agi_genius

The Mind (Δ) — Truth & Reasoning Engine

Checks if the response is truthful and clear. Enforces F2 (Truth), F4 (Clarity), and F7 (Humility).

Parameters

Name	Type	Default	Description
action	string	"sense"	Action to perform
query	string	""	The query to analyze
session_id	string	""	Session identifier
thought	string	""	Previous reasoning to build on

Actions

Action	Description
sense	Analyze input, detect intent, extract signals
think	Apply logical reasoning, check truth score
atlas	Route through ATLAS-333 lanes
forge	Generate reasoned output with citations
full	Complete pipeline (sense→think→atlas→forge)

Floors Enforced

Floor	Threshold	Check
F2 Truth	≥ 0.99	Is this factually accurate?
F4 Clarity	ΔS ≥ 0	Does this reduce confusion?
F7 Humility	[0.03, 0.05]	Is uncertainty stated?

Returns

{
  "action": "full",
  "query": "What is the capital of France?",
  "truth_score": 0.99,
  "clarity_delta": 0.15,
  "lane": "FACTUAL",
  "reasoning": "Verified geographic fact with high confidence.",
  "floor_results": {
    "F2": {"passed": true, "score": 0.99},
    "F4": {"passed": true, "score": 0.15},
    "F7": {"passed": true, "score": 0.04}
  }
}

Example Usage

Python

from arifos.mcp.tools.mcp_trinity import mcp_agi_genius

result = await mcp_agi_genius(
    action="full",
    query="What is the time complexity of quicksort?",
    session_id="abc123"
)

print(f"Truth Score: {result['truth_score']}")
print(f"Clarity Delta: {result['clarity_delta']}")

MCP Call

{
  "method": "tools/call",
  "params": {
    "name": "agi_genius",
    "arguments": {
      "action": "full",
      "query": "What is the time complexity of quicksort?",
      "session_id": "abc123"
    }
  }
}

Pipeline Flow

flowchart LR
    SENSE[SENSE<br/>Extract signals] --> THINK[THINK<br/>Apply reasoning]
    THINK --> ATLAS[ATLAS<br/>Route to lane]
    ATLAS --> FORGE[FORGE<br/>Generate output]