Verdict Rendering¶

Every Verdict has a render() method that formats the result for different audiences.

verdict.render()            # plain text — default
verdict.render("text")      # same as above
verdict.render("markdown")  # structured markdown
verdict.render("json")      # full JSON (all fields)

Text¶

A single readable paragraph — good for logs, CLI output, or passing to another LLM.

print(verdict.render("text"))

Proposal by agent 'a2' won with score 1.125. Highest weighted confidence score.
Dissenting proposals: Proposal p1 (agent a1) scored 0.850; Proposal p3 (agent a3) scored 0.600.

Markdown¶

Structured output with headers and a score table. Good for Slack messages, GitHub comments, audit reports, or any Markdown-aware surface.

print(verdict.render("markdown"))

## Decision Summary
Proposal by agent 'a2' won with score 1.125

**Strategy:** WeightedVote  **Status:** resolved

### Rationale
Highest weighted confidence score

### Score Breakdown
| Proposal | Score |
|---|---|
| p2 | 1.125 |
| p1 | 0.850 |
| p3 | 0.600 |

### Dissenting Proposals
- Proposal p1 (agent a1) scored 0.850
- Proposal p3 (agent a3) scored 0.600

JSON¶

Full machine-readable output with all fields. Use this for downstream processing, storage, or API responses.

import json
data = json.loads(verdict.render("json"))

{
  "id": "3a2b1c...",
  "decision_id": "9f8e7d...",
  "winner_proposal_id": "p2",
  "strategy_name": "WeightedVote",
  "status": "resolved",
  "created_at": "2025-01-15T10:23:45.123456+00:00",
  "explanation": {
    "summary": "Proposal by agent 'a2' won with score 1.125",
    "rationale": "Highest weighted confidence score",
    "dissents": [
      "Proposal p1 (agent a1) scored 0.850",
      "Proposal p3 (agent a3) scored 0.600"
    ],
    "score_breakdown": {
      "p1": 0.850,
      "p2": 1.125,
      "p3": 0.600
    }
  },
  "policy_result": {
    "allowed": true,
    "reason": "",
    "matched_rule": null
  }
}

Verdict status values¶

Status	Meaning
`resolved`	A winner was selected
`pending_human`	`DeferToHuman` — waiting for human input
`policy_blocked`	A policy rule blocked the decision

Accessing fields directly¶

You can also read fields directly from the verdict object without rendering:

verdict.winner_proposal_id     # str | None
verdict.strategy_name          # str
verdict.status                 # VerdictStatus
verdict.explanation.summary    # str
verdict.explanation.rationale  # str
verdict.explanation.score_breakdown  # dict[str, float]
verdict.policy_result.allowed  # bool
verdict.policy_result.reason   # str