Bulk scans
A bulk scan launches every query under a topic (or a subset you select) at once, in a single HTTP roundtrip. You'd use it to seed a new workspace, take a snapshot before a launch, or refresh data when you've added new prompts.
When to use bulk vs single scan
| Use case | Pick |
|---|---|
| Iterating on a new prompt and want to see if it produces signal | Single scan |
| Snapshotting your whole topic at a point in time | Bulk scan |
| Seeding a brand-new workspace with baseline data | Bulk scan |
| Testing a hypothesis about a specific competitor surge | Single scan on a targeted query |
| Pre/post comparison for a launch (run before, run after) | Bulk scan twice |
How to run a bulk scan
- Open any topic detail page
- Click Bulk run (top-right action button)
- Pick how many queries to include — up to 25 in one bulk batch
- Confirm and run
The dispatcher launches each query's four-LLM call sequentially per query (so per-query free-tier enforcement still applies), with the platform calls within each query running in parallel.
Typical bulk timing:
| Queries | Wall-clock time |
|---|---|
| 5 | ~2 minutes |
| 10 | ~4 minutes |
| 25 | ~10 minutes |
You can leave the page while it runs. Each individual scan's status updates live in the scans list — pending → running → completed.
Quota accounting
Each query in a bulk counts as one scan. A 25-query bulk uses 25 against your monthly scan cap.
For free-tier orgs:
- The dispatcher does an up-front check before any scan starts: "is there room for this whole bulk in both the lifetime cap and the hourly cap?"
- If not, you get a clean error before any work begins — no half-completed bulks
For BYOK orgs there's no cap, but the dispatcher still serializes the per-query loop so the free-tier flow stays uniform.
What happens if a scan in the bulk fails
A single scan failing inside a bulk doesn't abort the rest. You'll see:
- 24/25 scans completed
- 1/25 scan marked
failedwith the platform-level error
The Inngest worker retries transient failures (5xx, timeouts) automatically. Permanent failures (content policy blocks, malformed prompts) mark the scan failed and the dispatcher moves on.
Bulk + scheduled
A scheduled scan can run a bulk on a cadence — daily, weekly, or monthly. Set the schedule on the topic detail page; choose which queries to include. The schedule fires through Inngest, off the Vercel 60-second function-timeout path, so even a 25-query Monday-morning bulk completes reliably.
See Scheduled scans for the full setup.
Cost expectations (BYOK)
A 25-query bulk on BYOK across all four platforms costs approximately:
- Free models / cheap-tier (gated) — N/A; bulk requires BYOK or pro tier
- GPT-4o / Claude Sonnet / Gemini Pro / Perplexity Sonar — $0.15 – $0.40 per bulk
That's a small fraction of the per-month value of having weekly snapshots. Most workspaces run 1–2 bulks per week.
See also
- Scheduled scans — putting bulks on autopilot
- Free tier limits — what the shared-key gates are
- Scans & prompts — what each scan actually does