Skip to main content
If you don’t find your question here, follow the link inside each answer to the page that goes deeper, or contact support.

​
Documents and processing

PDF, JPEG, PNG, TIFF, GIF, WebP, and CSV, plus Word, Excel, PowerPoint, OpenDocument, and Rich Text documents (the legacy .doc, .xls, and .ppt binaries are not accepted). Multi-page TIFFs are supported. PDF is preferred for multi-page documents. Office and CSV documents are read as text, so they carry no bounding boxes and count as a single page. See supported formats.
Yes. Multi-page PDFs and multi-page TIFFs run through the workflow as a single document. When the workflow splits the document into chunks, Ingestly creates a child run for each chunk and the parent run waits in Awaiting Children until they all finish.
The run moves to Failed status and downstream nodes do not execute. Open the run detail to inspect the failed step execution for the input it received and the error it produced.
Drop a review node after the step you want to gate. Choose Always to pause every run, LowConfidence to pause when confidence drops below a threshold, or Unverified Fields to pause whenever the upstream extract node returns unverified fields. See the human review guide.
Each workflow has its own retention setting in days. After that window the document file and run history are removed. Configure retention on the workflow detail page.

​
Workflows and runs

The workflow is probably not activated. A workflow must be active and have valid configuration (at least one trigger, all nodes connected, validation passing) before it accepts documents.
Yes. Branch upstream with a classify node (rules or sample documents) and route each document type to its own subgraph. Classify is what picks the processing path; split only finds where one upload’s sub-documents start and end, so a bundle holding several document types goes split then classify, with classify choosing the path for each sub-document. See conditional routing.
The parent run paused while child runs finish. This happens when Ingestly splits a multi-page document into chunks and processes each chunk independently. The parent resumes once every child run completes.
Yes. Open the document on the workflow detail page and re-run it. Each re-run creates a new run and consumes credits.
Create an alert rule on a metric like Failure Rate, Callback Failure Rate, or Credit Balance. Pick In-App or Email as the channel.

​
API and integration

POST the document as multipart/form-data to https://api.ingestly.ai/workflows/{workflowId}/trigger with your API key in the X-Api-Key header. The workflow must have a webhook trigger. See the webhook trigger guide for code samples.
Add the X-Workspace-Id header. If you omit it, the request targets your default workspace. See API overview.
Limits scale with your plan: 10 requests per minute on Free, 60 on Basic, 300 on Standard, 1,000 on Advanced. Every response includes RateLimit-Limit, RateLimit-Remaining, and RateLimit-Reset headers. See rate limits.
Add a callback output node to your workflow. When the run completes, Ingestly posts the results to the URL you configure. See webhooks and callbacks.
Generate a webhook signing key and verify the HMAC-SHA256 signature on the X-Ingestly-Signature header against the raw request body. The same guide covers zero-downtime key rotation.
Rotation keeps both the old and new key active and signs callbacks with both, so your integration keeps working while you cut over. Revoking removes the signature header entirely; update your application to stop verifying signatures before you revoke.

​
Billing and credits

Ingestly reserves credits upfront at run start, based on the most expensive branch your workflow could take multiplied by the number of pages. The actual deduction matches that worst-case path, not just the steps that ended up running. See credits for the per-node breakdown and worked examples.
No. Ingestly reserves credits at run start based on the most expensive branch your workflow could take. That reservation isn’t refunded if a later step fails, so the deduction matches the worst-case path regardless of how far the run actually got.
New organizations get 200 welcome credits so you can build and test workflows before subscribing. It is a one-time grant, not a monthly allowance.
No. The 200 welcome credits are granted once, when your organization is created, and they are not topped up at the start of each month. Monthly credit allocation begins when you move to a paid plan. See pricing.
Create an alert rule on the Credit Balance metric with the Less Than operator so you get notified when your balance dips below a threshold.
The Upload, Email and Webhook triggers cost 1 credit per page, collected when the document arrives and is parsed rather than when a run starts (see document ingestion); the first run on the document shows that credit on its trigger step. The internal forward trigger is free. Outputs (callback, email output, forward output) are free, as are review, transform, merge, HTTP action, and non-AI validation. Action cost comes from parse, extract, classify (both modes), AI validation, and filter or split when they read text to decide. See credits.

Glossary

Definitions for every term used in Ingestly

Credits

Per-node credit costs and worked examples

API overview

Authentication, headers, and the public endpoint

Quickstart

Build your first workflow in 5 minutes