Skip to main content

​
What is Context?

Context is everything for LoopOS AI. Context provides agents with the information they need to:
  • Understand your business and use cases
  • Make informed decisions
  • Generate relevant responses
  • Validate data against your catalog
All context must be managed and provided by LoopOS applications, particularly LoopOS Core. The AI server does not maintain business context independently.

​
Context Types

​
Business Context (loopos_core_context)

High-level description of your business/platform: 
{
  "loopos_core_context": "Circular economy marketplace for consumer goods in Portugal. Focus on electronics, furniture, and clothing. Active marketplace with pricing data and trends."
}
Best practices:
  • Be specific about your business model
  • Include geographic focus if relevant
  • Mention key categories or product types
  • Describe your marketplace characteristics

​
Item Context

Information about the current item being processed: 
{
  "agent_specific": {
    "loopos_item": {
      "item": {
        "data": {
          "id": 2069,
          "product": {
            "name": "iPhone 12",
            "brand": "Apple"
          },
          "categories": ["Electronics", "Smartphones"]
        },
        "protocol_answers": {
          "last_answers": [
            {"question": "Condition", "answer": "Good"},
            {"question": "Description", "answer": "Minor scratches"}
          ]
        }
      }
    }
  }
}

​
Session Context

Conversation history and state maintained across requests: 
{
  "session_id": "session-123",
  "messages": [
    {"role": "user", "text": "I want to sell an iPhone"},
    {"role": "assistant", "text": "Great! Let me help you..."}
  ],
  "session_extra_data": {
    "core_id": "core123",
    "script_id": "script456",
    "item_id": "item789"
  }
}

​
Agent-Specific Context

Service-specific parameters and configuration: 
{
  "agent_specific": {
    "category_name": "Electronics",
    "product_name": "iPhone 12 Pro",
    "brand_name": "Apple",
    "user_description": "iPhone em excelente estado"
  }
}

​
Context Flow

Context flows through the system: 
Request
  ↓
Service Context (initialized from request)
  ↓
Agent Context (passed to agent)
  ↓
Tools (context used in tool calls)
  ↓
Response (context included in output)

​
Providing Context

​
In API Requests

Include context in your API requests: 
{
  "language": "PT-PT",
  "loopos_core_context": "Your business description here",
  "agent_specific": {
    "category_name": "Electronics",
    "product_name": "iPhone 12"
  },
  "session_extra_data": {
    "core_id": "core123",
    "item_id": "item789"
  }
}

​
Context from LoopOS Core

LoopOS Core provides context through:
  • Settings: Business context stored in LoopOS Core settings
  • Item Data: Current item information
  • Catalog Data: Products, brands, categories
  • Protocol Data: Protocol questions and answers

​
Context Best Practices

Be specific: Provide detailed, specific context rather than generic descriptions.
Include relevant data: Only include context that’s relevant to the current operation.
Maintain consistency: Use consistent context across related requests in a session.
Update context: Update context as items progress through workflows.

​
Context Examples

​
Example 1: Value Estimation

{
  "loopos_core_context": "Portuguese circular economy marketplace, active used goods market with pricing data and trends",
  "agent_specific": {
    "loopos_item": {
      "item": {
        "data": {
          "product": {
            "name": "iPhone 12",
            "brand": "Apple"
          },
          "categories": ["Electronics", "Smartphones"]
        },
        "protocol_answers": {
          "last_answers": [
            {"question": "Condition", "answer": "Grade B - Good condition"}
          ]
        }
      }
    }
  }
}

​
Example 2: C2C Descriptor

{
  "loopos_core_context": "Portuguese marketplace for second-hand items",
  "agent_specific": {
    "category_name": "Electronics",
    "product_name": "iPhone 12 Pro",
    "brand_name": "Apple",
    "user_description": "iPhone em excelente estado, pouco uso"
  },
  "images": ["https://example.com/photo.jpg"]
}

​
Example 3: ID Validation

{
  "loopos_core_context": "Portuguese circular economy platform, ID validation for marketplace",
  "agent_specific": {
    "doc_type": "passport,id-card",
    "name": "JOHN DOE",
    "number": "123456789",
    "validity": true
  },
  "images": ["https://example.com/id-front.jpg", "https://example.com/id-back.jpg"]
}

​
Context Limitations

Context size is limited by:
  • Request size limits (default: 10MB)
  • Token limits for LLM context windows
  • Database storage for session context
Keep context focused and relevant to avoid hitting limits.

​
Context in Multi-Agent Workflows

In multi-agent workflows, context is passed between agents:
  1. Initial Context: Set in first request
  2. Context Updates: Tools update context (e.g., create_item adds item_token)
  3. Handoff Context: Context passed to next agent via handoff data
  4. Final Context: Final context included in response