Agent Planner Fallback

Fallback protocol when a specific tool cannot resolve intent. This capability represents a core functional node within the pConAI engine, designed to be executed efficiently via natural language or programmatic MCP invocation.

Technical Overview

The Agent Planner Fallback module is engineered to handle high-level architectural intents. Instead of manually interacting with the CAD environment, external agents can utilize this capability to perform complex calculations and scene mutations autonomously.

• Deterministic Execution: Follows strict routing rules to ensure predictable outcomes.
• Parametric Awareness: Understands OFML data schemas and reacts to spatial constraints.
• Non-Destructive Operations: Unless explicitly marked, operations can be undone or iterated upon conversationally.

MCP Integration Details

To invoke this capability via the Model Context Protocol, your AI agent should construct an intent payload targeting the specific tool endpoint.

{
  "intentOperation": "agent_planner_fallback",
  "target": "scene_graph",
  "parameters": {
    "context": "auto_resolve",
    "dry_run": false
  }
}

[!IMPORTANT] Ensure that all required parameters are resolved before invoking this capability. If the agent detects missing context, it will automatically fallback to a clarification prompt.

Natural Language Examples

When interacting with the pConAI conversational interface, the intent router will automatically map the following natural language structures to this capability:

• "Apply the agent planner fallback action to the current selection."
• "Execute the agent planner fallback macro on the east wall."
• "Run the standard workflow for ensures continuous operation gracefully."

Verification & Fallback

Before finalizing any scene mutations, the system triggers internal read-backs. If the operation results in a geometric collision or constraint violation, the operation will be aborted, and a detailed telemetry report is generated for debugging.