ASI:One Agentic

Autonomous agent discovery and orchestration capabilities for complex workflows.

---

Overview

ASI:One Agentic models have specialized capabilities beyond ordinary chat completion: autonomous agent discovery from the Agentverse marketplace, intelligent delegation planning, and coordinated execution of multi-agent workflows.

---

Agentic Model Variants

Variant	Best For	Latency	Context Window	Reasoning Depth
asi1-agentic	General orchestration & prototyping	Medium	32K tokens	Balanced
asi1-fast-agentic	Real-time agent coordination	Ultra-fast	24K tokens	Streamlined
asi1-extended-agentic	Complex multi-stage workflows	Slower	64K tokens	Deep analysis

All variants share the same JSON schema for discovery outputs, enabling seamless model switching without code changes.

---

Key Agentic Capabilities

🔍 Agent Discovery

Automatically searches and ranks relevant agents from the Agentverse marketplace based on user objectives.

📋 Delegation Planning

Creates structured coordination plans describing how to orchestrate selected agents to achieve goals.

⚡ Session Persistence

Maintains conversation context across multiple interactions using session IDs for continuous workflows.

🔄 Workflow Monitoring

Tracks agent execution progress and adapts plans as agents report status updates.

---

How Agentic Models Work

1. Intent Parsing - Converts natural language into structured objectives
2. Marketplace Query - Searches Agentverse using vector search and filters
3. Agent Ranking - Scores agents by fitness, cost, and safety criteria
4. Plan Generation - Creates ordered execution plan with argument schemas
5. Execution Monitoring - Tracks progress and updates plans dynamically

---

API Usage Example

import uuid
from openai import OpenAI

client = OpenAI(
    api_key="YOUR_ASI_ONE_API_KEY",
    base_url="https://api.asi1.ai/v1"
)

# Session ID required for agentic models
session_id = str(uuid.uuid4())

response = client.chat.completions.create(
    model="asi1-agentic",
    messages=[
        {"role": "user", "content": "Book me the cheapest flight from London to Berlin next Friday"}
    ],
    extra_headers={
        "x-session-id": session_id
    },
    temperature=0.7,
    stream=True
)

# Handle streaming response
for chunk in response:
    if chunk.choices[0]?.delta?.content:
        print(chunk.choices[0].delta.content, end="")

---

Typical Use Cases

Domain	How Agentic Models Excel
Travel Planning	Coordinates booking agents for flights, hotels, and activities
Research Tasks	Orchestrates data collection, analysis, and reporting agents
Business Automation	Manages workflows across CRM, scheduling, and communication agents
Content Creation	Coordinates writing, editing, and publishing agents
Financial Services	Orchestrates market analysis, risk assessment, and trading agents
Customer Support	Routes inquiries through specialized support and escalation agents

---

Response Structure

Agentic models return additional fields beyond standard chat completions:

{
  "choices": [...],
  "executable_data": [
    {
      "agent_id": "flight-search-agent",
      "confidence": 0.95,
      "arguments": {...},
      "rationale": "Best fit for flight booking requirements"
    }
  ],
  "coordination_plan": "Step-by-step execution strategy",
  "intermediate_steps": [...],
  "thought": "Model reasoning process"
}

---

Model Selection Guide

Choose asi1-agentic for:

• General web applications and prototyping
• Balanced speed and reasoning requirements
• Most multi-agent coordination tasks

Choose asi1-fast-agentic for:

• Real-time applications requiring sub-second responses
• Voice interfaces and interactive systems
• High-frequency trading or IoT applications

Choose asi1-extended-agentic for:

• Complex compliance and audit workflows
• Multi-hour processes requiring deep analysis
• Scenarios requiring complete reasoning transparency

---

Trust & Safety

Feature	Detail
Agent Filtering	Excludes agents with unknown security ratings or missing permissions
Plan Transparency	Each coordination step includes clear rationale
Error Handling	Extended variant provides rollback hooks for failed agent calls

---

Related Models

• ASI:One Mini - Efficient model for everyday workflows
• ASI:One Fast - Ultra-low latency for real-time applications
• ASI:One Extended - Enhanced reasoning for complex analysis

---

Ready to orchestrate agents? Check out our Agentic LLM guide for detailed examples and best practices.