mirror of
https://github.com/proffesor-for-testing/agentic-qe.git
synced 2026-09-19 08:45:47 +08:00
Dragan Spiridonov
28fa643e4e
Audited AQE routing against two router bugs ruflo fixed 2026-05-29: - #2229 stale route-cache: NOT PRESENT — Q-values read live from SQLite (qe-reasoning-bank buildQValueLookup) and in-memory qTable; no cached routing decision exists to go stale, so no invalidation needed. - #2239 state-encoder truncation: NOT PRESENT — state keys are plain string joins (agent-routing buildRoutingStateKey: taskType|priority|domain), no bit-folded quantized vector, no mask. Distinct keyword tasks map to distinct states. Adds a guard test locking in non-collapsing state encoding so a future refactor toward a bit-folding encoder fails here. 4 tests passing. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
67 lines
2.8 KiB
TypeScript
67 lines
2.8 KiB
TypeScript
/**
|
|
* Regression guard for the routing state encoder (audit of ruflo #2239 —
|
|
* "state-encoder feature truncation").
|
|
*
|
|
* ruflo's bug folded a feature vector with `(hash << 4) ^ q[i] & 0x7fffffff`,
|
|
* shifting an entire feature block past the 31-bit mask so every
|
|
* keyword-distinct task collapsed into ONE state.
|
|
*
|
|
* AQE does NOT fold a quantized feature vector. `buildRoutingStateKey`
|
|
* concatenates `taskType|priority|domain` as plain strings, and `taskType`
|
|
* is derived from description keywords by `deriveTaskType`. These tests lock
|
|
* in that distinct task semantics produce distinct state keys, so a future
|
|
* refactor toward a bit-folding encoder would fail here.
|
|
*/
|
|
|
|
import { describe, it, expect } from 'vitest';
|
|
import { deriveTaskType, buildRoutingStateKey } from '../../../src/learning/agent-routing.js';
|
|
|
|
describe('routing state encoder — no feature collapse (ruflo #2239 guard)', () => {
|
|
it('maps keyword-distinct tasks to distinct taskTypes', () => {
|
|
expect(deriveTaskType('write code for the parser')).not.toBe(
|
|
deriveTaskType('fix bug in the parser'),
|
|
);
|
|
expect(deriveTaskType('generate tests for the parser')).toBe('test-generation');
|
|
expect(deriveTaskType('fix bug in the parser')).toBe('defect-intelligence');
|
|
expect(deriveTaskType('analyze coverage gaps')).toBe('coverage-analysis');
|
|
expect(deriveTaskType('audit security vulnerabilities')).toBe('quality-assessment');
|
|
});
|
|
|
|
it('produces distinct state keys for tasks differing only in keyword content', () => {
|
|
const keys = new Set(
|
|
[
|
|
'generate tests for the parser',
|
|
'fix bug in the parser',
|
|
'analyze coverage gaps',
|
|
'validate requirements spec',
|
|
'refactor the module',
|
|
].map((task) =>
|
|
buildRoutingStateKey({
|
|
taskType: deriveTaskType(task),
|
|
priority: 'normal',
|
|
domain: 'test-generation',
|
|
}),
|
|
),
|
|
);
|
|
// 5 keyword-distinct tasks must yield 5 distinct keys — never collapse to 1.
|
|
expect(keys.size).toBe(5);
|
|
});
|
|
|
|
it('preserves the full state key (taskType + priority + domain), no truncation/mask', () => {
|
|
const key = buildRoutingStateKey({
|
|
taskType: 'security-compliance',
|
|
priority: 'high',
|
|
domain: 'security-compliance',
|
|
});
|
|
// All three segments survive — nothing is shifted past a bit mask.
|
|
expect(key).toBe('security-compliance|high|security-compliance');
|
|
expect(key.split('|')).toHaveLength(3);
|
|
});
|
|
|
|
it('distinguishes states that differ only in a later segment (domain)', () => {
|
|
const a = buildRoutingStateKey({ taskType: 'test-generation', priority: 'normal', domain: 'test-generation' });
|
|
const b = buildRoutingStateKey({ taskType: 'test-generation', priority: 'normal', domain: 'coverage-analysis' });
|
|
expect(a).not.toBe(b);
|
|
});
|
|
});
|