Consensus Engine

For high-stakes governance decisions, GovernLayer's consensus engine queries multiple LLMs independently and requires agreement before acting. Three strategies eliminate single-model hallucination risk.

Strategies

Voting — 3+ models evaluate independently; majority rules
Chain-of-Verification — Generate, question, verify, synthesize across models
Adversarial Debate — One model claims, another critiques, a third judges

POST/v1/achonye/consensus

Run consensus

Run a governance task through multi-model consensus.

Required attributes

Name
task
Type
string
Description
The governance question or decision to evaluate.
Name
strategy
Type
string
Description
Consensus strategy: voting, chain_of_verification, or adversarial_debate.

Optional attributes

Name
models
Type
array
Description
Specific models to use. Defaults to the Board (Sonnet, Gemini, GPT-4o).
Name
min_agreement
Type
float
Description
Minimum agreement threshold for voting (default: 0.67).

Request

POST

/v1/achonye/consensus

curl -X POST https://api.governlayer.ai/v1/achonye/consensus \
  -H "X-API-Key: gl_your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "task": "Should agent-12 be allowed to process refunds above $5000 without human approval?",
    "strategy": "voting"
  }'

Response

{
  "consensus_reached": true,
  "decision": "DENY",
  "confidence": 1.0,
  "strategy": "voting",
  "votes": [
    {
      "model": "claude-sonnet",
      "vote": "DENY",
      "reasoning": "Refunds above $5000 carry significant financial risk and should require human oversight per SOC 2 CC6.1."
    },
    {
      "model": "gemini-pro",
      "vote": "DENY",
      "reasoning": "High-value financial transactions without human approval violate separation of duties principles."
    },
    {
      "model": "gpt-4o",
      "vote": "DENY",
      "reasoning": "Agent autonomy for transactions exceeding $5000 creates unacceptable fraud risk."
    }
  ],
  "summary": "Unanimous denial (3/3). All models agree that high-value refunds require human-in-the-loop approval.",
  "total_tokens": 2847,
  "total_cost": "$0.038"
}