IFS Analyzer
IFS Analyzer API
Arbitrate IFS Food v8 audit observations graded "D": the AI says whether D is appropriate or the observation should be requalified as a Major non-conformity, with a technical justification. An optional second pass checks that conclusion against the technical validator's remark.
Endpoints
- POST /api/v1/ifs-analyzer/analyze — one observation, answered synchronously.
- POST /api/v1/ifs-analyzer/batches — up to 100 observations, processed asynchronously.
- GET /api/v1/ifs-analyzer/jobs/:jobId — batch status and results.
- GET /api/v1/standards/IFS-FOOD/clauses — the 232 IFS Food v8 requirements you can reference.
Permission required
ifs-analyzer:analyze on the API key used. See Authentication.
What you submit
- The IFS Food v8 requirement number (e.g. 4.12.1). We use the official requirement text and return its KO flag.
- The scope (périmètre) of the audited process and the auditor's observation.
- Optionally, the technical validator's remark, your own reference for the observation, and critic: true or false to run or skip the critic (on by default).
What you get back
- A recommended grade (D, Majeur, …) and a technical justification in French, in markdown.
- The critic's verdict on the validator's remark, unless you sent critic: false.
- The model, prompt and settings versions used, token usage and processing time.
Every result is a suggestion flagged requires_human_validation: an IFS expert must confirm it.
Processing model
One observation takes about 15 to 40 seconds. For several, send a batch and poll the job, or pass a callback_url to receive a signed webhook.