures (2 retries)
• 24-hour idempotency window
• Stripe-style webhook retry schedule (5min → 30min → ... → 24h)
• 100-row default invocation log on the Studio Logs tab
• Async worker enabled (1 worker per API pod)

That's a working production setup. Tune only what you need.

Configuration order of precedence

When the same setting can be controlled in multiple places:

1. Per-call SDK option — wins always (caller's intent is explicit).
2. Per-API Studio setting — applied when the SDK didn't override.
3. Server env var default — when neither of the above is set.

Example: rate limit

• API key has rateLimitPerMinute = 200 set in Studio
• STUDIO_DEFAULT_RATE_LIMIT_PER_MIN = 60 env var
• → Effective limit: 200 (Studio setting wins).

Where to file feedback

Configuration UX is iterated based on what hospitals actually deploy. If a setting is missing or default is wrong for your workflow, drop a note in your deployment channel — we tune the defaults across releases.