SDS-040: AI Agent Configuration Service

Addresses Requirements

[!NOTE] Evolution Path: This service implements ADR-009’s Prompt Management DSL approach. As SEA™ evolves toward the Cognitive Supply Chain ADR-026, agent configuration will integrate with versioned prompt SOPs and supply chain governance. This service remains the runtime configuration layer; ADR-026 extends the lifecycle and versioning model.

MVP Status

MVP


Component Description

The AI Agent Configuration Service manages the definitions and configurations of all AI agents within the SEA™ Intelligence Layer. It uses a Prompt Management DSL to define prompt templates, context sources, skills, and behavior adapters, enabling dynamic and modular AI agent behavior.


Technical Details

Interfaces

Direction Description Format
Input AI agent configuration JSON/YAML (Prompt Management DSL)
Output Configured AI agent instances Runtime configuration
Output Agent metadata JSON

Protocols


Data Model

Refer to SDS-001: Data Model Schemas for:


API Specification

Endpoints

Endpoint Method Description
/ai-agents/configure POST Create/update agent configuration
/ai-agents/{id}/config GET Retrieve agent configuration
/ai-agents GET List all agents
/ai-agents/{id}/deploy POST Deploy agent with configuration

Request/Response

Configure Agent

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
POST /ai-agents/configure
{
  "agentId": "artifact-recommender",
  "name": "Artifact Recommendation Agent",
  "role": "ArtifactRecommender",
  "baseModel": "GPT-4",
  "promptTemplate": "You are an expert at recommending cognitive artifacts. Based on the following context: {{context}}, suggest the most appropriate artifact type.",
  "contextSources": [
    {
      "sourceType": "knowledgeGraph",
      "query": "MATCH (e:Entity) WHERE e.domain = $domain RETURN e",
      "injectionPoint": "{{entities}}"
    },
    {
      "sourceType": "userConversation",
      "query": "last_5_messages",
      "injectionPoint": "{{conversation}}"
    }
  ],
  "skills": [
    {
      "skillId": "generate-checklist",
      "description": "Generate a checklist artifact",
      "toolCallDefinition": {...}
    }
  ],
  "behaviorAdapters": [
    {
      "adapterType": "LoRA",
      "configuration": {
        "modelPath": "adapters/artifact-specialist",
        "rank": 16
      }
    }
  ]
}

// Response: 201 Created
{
  "agentId": "artifact-recommender",
  "status": "configured",
  "deploymentUrl": "/ai-agents/artifact-recommender/invoke"
}

Error Codes

Code Description
400 Bad Request Invalid configuration syntax
404 Not Found Agent not found
409 Conflict Configuration conflict

Dependencies


Error Handling


Performance Considerations


Security Considerations


Testing Strategy