API Documentation

Stripe-style developer experience for behavioral intelligence

Getting Started

B_Act Labs provides a simple, RESTful API for integrating behavioral intelligence into your AI systems. The API handles voice, visual, and semantic signal parsing as essential infrastructure.

Authentication

All API requests require an API key. Include it in the Authorization header:

bash

curl -H "Authorization: Bearer YOUR_API_KEY" \
  https://api.b-act.com/v1/analyze

1. Initialize an Agent

Create a behavioral intelligence agent instance for your AI system:

python

import b_act

agent = b_act.Agent(
    name="my-vision-agent",
    model="claude-3.5-sonnet",
    provider="anthropic",
    signals=["voice", "visual", "semantic"]
)

# Agent is now ready to receive signals
print(f"Agent {agent.id} initialized")

2. Send Behavioral Signals

Stream multimodal signals to the behavioral intelligence layer:

javascript

const response = await fetch('https://api.b-act.com/v1/analyze', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_API_KEY',
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({
    agent_id: 'my-vision-agent',
    signals: {
      voice: {
        transcript: 'I need help with this',
        emotion: 0.78,
        clarity: 0.96
      },
      visual: {
        objects: ['person', 'screen'],
        confidence: 0.92
      },
      semantic: {
        intent: 'request_help',
        context_depth: 0.85
      }
    }
  })
});

const result = await response.json();
console.log('Behavioral analysis:', result.emotional_state);

3. Receive Behavioral Analysis

The API returns real-time behavioral intelligence to inform your model's response:

json

{
  "agent_id": "my-vision-agent",
  "timestamp": "2024-03-25T14:30:00Z",
  "emotional_state": {
    "valence": 0.62,
    "arousal": 0.71,
    "confidence": 0.92
  },
  "signal_quality": {
    "voice": 0.96,
    "visual": 0.89,
    "semantic": 0.94
  },
  "behavioral_recommendations": {
    "confidence_adjustment": 1.08,
    "response_formality": "professional",
    "tone_suggestion": "supportive"
  },
  "latency_ms": 14
}

4. Complete Integration Pattern

Here's how to integrate behavioral intelligence into your AI pipeline:

python

import anthropic
import b_act

# Initialize both agents
b_act_agent = b_act.Agent("my-agent", model="claude-3.5-sonnet")
client = anthropic.Anthropic()

def process_with_behavioral_intelligence(user_input, voice_signal=None):
    # Step 1: Get behavioral analysis
    behavior = b_act_agent.analyze({
        "text": user_input,
        "voice": voice_signal,
    })
    
    # Step 2: Adjust system prompt based on behavior
    system_prompt = f"""You are a helpful AI assistant.
The user's emotional state is {behavior.emotional_state}.
Respond with appropriate {behavior.tone_suggestion} tone.
Confidence level: {behavior.confidence_adjustment}x"""
    
    # Step 3: Generate response with behavioral context
    response = client.messages.create(
        model="claude-3.5-sonnet",
        max_tokens=1024,
        system=system_prompt,
        messages=[
            {"role": "user", "content": user_input}
        ]
    )
    
    return response.content[0].text

# Usage
result = process_with_behavioral_intelligence(
    "Can you help me with this task?"
)

API Reference

POST

/v1/analyze

Analyze multimodal signals and return behavioral intelligence.

request

POST /v1/analyze
Content-Type: application/json
Authorization: Bearer YOUR_API_KEY

{
  "agent_id": "string",
  "signals": {
    "voice": { ... },
    "visual": { ... },
    "semantic": { ... }
  }
}

GET

/v1/agents

List all behavioral intelligence agents.

request

GET /v1/agents
Authorization: Bearer YOUR_API_KEY

Returns array of agent objects with status and signal quality metrics.

Supported AI Providers

B_Act Labs integrates seamlessly with leading AI platforms:

Anthropic

Claude 3.5 Sonnet

Recommended

OpenAI

GPT-4o

Supported

Robotics / Humanoids

Custom Models

Supported

Open Source

Llama, Mistral

Supported

Best Practices

→Use signal batching to reduce API calls during high-frequency interactions
→Configure confidence thresholds to filter low-quality behavioral signals
→Cache emotional state analysis for repeated user patterns
→Monitor signal quality metrics to maintain behavioral intelligence accuracy
→Use webhook callbacks for real-time behavioral state updates