Files
msd-core/tests/feat-3024-dynamic-routing.test.cjs
Tom Boucher e1d661ece0 feat(#3024): dynamic routing with failure-tier escalation (#3031)
* feat(#3024): dynamic routing with failure-tier escalation

Adds a `dynamic_routing` block to .planning/config.json that lets
the resolver start agents on a cheap tier and escalate one tier up
when the orchestrator detects a soft failure (verification
inconclusive, plan-check FLAG, etc.). Solves the "pay Opus rates as
insurance" anti-pattern by making escalation observed-quality-driven.

Architecture:
- AGENT_DEFAULT_TIERS map (light/standard/heavy) — every agent in
  MODEL_PROFILES declares a default tier; tests assert coverage
  so adding a new agent without updating the map fails CI.
- nextTier(currentTier) helper — light → standard → heavy → heavy
  (heavy stays at heavy; can't go further).
- resolveModelForTier(cwd, agentType, attempt) — new resolver. The
  orchestrator tracks the attempt counter and passes 0 for the
  first spawn, 1+ on escalation. The resolver caps internally at
  max_escalations so the orchestrator can blindly bump the counter.
- Schema validation: dynamic_routing.enabled / escalate_on_failure /
  max_escalations / tier_models.<light|standard|heavy>. Unknown
  tiers and unknown sub-keys rejected at config-set time.
- SDK schema mirror updated to keep CJS/SDK in lockstep (#2653).

Resolution precedence (highest → lowest):
  1. model_overrides[<agent>]              (full IDs accepted)
  2. dynamic_routing.tier_models[<tier>]   (NEW; escalation-aware)
  3. models[<phase_type>]                  (#3023 phase-type map)
  4. model_profile                         (per-agent column)
  5. Runtime default

Backward compatibility: dynamic_routing is disabled by default
(enabled: false or block omitted). resolveModelForTier short-
circuits to resolveModelInternal in that case, so callers can
adopt unconditionally without breaking existing behavior.

This PR delivers the JS-layer infrastructure: schema + tier map +
resolver. Orchestrator adoption (workflow markdown updates that
detect soft failures and call resolveModelForTier with attempt+1)
is incremental follow-up — verifier / plan-checker / integration-
checker each adopt the protocol when ready.

Tests (23 cases, all structural-IR — no stdout grep):
- Schema invariants: AGENT_DEFAULT_TIERS coverage, VALID_AGENT_TIERS
  exact match, every assignment uses a valid tier
- nextTier helper: light→standard→heavy→heavy, null on invalid input
- Disabled mode: no block + enabled:false both no-op (back-compat)
- Enabled mode: attempt=0 returns default tier model, attempt=1
  escalates, beyond max_escalations caps, heavy agents stay heavy,
  default max_escalations=1 when omitted
- Precedence: per-agent override beats dynamic_routing,
  dynamic_routing beats phase-type models
- Validation: every settings key accepted, unknown tiers/sub-keys
  rejected, bare `dynamic_routing` rejected as config-set target

Documentation:
- get-shit-done/references/model-profiles.md — full reference section
- docs/CONFIGURATION.md — full settings table + escalation flow
- docs/USER-GUIDE.md — task-oriented "Cheap-by-default" section
- docs/FEATURES.md — config row cross-link

Verification:
- 23/23 pass on regression test
- 6843/6843 full suite (23 net new from 6820)
- lint-no-source-grep clean (376 test files)
- SDK schema mirror keeps CJS/SDK in sync per #2653 parity test

Closes #3024

* fix(#3024): honor escalate_on_failure:false + 3 CR follow-ups

CodeRabbit on PR #3031 (4 findings — 1 Major + 2 Minor + 1 Nitpick):

1. **Major (inline)** — get-shit-done/bin/lib/core.cjs:1668
   resolveModelForTier ignored dynamic_routing.escalate_on_failure.
   When the user set it to false, escalation should be disabled, but
   the resolver only checked attempt/max_escalations. An orchestrator
   that always passes attempt+1 on retry would silently escalate
   despite the user opting out.
   Fix: gate effectiveAttempt on `dr.escalate_on_failure !== false`
   so false short-circuits every attempt back to the default tier.

2. **Minor (inline)** — docs/CONFIGURATION.md:123-126
   The dynamic_routing rows in the Core Settings table had 4 cells
   instead of 5 (missing the Options column), breaking the table
   structure. Added explicit Options values for enabled / escalate_on_failure
   / max_escalations rows.

3. **Minor (outside-diff)** — references/model-profiles.md:179-195
   "Resolution Logic" sketch was pre-#3024 and didn't include
   dynamic_routing in the precedence ladder. Updated to a 6-step
   block with dynamic_routing at step 3 (between override and
   phase-type).

4. **Nitpick** — tests/feat-3024-dynamic-routing.test.cjs:189+
   Tests used `if (lightAgent) { ... }` guards that silent-pass
   when AGENT_DEFAULT_TIERS drifts. Replaced all 5 conditional
   skips with `assert.ok(lightAgent, '...')` preconditions so a
   tier-mapping change surfaces as a test failure.

Plus: 2 new regression tests for the Major fix:
- escalate_on_failure:false caps every attempt at default tier
- escalate_on_failure:true (explicit) still escalates normally

Verification:
- 25/25 pass on regression test (23 prior + 2 escalate_on_failure)
- 6845/6845 full suite (2 net new)
- lint-no-source-grep clean

* docs(#3024): align precedence + add fence language tags (CR follow-up)

CodeRabbit (3 minor):

1. docs/CONFIGURATION.md:691 — "Per-Phase-Type Models → Resolution
   precedence" was a 4-step block written pre-#3024; readers got
   contradictory rules between the per-phase-type section and the
   later dynamic_routing section. Updated to the same 5-step ladder
   with dynamic_routing at step 2, and noted that dynamic_routing
   is disabled by default so this section's behavior is unchanged
   when the kill-switch is off.

2. docs/CONFIGURATION.md:770 — escalation-flow code fence missing
   language tag (MD040). Added `text`.

3. references/model-profiles.md:184 — resolution-ladder code fence
   missing language tag (MD040). Added `text`.

No code changes; docs only. Verification: regression test still 25/25.

* docs(#3024): clarify precedence prose — five layers, not four (CR nitpick)

CodeRabbit nitpick: the "Per-Phase-Type Models → Resolution
precedence" prose said "The four layers compose..." but the ladder
above lists five (including Runtime default). Also "dynamic_routing
escalates per-attempt above all of them" misreads as suggesting
dynamic_routing wins over model_overrides — actually overrides still
win at step 1.

Reworded top-down so the precedence direction is unambiguous:
  - model_profile = base
  - models = phase-level override
  - dynamic_routing = per-attempt escalation
  - model_overrides = per-agent exception (top)
  - runtime default = fallback

No code changes; docs only.

* docs(#3024): note escalate_on_failure:false in escalation-flow diagram (CR)

CodeRabbit nitpick: the escalation-flow diagram in
docs/CONFIGURATION.md described the soft-failure → respawn →
tier_models[next_tier_up] path, but didn't surface the
`dynamic_routing.escalate_on_failure: false` kill-switch right next
to it. Users reading the flow diagram (which is the canonical place
to understand attempt behavior) wouldn't see that the kill-switch
overrides the soft-failure branch.

Added a one-paragraph note immediately after the flow listing,
before the tier-sequence example, so the kill-switch is visible
exactly where users decide whether escalation will happen.

No code changes; docs only.
2026-05-02 14:26:35 -04:00

367 lines
16 KiB
JavaScript

/**
* Feature test for issue #3024 — dynamic routing with failure-tier escalation.
*
* Adds a `dynamic_routing` block to .planning/config.json:
*
* {
* "dynamic_routing": {
* "enabled": true,
* "tier_models": {
* "light": "haiku",
* "standard": "sonnet",
* "heavy": "opus"
* },
* "escalate_on_failure": true,
* "max_escalations": 1
* }
* }
*
* Each agent has a default tier (light/standard/heavy). When dynamic
* routing is enabled, the resolver picks `tier_models[default_tier]`
* for the first attempt. On orchestrator-detected soft failure, the
* orchestrator calls the resolver again with `attempt: 1`, which
* returns the next tier up (capped at `max_escalations`).
*
* This PR delivers the JS-layer infrastructure: schema + tier map +
* resolver + escalation helpers. Orchestrator adoption is incremental
* follow-up — this PR's contract is the resolver function and the
* config it consumes.
*
* Resolution precedence (highest → lowest):
* 1. model_overrides[agent] (full IDs accepted; targeted)
* 2. dynamic_routing.tier_models[tier] (NEW; escalation-aware)
* 3. models[phase_type] (#3023; coarse phase-level)
* 4. model_profile (per-agent column)
* 5. Runtime default
*
* Tests are typed-IR / structural — assert on the value returned by
* resolveModelForTier or isValidConfigKey, not stdout/grep.
*/
'use strict';
process.env.GSD_TEST_MODE = '1';
const { test, describe, beforeEach, afterEach } = require('node:test');
const assert = require('node:assert/strict');
const fs = require('node:fs');
const path = require('node:path');
const os = require('node:os');
const {
resolveModelInternal,
resolveModelForTier,
} = require('../get-shit-done/bin/lib/core.cjs');
const {
AGENT_DEFAULT_TIERS,
VALID_AGENT_TIERS,
MODEL_PROFILES,
nextTier,
} = require('../get-shit-done/bin/lib/model-profiles.cjs');
const { isValidConfigKey } = require('../get-shit-done/bin/lib/config-schema.cjs');
function makeTmp(prefix) {
return fs.mkdtempSync(path.join(os.tmpdir(), `gsd-3024-${prefix}-`));
}
function writeConfig(dir, config) {
const planningDir = path.join(dir, '.planning');
fs.mkdirSync(planningDir, { recursive: true });
fs.writeFileSync(path.join(planningDir, 'config.json'), JSON.stringify(config, null, 2));
}
function rmr(p) { try { fs.rmSync(p, { recursive: true, force: true }); } catch { /* noop */ } }
// ─── Schema: AGENT_DEFAULT_TIERS coverage + valid tier set ──────────────────
describe('#3024 schema: every agent has a default tier (light/standard/heavy)', () => {
test('AGENT_DEFAULT_TIERS exported as a non-empty object', () => {
assert.equal(typeof AGENT_DEFAULT_TIERS, 'object');
assert.ok(AGENT_DEFAULT_TIERS !== null);
assert.ok(Object.keys(AGENT_DEFAULT_TIERS).length > 0);
});
test('VALID_AGENT_TIERS exposes exactly {light, standard, heavy}', () => {
assert.deepStrictEqual([...VALID_AGENT_TIERS].sort(), ['heavy', 'light', 'standard']);
});
test('every agent in MODEL_PROFILES has a default tier', () => {
const missing = Object.keys(MODEL_PROFILES).filter((a) => !AGENT_DEFAULT_TIERS[a]);
assert.deepStrictEqual(missing, []);
});
test('every assigned tier is one of the three valid tiers', () => {
const invalid = Object.entries(AGENT_DEFAULT_TIERS).filter(
([, t]) => !VALID_AGENT_TIERS.has(t)
);
assert.deepStrictEqual(invalid, []);
});
});
// ─── nextTier helper ────────────────────────────────────────────────────────
describe('#3024 nextTier helper', () => {
test('exported as a function', () => {
assert.equal(typeof nextTier, 'function');
});
test('light → standard → heavy → heavy (caps at heavy)', () => {
assert.equal(nextTier('light'), 'standard');
assert.equal(nextTier('standard'), 'heavy');
assert.equal(nextTier('heavy'), 'heavy', 'already at top — stays at heavy');
});
test('returns null for invalid input', () => {
assert.equal(nextTier('jumbo'), null);
assert.equal(nextTier(null), null);
assert.equal(nextTier(undefined), null);
});
});
// ─── Resolver behavior: dynamic routing, disabled mode ──────────────────────
describe('#3024 resolveModelForTier: disabled mode is a no-op (acceptance criterion 1)', () => {
let projectDir;
beforeEach(() => { projectDir = makeTmp('disabled'); });
afterEach(() => { rmr(projectDir); });
test('exported as a function', () => {
assert.equal(typeof resolveModelForTier, 'function');
});
test('with no dynamic_routing block, falls back to resolveModelInternal', () => {
writeConfig(projectDir, { model_profile: 'balanced' });
// resolveModelForTier with attempt=0 must match resolveModelInternal.
const baseline = resolveModelInternal(projectDir, 'gsd-phase-researcher');
assert.equal(resolveModelForTier(projectDir, 'gsd-phase-researcher', 0), baseline);
});
test('with dynamic_routing.enabled=false, attempt argument is ignored — same as resolveModelInternal', () => {
writeConfig(projectDir, {
model_profile: 'balanced',
dynamic_routing: {
enabled: false,
tier_models: { light: 'haiku', standard: 'sonnet', heavy: 'opus' },
},
});
const baseline = resolveModelInternal(projectDir, 'gsd-phase-researcher');
// attempt=0 and attempt=1 both ignored when disabled
assert.equal(resolveModelForTier(projectDir, 'gsd-phase-researcher', 0), baseline);
assert.equal(resolveModelForTier(projectDir, 'gsd-phase-researcher', 1), baseline);
});
});
// ─── Resolver behavior: dynamic routing, enabled ────────────────────────────
describe('#3024 resolveModelForTier: enabled mode picks tier_models[default_tier]', () => {
let projectDir;
beforeEach(() => { projectDir = makeTmp('enabled'); });
afterEach(() => { rmr(projectDir); });
test('attempt=0 returns tier_models[agent_default_tier] (acceptance criterion 2)', () => {
writeConfig(projectDir, {
model_profile: 'balanced',
dynamic_routing: {
enabled: true,
tier_models: { light: 'haiku', standard: 'sonnet', heavy: 'opus' },
},
});
// gsd-codebase-mapper has light default tier per AGENT_DEFAULT_TIERS.
// CR nitpick (#3031): assert preconditions explicitly so a tier
// re-mapping in AGENT_DEFAULT_TIERS surfaces as a test failure
// instead of a silent skip.
assert.equal(AGENT_DEFAULT_TIERS['gsd-codebase-mapper'], 'light',
'gsd-codebase-mapper expected to be light tier');
assert.equal(resolveModelForTier(projectDir, 'gsd-codebase-mapper', 0), 'haiku');
assert.equal(AGENT_DEFAULT_TIERS['gsd-planner'], 'heavy',
'gsd-planner expected to be heavy tier');
assert.equal(resolveModelForTier(projectDir, 'gsd-planner', 0), 'opus');
});
test('attempt=1 escalates to next tier up (acceptance criterion 3)', () => {
writeConfig(projectDir, {
model_profile: 'balanced',
dynamic_routing: {
enabled: true,
tier_models: { light: 'haiku', standard: 'sonnet', heavy: 'opus' },
escalate_on_failure: true,
max_escalations: 1,
},
});
// For an agent with default tier 'light', attempt=1 should give 'standard' tier model.
const lightAgent = Object.entries(AGENT_DEFAULT_TIERS).find(([, t]) => t === 'light')?.[0];
assert.ok(lightAgent, 'AGENT_DEFAULT_TIERS must contain at least one light agent');
assert.equal(resolveModelForTier(projectDir, lightAgent, 0), 'haiku');
assert.equal(resolveModelForTier(projectDir, lightAgent, 1), 'sonnet');
// For a 'standard' agent, attempt=1 should give 'heavy' model.
const stdAgent = Object.entries(AGENT_DEFAULT_TIERS).find(([, t]) => t === 'standard')?.[0];
assert.ok(stdAgent, 'AGENT_DEFAULT_TIERS must contain at least one standard agent');
assert.equal(resolveModelForTier(projectDir, stdAgent, 0), 'sonnet');
assert.equal(resolveModelForTier(projectDir, stdAgent, 1), 'opus');
});
test('attempts beyond max_escalations cap at the highest reachable tier (acceptance criterion 4)', () => {
writeConfig(projectDir, {
model_profile: 'balanced',
dynamic_routing: {
enabled: true,
tier_models: { light: 'haiku', standard: 'sonnet', heavy: 'opus' },
escalate_on_failure: true,
max_escalations: 1, // cap at 1 escalation total
},
});
const lightAgent = Object.entries(AGENT_DEFAULT_TIERS).find(([, t]) => t === 'light')?.[0];
assert.ok(lightAgent, 'AGENT_DEFAULT_TIERS must contain at least one light agent');
// attempts beyond max_escalations should not exceed max_escalations'
// tier — i.e. attempt=2 with max=1 = same as attempt=1.
assert.equal(resolveModelForTier(projectDir, lightAgent, 2), 'sonnet',
'attempt=2 with max_escalations=1 caps at attempt=1 tier');
assert.equal(resolveModelForTier(projectDir, lightAgent, 5), 'sonnet');
});
test('"heavy" agents stay at heavy (no tier above)', () => {
writeConfig(projectDir, {
model_profile: 'balanced',
dynamic_routing: {
enabled: true,
tier_models: { light: 'haiku', standard: 'sonnet', heavy: 'opus' },
escalate_on_failure: true,
max_escalations: 2,
},
});
const heavyAgent = Object.entries(AGENT_DEFAULT_TIERS).find(([, t]) => t === 'heavy')?.[0];
assert.ok(heavyAgent, 'AGENT_DEFAULT_TIERS must contain at least one heavy agent');
assert.equal(resolveModelForTier(projectDir, heavyAgent, 0), 'opus');
// Already at heavy — escalation cannot go higher.
assert.equal(resolveModelForTier(projectDir, heavyAgent, 1), 'opus');
assert.equal(resolveModelForTier(projectDir, heavyAgent, 5), 'opus');
});
test('default max_escalations is 1 when omitted', () => {
writeConfig(projectDir, {
model_profile: 'balanced',
dynamic_routing: {
enabled: true,
tier_models: { light: 'haiku', standard: 'sonnet', heavy: 'opus' },
// max_escalations omitted — default to 1
},
});
const lightAgent = Object.entries(AGENT_DEFAULT_TIERS).find(([, t]) => t === 'light')?.[0];
assert.ok(lightAgent, 'AGENT_DEFAULT_TIERS must contain at least one light agent');
// attempt=1 escalates; attempt=2 should cap at attempt=1 (default max=1)
assert.equal(resolveModelForTier(projectDir, lightAgent, 1), 'sonnet');
assert.equal(resolveModelForTier(projectDir, lightAgent, 2), 'sonnet');
});
// ─── CR Major (#3031): escalate_on_failure: false honored ──────────────
test('escalate_on_failure:false disables escalation even when attempt > 0 (CR Major)', () => {
// Pre-fix bug: an orchestrator that always passes attempt+1 on retry
// would silently escalate even though the user opted out via
// escalate_on_failure:false. The kill-switch must short-circuit
// every attempt back to the default tier.
writeConfig(projectDir, {
model_profile: 'balanced',
dynamic_routing: {
enabled: true,
tier_models: { light: 'haiku', standard: 'sonnet', heavy: 'opus' },
escalate_on_failure: false, // ← kill-switch
max_escalations: 5,
},
});
const lightAgent = Object.entries(AGENT_DEFAULT_TIERS).find(([, t]) => t === 'light')?.[0];
assert.ok(lightAgent, 'AGENT_DEFAULT_TIERS must contain at least one light agent');
// Every attempt must resolve to the default (light → haiku),
// regardless of how high the orchestrator bumped the counter.
assert.equal(resolveModelForTier(projectDir, lightAgent, 0), 'haiku');
assert.equal(resolveModelForTier(projectDir, lightAgent, 1), 'haiku',
'escalate_on_failure:false must not escalate even at attempt=1');
assert.equal(resolveModelForTier(projectDir, lightAgent, 5), 'haiku');
});
test('escalate_on_failure:true (explicit) escalates normally', () => {
// Sanity: explicit true matches the default truthy behavior.
writeConfig(projectDir, {
model_profile: 'balanced',
dynamic_routing: {
enabled: true,
tier_models: { light: 'haiku', standard: 'sonnet', heavy: 'opus' },
escalate_on_failure: true,
max_escalations: 1,
},
});
const lightAgent = Object.entries(AGENT_DEFAULT_TIERS).find(([, t]) => t === 'light')?.[0];
assert.ok(lightAgent);
assert.equal(resolveModelForTier(projectDir, lightAgent, 1), 'sonnet');
});
});
// ─── Resolver precedence ────────────────────────────────────────────────────
describe('#3024 precedence: per-agent override > dynamic_routing > models > profile', () => {
let projectDir;
beforeEach(() => { projectDir = makeTmp('precedence'); });
afterEach(() => { rmr(projectDir); });
test('per-agent model_overrides beats dynamic_routing (acceptance criterion: override wins)', () => {
writeConfig(projectDir, {
model_profile: 'balanced',
model_overrides: { 'gsd-codebase-mapper': 'openai/gpt-5' },
dynamic_routing: {
enabled: true,
tier_models: { light: 'haiku', standard: 'sonnet', heavy: 'opus' },
},
});
// Per-agent override always wins, even at escalated attempt.
assert.equal(resolveModelForTier(projectDir, 'gsd-codebase-mapper', 0), 'openai/gpt-5');
assert.equal(resolveModelForTier(projectDir, 'gsd-codebase-mapper', 1), 'openai/gpt-5');
});
test('dynamic_routing beats phase-type models (#3023)', () => {
writeConfig(projectDir, {
model_profile: 'balanced',
models: { research: 'opus' }, // phase-type would say opus
dynamic_routing: {
enabled: true,
tier_models: { light: 'haiku', standard: 'sonnet', heavy: 'opus' },
},
});
// gsd-codebase-mapper is research phase-type; phase-type would give 'opus',
// but dynamic routing (light default → haiku) wins.
if (AGENT_DEFAULT_TIERS['gsd-codebase-mapper'] === 'light') {
assert.equal(resolveModelForTier(projectDir, 'gsd-codebase-mapper', 0), 'haiku');
}
});
});
// ─── Schema validation ──────────────────────────────────────────────────────
describe('#3024 config-schema: dynamic_routing.* validation', () => {
test('dynamic_routing.enabled is a valid config key', () => {
assert.equal(isValidConfigKey('dynamic_routing.enabled'), true);
});
test('dynamic_routing.escalate_on_failure is a valid config key', () => {
assert.equal(isValidConfigKey('dynamic_routing.escalate_on_failure'), true);
});
test('dynamic_routing.max_escalations is a valid config key', () => {
assert.equal(isValidConfigKey('dynamic_routing.max_escalations'), true);
});
test('dynamic_routing.tier_models.<tier> for each valid tier', () => {
for (const t of ['light', 'standard', 'heavy']) {
assert.equal(isValidConfigKey(`dynamic_routing.tier_models.${t}`), true);
}
});
test('unknown tier in tier_models is rejected', () => {
assert.equal(isValidConfigKey('dynamic_routing.tier_models.jumbo'), false);
assert.equal(isValidConfigKey('dynamic_routing.tier_models.medium'), false);
});
test('unknown dynamic_routing.* keys are rejected', () => {
assert.equal(isValidConfigKey('dynamic_routing.foo'), false);
assert.equal(isValidConfigKey('dynamic_routing'), false,
'bare dynamic_routing (no field) must not be a config-set target');
});
});