Garry's Opinionated OpenClaw/Hermes Agent Brain
TypeScript
/**
* Retrieval feedback operations: `rate_answer` (CLI `gbrain rate`).
* Never import from '../operations.ts' here (cycle).
*/
import type { Operation } from './contract.ts';
import { rateAnswer } from '../feedback/rate.ts';
const rate_answer: Operation = {
name: 'rate_answer',
area: 'search',
outputRedaction: 'no_stored_text',
mutating: true,
writeInference: 'none',
idempotent: true,
description:
'Rate how useful an answer\'s retrieved evidence was, so this brain ranks better next time (zero LLM calls). ' +
'Pass the answer_id from a query/search/think/synthesize/recall response and rating 1-5, or rate single pages with ' +
'pages: [{ ref: "source_id:slug", rating }]. Example: { answer_id: "ans_01J…", pages: [{ ref: "default:people/alice-example", rating: 1 }] }. ' +
'Returns a receipt with each weight before/after; it never edits stored facts.',
scope: 'write',
params: {
answer_id: { type: 'string', required: true, description: 'The ans_… id from the answer you are rating.' },
rating: { type: 'number', description: 'Whole-answer rating 1-5 (omit when using pages).' },
pages: {
type: 'array',
description: 'Per-page ratings: [{ ref: "source_id:slug", rating: 1-5 }]. Pages only; edges are not touched.',
items: { type: 'object' },
},
},
handler: async (ctx, p) => rateAnswer(ctx, { answer_id: p.answer_id, rating: p.rating, pages: p.pages }),
cliHints: { name: 'rate', positional: ['answer_id', 'rating'] },
};
export const feedbackOperations: Operation[] = [rate_answer];