API Documentation — myQA.team

IFS Analyzer

POST/api/v1/ifs-analyzer/analyze

Arbitrate one IFS Food v8 observation graded "D": keep D or requalify as a Major non-conformity.

Authentication

Requires ifs-analyzer:analyze permission. See Authentication.

Request — JSON

1curl -X POST https://api.myqateam.com/api/v1/ifs-analyzer/analyze \
2  -H "Authorization: Bearer myqa_..." \
3  -H "Content-Type: application/json" \
4  -d '{
5  "requirement": "4.12.1",
6  "scope": "Production de plats cuisinés réfrigérés",
7  "observation": "Le test du détecteur de métaux de la ligne 2 n'\''a pas été réalisé au démarrage de 6h.",
8  "validator_remark": "Risque de corps étrangers sur les lots de 6h à 10h.",
9  "reference": "NC-2026-0042",
10  "critic": true
11}'

Fields

Field	Type	Required	Description
`requirement`	string	Yes	IFS Food v8 requirement number, e.g. `4.12.1`. We use the official text (its KO flag comes back as `requirement.is_ko`); unknown numbers return `REQUIREMENT_NOT_FOUND`. List them with GET /api/v1/standards/IFS-FOOD/clauses.
`scope`	string	Yes	Scope (périmètre) of the audited process. The analysis weighs the hazards of this process. ≤ 1,000 characters.
`observation`	string	Yes	The auditor's observation graded D. ≤ 5,000 characters.
`validator_remark`	string	No	The technical validator's remark (VT). Used by the critic. ≤ 3,000 characters.
`reference`	string	No	Your own id for the observation, echoed back as-is. ≤ 200 characters.
`critic`	boolean	No	`true` runs the critic, a second opinion that checks the analysis against `validator_remark`; `false` skips it (faster, fewer tokens, `critic` is `null` in the response). Omit it to use the default, which is on.

Text is sent to Google Gemini on the EU multi-region endpoint. Emails, phone numbers and similar identifiers are removed before the model sees them. The analysis is written in French.

Response — synchronous

json

1{
2  "success": true,
3  "data": {
4    "analysis_id": "6f0c…",
5    "reference": "NC-2026-0042",
6    "requirement": { "number": "4.12.1", "text": "Sur la base des risques, des procédures…", "is_ko": true },
7    "grade": "Majeur",
8    "reasoning": "1. **Analyse du Risque lié au Scope :** …",
9    "critic": {
10      "verdict": "L'analyse de l'IA ne valide pas la notation en D, et la demande de la VT est pertinente"
11    },
12    "model": { "analyst": "gemini-3.5-flash", "critic": "gemini-3.5-flash" },
13    "prompt_versions": { "analyst": 1, "critic": 1 },
14    "settings_version": 1,
15    "processing_time_ms": 18450,
16    "token_usage": { "input_tokens": 2100, "output_tokens": 950, "total_tokens": 3050 },
17    "validation_status": "pending_validation",
18    "requires_human_validation": true,
19    "created_at": "2026-10-01T12:00:00Z"
20  }
21}

grade: one of A, B, C, D, Majeur, KO.
reasoning: the technical justification, in markdown.
critic: null when you sent critic: false (or the critic is turned off by default). Otherwise verdict is one of: "L'analyse de l'IA valide la notation "D"", "L'analyse de l'IA ne valide pas la notation en D, et la demande de la VT est pertinente", or "L'analyse de l'IA ne valide pas la notation en D, Pas de remarque de la VT".
prompt_versions / settings_version: which configuration produced the result, for traceability.
requires_human_validation is always true: an IFS expert must confirm the suggestion.

Errors

HTTP	Code	Description
400	`INVALID_INPUT`	Missing field, wrong type, too long, or invalid JSON. `details.errors` lists every problem.
400	`REQUIREMENT_NOT_FOUND`	The requirement number does not exist in IFS Food v8.
402	`COIN_LIMIT_EXCEEDED`	Your organization has used its coins for the month.
502	`MODEL_ERROR`	The model answer was cut off or unreadable. Safe to retry.
503	`AI_RATE_LIMITED`	The AI model is temporarily overloaded. We already retried for about a minute; retry after the `Retry-After` header (60 s).
500	`PROCESSING_ERROR`	Unexpected failure. Retry, and quote the `X-Request-Id` if it persists.

See the full error code reference on the authentication page.