RISE Specification: Agent Orchestration Participation

Desired Outcome

The DSL server operates as a concept-resolution specialist within multi-agent workflows. When an orchestrator decomposes a task, any sub-agent that encounters IAIP framework terminology can query the DSL server for authoritative definitions, relational context, and terminology validation — enabling framework-coherent outputs across the entire agent fleet.

Current Reality

• Agents in the IAIP ecosystem (mia-code, miadi-code, Heyva, Anikwag-Ayaaw, Tushell) each carry framework knowledge in system prompts
• System prompts are static snapshots; concepts evolve but prompts lag
• No shared concept authority exists across agents
• Agent outputs may drift from framework terminology without detection
• Orchestrators (LangGraph, A2A, NATS-based) have no concept-aware routing

Structural Tension

Between agents that each carry partial, potentially stale framework knowledge in isolation, and a fleet of agents that share a living, authoritative concept backbone enabling framework-coherent collaboration.

Key Features

1. Agent Registration

Agents register with the DSL server to declare their role and directional alignment:

```python class AgentRegistration(BaseModel): agent_id: str # e.g., "mia-code" agent_type: str # e.g., "development", "narrative", "ceremony" primary_direction: Direction # Which direction this agent primarily serves capabilities: list[str] # What this agent can do concept_subscriptions: list[str] # Concepts this agent cares about (for change notifications) ```

Registration is lightweight and optional — agents that don't register can still query tools.

2. Concept-Aware Routing Hints

When an orchestrator queries the DSL server about a task, the server can suggest which agents should be involved based on concept alignment:

```python

MCP tool: iaip_agent_suggest

{ "name": "iaip_agent_suggest", "description": "Given a task description, suggest which registered agents are best aligned based on concept relevance and directional alignment.", "inputSchema": { "properties": { "task_description": { "type": "string" }, "required_concepts": { "type": "array", "items": { "type": "string" } } }, "required": ["task_description"] } } ```

Response example: ```json { "suggestions": [ { "agent_id": "mia-code", "relevance": 0.9, "reason": "Task involves structural-tension and creative-orientation — Mia's primary domain", "direction": "south" }, { "agent_id": "miette-narrator", "relevance": 0.7, "reason": "Task touches eight-feelings — Miette illuminates emotional context", "direction": "east" } ] } ```

3. Coherence Validation for Agent Outputs

After an agent produces output, the orchestrator can validate framework coherence:

```python

MCP tool: iaip_validate_agent_output

{ "name": "iaip_validate_agent_output", "description": "Validate that an agent's output uses IAIP framework concepts coherently. Returns coherence score and suggestions.", "inputSchema": { "properties": { "agent_id": { "type": "string" }, "output_text": { "type": "string" }, "expected_concepts": { "type": "array", "items": { "type": "string" } } }, "required": ["output_text"] } } ```

4. Concept Change Notifications

When concepts evolve, registered agents receive notifications:

```python

Server-side notification mechanism

class ConceptChangeEvent(BaseModel): concept_id: str change_type: str # "added" | "modified" | "deprecated" old_definition: Optional[str] new_definition: str ceremony_id: Optional[str] # If change was ceremony-governed affected_agents: list[str] # Agents subscribed to this concept ```

This ensures agents don't operate with stale concept understanding.

5. Orchestration Shape Compatibility

The DSL server supports multiple orchestration patterns:

6. A2A (Agent-to-Agent) Protocol Compatibility

For agents communicating via the emerging A2A protocol (or NATS-based messaging):

```python class A2AConceptMessage(BaseModel): """Message format for concept queries over A2A/NATS""" message_type: str = "concept_query" query: dict # Same as MCP tool input reply_to: str # NATS reply subject or A2A callback correlation_id: str # For request-response correlation requesting_agent: str requesting_direction: Direction # Which direction the request comes from ```

Integration with Ceremonial Inquiry Ecosystem

The Six Thematic Suns from the Ceremonial Inquiry Ecosystem Framework map to concept clusters:

Success Metrics

• At least 3 agents successfully register and receive concept change notifications
• Coherence validation catches at least 80% of terminology inconsistencies in agent outputs
• Concept-aware routing suggestions match human judgment in ≥70% of test cases
• DSL server handles concurrent queries from 5+ agents without latency degradation
• Agent outputs measurably improve in framework coherence when DSL server is available vs. absent

Dependencies

• MCP tool layer (from rispecs-mcp-tools.md)
• Concept Registry and Relational Graph (from rispecs.md)
• Optional: NATS client library for A2A messaging
• Optional: WebSocket support for real-time notifications