v1.0.4-stableUpdated 2 days ago

The Analyze Endpoint

POST/api/analyze

The core engine of the API. This endpoint accepts a screenshot (base64) or raw conversation text, and uses Mistral Large with structured JSON output mode to generate evidence-based verdicts. Maude processes your data through a multi-step reasoning pipeline to ensure the analysis is both accurate and devastating.

Request Body

imageBase64Required*

Base64-encoded screenshot. Pixtral Large extracts text via OCR. *Either this or conversationText is required.

string

conversationTextRequired*

Raw conversation text (Person A / Person B format). Bypasses OCR for direct text analysis.

string

styleOptional

Maude's personality mode. Same reasoning engine, different delivery.

"default" | "analyst" | "coach" | "roast"

Example Request

{
  "conversationText": "Person A: Did you eat my leftovers?...",
  "style": "roast",
  "context": "We share a fridge at work"
}

Maude's Response JSON

{
  "success": true,
  "verdict": {
    "winner": "person_b",
    "gotCooked": "person_a",
    "confidence": 84,
    "oneLiner": "The label was right there and so was the audacity."
  },
  "meta": {
    "modelUsed": "mistral/mistral-large-latest",
    "promptVersion": "v3"
  }
}