Clairic vs Google Cloud Document AI

Clairic

Clairic's price is public at $0.10 per million input tokens with no output charge, and its docs publish JSON Schemas, limitations and when to use a decision model instead. It is a preview that Cloudflare calls an early checkpoint, served under an unversioned ID with no SLA, and its promised open weights have no date or licence.

Google Cloud Document AI

A public Discovery document with 42 methods, IAM roles that can limit a caller to processing, a fieldMask that trims responses, and a 99.9 per cent SLA on the US and EU endpoints. A processor has to be created before the first call, online requests stop at 15 pages, and a Google Cloud billing account with a card comes first.

Clairic

1. Send model as clairic in the body as well as in the URL. The input schema requires model and fields
2. Set reference_date, or add a line giving today's date to the state, for any field that depends on the current date
3. Treat a null value as not found, and read p_null, confidence or the 80% interval before acting. Typed output is well formed, not necessarily correct
4. Extract the inputs and do multi-step sums and table totals in code. Cloudflare says long arithmetic chains aren't reliable yet
5. Embed images as data URLs or base64 objects, at most four, 4 MiB each and 8 MiB in all. Remote image URLs are refused
6. Read the internal code on a 429. 3036 means the day's 10,000 free neurons are spent, 3040 means capacity, so retry later

Google Cloud Document AI

1. Create a processor first (processors.create or the console), then POST to https://LOCATION-documentai.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/processors/PROCESSOR_ID:process
2. Use the host that matches the processor's location, us-documentai.googleapis.com or eu-documentai.googleapis.com
3. Set fieldMask (for example text,entities) and imagelessMode to keep page images and token geometry out of the response
4. Send more than 15 pages through :batchProcess with Cloud Storage input and output, then poll the operation. Jobs unfinished after 24 hours are cancelled
5. Grant the service account roles/documentai.apiUser only. Failed requests (4xx or 5xx) are not billed, so a retry costs nothing extra
6. Treat extracted text as untrusted input