Bivia API

Bivia API

Bivia is a calibrated decision engine. Ask a question with context — text, images, audio, documents, or signals — and get back a typed, structured answer with a probability distribution and honest abstention when the model isn’t sure.

Decision types

Type	What you get
Boolean (/boolean)	yes or no with P(value)
Choice (/choice)	One label from your list, with full distribution
Auto (/decide)	Let the model pick the best type
Score (/score)	A number on your scale, with per-level mass
Multi-select (/set)	Zero or more labels from your options
Span (/span)	Character offsets into the input text
Ground (/ground)	Bounding boxes in images or time intervals in video

Quick start

curl -s https://api.articul8.ai/bivia/boolean \
  -H "Authorization: Bearer $A8_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "state": "Customer threatened to cancel today.",
    "question": "Is this urgent?"
  }'

{
  "type": "boolean",
  "value": "yes",
  "probability": 0.92,
  "distribution": { "yes": 0.92, "no": 0.08 },
  "status": "answered",
  "model": "dec-1"
}

Key concepts

• Calibrated — probabilities are meaningful, not just softmax scores
• Abstention — when unsure, the model says so (status: "abstained")
• Multimodal — attach text, images, audio, documents, or signals
• Evidence — responses include evidence.spans for traceability
• Typed — every answer is machine-readable, ready for downstream logic

SDKs

Install the SDK for your language:

pip install bivia          # Python
npm install @articul8/bivia   # Node.js
go get github.com/articul8/bivia-go  # Go