@metaharness/darwin 0.2.1 → 0.2.3
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +32 -0
- package/LEARNINGS.md +63 -0
- package/README.md +14 -9
- package/dist/cli.js +10 -1
- package/dist/cli.js.map +1 -1
- package/dist/index.d.ts +1 -0
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +1 -0
- package/dist/index.js.map +1 -1
- package/dist/ruvllm-mutator.d.ts +32 -0
- package/dist/ruvllm-mutator.d.ts.map +1 -0
- package/dist/ruvllm-mutator.js +73 -0
- package/dist/ruvllm-mutator.js.map +1 -0
- package/package.json +5 -3
package/CHANGELOG.md
ADDED
|
@@ -0,0 +1,32 @@
|
|
|
1
|
+
# Changelog — @metaharness/darwin
|
|
2
|
+
|
|
3
|
+
All notable changes to this package. Dates UTC.
|
|
4
|
+
|
|
5
|
+
## 0.2.3 — 2026-06-19
|
|
6
|
+
|
|
7
|
+
- Add `LEARNINGS.md` — the measured findings distilled into harness defaults (repair=2x lever, cost-routing, Barbarian&Scholar tiering, format-contract, batch-eval discipline, capability floor).
|
|
8
|
+
|
|
9
|
+
## 0.2.2 — 2026-06-19
|
|
10
|
+
|
|
11
|
+
- **Docs: full SWE-bench Lite (300) evidence ladder** now in the description + README, all official `swebench` Docker harness, batch-verified:
|
|
12
|
+
- open-loop **7.7%** [5.2, 11.2] (ADR-144)
|
|
13
|
+
- + localization **8.0%** [5.4, 11.6] (ADR-146)
|
|
14
|
+
- + closed-loop repair **15.3%** [11.7, 19.8] (ADR-149)
|
|
15
|
+
- + Barbarian&Scholar hybrid (cheap base + frontier-tail escalation) **33.3%** [28.2, 38.8] (ADR-148)
|
|
16
|
+
- Blended cost ~$0.01/instance (cheap) → ~$0.34/instance (hybrid) vs $1–20/instance for frontier agents.
|
|
17
|
+
- README now links `bench/results/RESULTS.md` (the full reproducible evidence) for npm-only readers.
|
|
18
|
+
- `RuvllmMutator` (local/$0 air-gapped mutator, ADR-259) ships in `dist/`.
|
|
19
|
+
- Added `bench/swebench/KNOWN_FLAKY.md` (standing `psf__requests-2317` Docker-hang exclusion note).
|
|
20
|
+
|
|
21
|
+
## 0.2.1 — 2026-06-19
|
|
22
|
+
|
|
23
|
+
- Metadata: repositioned as "an LLM supercharger and cost optimizer"; keywords/description.
|
|
24
|
+
|
|
25
|
+
## 0.2.0 — 2026-06-18
|
|
26
|
+
|
|
27
|
+
- Integrated into the `metaharness` scaffolder (`npm run evolve`, ADR-147).
|
|
28
|
+
- Evolutionary stack (mutation + crossover + diverse selection + graded promotion) over a frozen core.
|
|
29
|
+
|
|
30
|
+
## 0.1.x — 2026-06
|
|
31
|
+
|
|
32
|
+
- Initial release: frozen-model / evolving-harness over 7 mutation surfaces; deterministic mutator default; `validateGeneratedCode` safety gate; pluggable `CodeGenerator`.
|
package/LEARNINGS.md
ADDED
|
@@ -0,0 +1,63 @@
|
|
|
1
|
+
# What the benchmarks taught us → harness defaults
|
|
2
|
+
|
|
3
|
+
Empirical findings from the full SWE-bench Lite (300) arc (official `swebench` Docker harness,
|
|
4
|
+
batch-verified — see `bench/results/RESULTS.md`). These are the *measured* reasons behind the
|
|
5
|
+
recommended harness patterns. The headline: **the harness, not the model, is the dominant lever.**
|
|
6
|
+
|
|
7
|
+
## 1. Closed-loop repair (test feedback) is the #1 lever — ~2× for free
|
|
8
|
+
- open-loop single-shot: **7.7%** → + closed-loop repair (run the failing tests, feed the
|
|
9
|
+
traceback back, retry ≤3): **15.3%** — on the *same cheap model*, ~$0.01/instance.
|
|
10
|
+
- **Recommendation:** make iteration against ground-truth (compiler/tests) first-class. A model
|
|
11
|
+
that can *see why it failed* beats a smarter model that can't. Prefer `retryPolicy` configs that
|
|
12
|
+
consume real failure signal over blind retries.
|
|
13
|
+
|
|
14
|
+
## 2. Localization fixes retrieval, not emission — beware the "emission wall"
|
|
15
|
+
- LLM file-localization lifted gold-file recall **+15pp** but resolve-rate stayed flat (8.0%).
|
|
16
|
+
The bottleneck was *writing a valid patch*, not *finding the file*.
|
|
17
|
+
- **Recommendation:** measure where you actually lose (selection vs emission) before optimizing
|
|
18
|
+
retrieval. Don't assume better context = better output.
|
|
19
|
+
|
|
20
|
+
## 3. Format contract + fit-in-context unblocks weak/local models (0 → 13/25 applied)
|
|
21
|
+
- A small local model emitted prose summaries instead of edits until the harness (a) served enough
|
|
22
|
+
context window, (b) carried the search/replace **format contract in a system message + worked
|
|
23
|
+
example**, and (c) **shrank per-file context to fit the window** (truncation silently dropped the
|
|
24
|
+
instruction). Apply-rate went 0 → ~50%.
|
|
25
|
+
- **Recommendation:** put the output-format contract in a *system* role with an example; size the
|
|
26
|
+
prompt to the model's real context; never let truncation eat the instruction.
|
|
27
|
+
|
|
28
|
+
## 4. Cheap-first + cost-aware routing — 31× cheaper per resolve
|
|
29
|
+
- Router probe: `pareto-code`→deepseek-v4-pro resolved at **$0.21/resolve** vs `fusion`→opus-4.8 at
|
|
30
|
+
**$6.57/resolve** — same task, 31× cost gap for +1 resolve.
|
|
31
|
+
- **Recommendation:** default to the cheapest model that clears the task; reserve frontier models for
|
|
32
|
+
measured capability gaps. Track **$/resolve**, not just resolve-rate.
|
|
33
|
+
|
|
34
|
+
## 5. Barbarian & Scholar — tier the models, escalate only the residual (33.3% at ~6× less cost)
|
|
35
|
+
- Cheap base (deepseek + repair) banks the easy 46/300; a frontier "Scholar" (sonnet-4 + repair)
|
|
36
|
+
escalated **only to the 254 it failed** cracks 55 more → **100/300 = 33.3%**, blended
|
|
37
|
+
~$0.34/instance vs ~$2 to run frontier on all 300.
|
|
38
|
+
- **Recommendation:** two-tier orchestration — cheap sweep, then frontier on the residual — is far
|
|
39
|
+
more cost-efficient than one strong model everywhere (you'd waste 5/6 of frontier spend re-solving
|
|
40
|
+
what cheap already gets).
|
|
41
|
+
|
|
42
|
+
## 6. The repair lift is model-bound below a capability floor (~14B)
|
|
43
|
+
- Repair did nothing for a 7B (4%→4%) but lifted a 14B (8%→12%) and doubled a hosted model. The loop
|
|
44
|
+
needs the model to *occasionally* produce a correct-ish patch to converge toward.
|
|
45
|
+
- **Recommendation:** don't expect harness scaffolding to rescue a model below the task's reasoning
|
|
46
|
+
floor; pick the smallest model *above* it, then let the harness multiply it.
|
|
47
|
+
|
|
48
|
+
## 7. Methodology: only batch-eval on final predictions is authoritative
|
|
49
|
+
- In-loop "resolved" counters drifted from clean batch eval by 1.5–5× (both directions — flaky
|
|
50
|
+
passes over-count; Docker-hang false-negatives under-count). Every reported number here is a
|
|
51
|
+
fresh batch eval on the final saved predictions.
|
|
52
|
+
- **Recommendation:** never report the in-loop signal; re-evaluate the artifact you'd actually ship.
|
|
53
|
+
|
|
54
|
+
## 8. Engineering robustness (or your run lies to you)
|
|
55
|
+
- Concurrency clones rate-limit (6-wide anon GitHub clones → 63 fetch failures): **cap at 2–3**,
|
|
56
|
+
retry-with-backoff, free each clone. One instance (`psf__requests-2317`) reliably hangs Docker
|
|
57
|
+
past timeout → known-flaky exclusion (`bench/swebench/KNOWN_FLAKY.md`). Watch for wedged containers.
|
|
58
|
+
|
|
59
|
+
---
|
|
60
|
+
|
|
61
|
+
Verdict: this paradigm (localize + search/replace + repair + tiered escalation) tops out ~33% on
|
|
62
|
+
SWE-bench Lite with a cheap base. The 65–88% agentic-SOTA tier needs a **multi-step autonomous agent**
|
|
63
|
+
(bash, dir-navigation, long-horizon discovery) — an architecture change, not more knob-tuning.
|
package/README.md
CHANGED
|
@@ -271,7 +271,9 @@ substrate is **no longer deferred** — it shipped (ADR-106→141) and now runs
|
|
|
271
271
|
- **Polyglot model frontier** (ADR-085): 15 models × 6 languages, execution-scored;
|
|
272
272
|
DeepSeek-V3 ($0.4/Mtok) tops quality-per-dollar — cheap beats frontier for code.
|
|
273
273
|
|
|
274
|
-
### Canonical SWE-bench Lite (real, official harness — ADR-142–
|
|
274
|
+
### Canonical SWE-bench Lite (real, official harness — ADR-142–149)
|
|
275
|
+
|
|
276
|
+
> Full reproducible evidence: [`bench/results/RESULTS.md`](bench/results/RESULTS.md) · measured best-practices: [`LEARNINGS.md`](LEARNINGS.md) · known-flaky exclusions: [`bench/swebench/KNOWN_FLAKY.md`](bench/swebench/KNOWN_FLAKY.md)
|
|
275
277
|
|
|
276
278
|
Run on the **full 300** SWE-bench Lite (test) instances, scored by the **official
|
|
277
279
|
`swebench` Docker harness** — no cherry-picking, tight CIs. Solver = relevance-ranked
|
|
@@ -281,14 +283,17 @@ context + symbol-aware localization + search/replace patch, `deepseek-chat`, ~$0
|
|
|
281
283
|
|---|---|---|---|
|
|
282
284
|
| baseline (open-loop, single-shot) | 23/300 = **7.7%** | [5.2, 11.2] | 144 |
|
|
283
285
|
| + LLM localization | 24/300 = **8.0%** | [5.4, 11.6] | 146 |
|
|
284
|
-
|
|
|
285
|
-
|
|
286
|
-
|
|
287
|
-
|
|
288
|
-
|
|
289
|
-
|
|
290
|
-
|
|
291
|
-
|
|
286
|
+
| **+ closed-loop repair (test-feedback, ≤3)** | 46/300 = **15.3%** | **[11.7, 19.8]** | 149 |
|
|
287
|
+
| **+ Barbarian&Scholar hybrid (cheap+frontier tail)** | 100/300 = **33.3%** | **[28.2, 38.8]** | 148 |
|
|
288
|
+
|
|
289
|
+
The closed-loop repair loop **~doubles** the resolve-rate (7.7% → 15.3%) on the *same cheap model*
|
|
290
|
+
(deepseek, <1¢/instance); the baseline and repair CIs are essentially disjoint (11.2 vs 11.7), so
|
|
291
|
+
it's a real lift, not noise. Honest framing: 15.3% is still a **cheap-model** number — leaderboard
|
|
292
|
+
leaders hit 65–88% on Verified using frontier models + deeper agentic loops at $1–20/instance. The
|
|
293
|
+
contribution is the **harness lift on a fixed cheap model**: open-loop 7.7% → closed-loop 15.3% at
|
|
294
|
+
near-constant cost. Localization lifted file-selection recall **44.7% → 59.7%**; the repair loop
|
|
295
|
+
then converts first-shot misses into fixes via test feedback. Next levers: hybrid cheap→frontier
|
|
296
|
+
escalation (ADR-148), local-model repair (ADR-150). Every number is reproducible under `bench/swebench/`.
|
|
292
297
|
|
|
293
298
|
## Status
|
|
294
299
|
|
package/dist/cli.js
CHANGED
|
@@ -13,6 +13,7 @@
|
|
|
13
13
|
// leaderboard + the winner's lineage. Dependency-free.
|
|
14
14
|
import { resolve } from 'node:path';
|
|
15
15
|
import { evolve } from './evolve.js';
|
|
16
|
+
import { RuvllmMutator } from './ruvllm-mutator.js';
|
|
16
17
|
import { profileRepo } from './repo_profiler.js';
|
|
17
18
|
import { loadSuite, makeSuite, saveSuite, verifySuite } from './bench/suite.js';
|
|
18
19
|
function flag(name, fallback) {
|
|
@@ -95,7 +96,7 @@ async function main() {
|
|
|
95
96
|
}
|
|
96
97
|
if (command !== 'evolve') {
|
|
97
98
|
process.stderr.write('usage: metaharness-darwin <evolve|bench> …\n' +
|
|
98
|
-
' evolve <repo> [--generations N] [--children N] [--concurrency N] [--seed N] [--bench <suite.json>] [--tie faster] [--selection quality-diversity|behavioral-diversity|niche-steering|clade|pareto] [--crossover] [--epistasis] [--risk-budget N] [--fdr Q] [--curriculum] [--sandbox real|mock|agent]\n' +
|
|
99
|
+
' evolve <repo> [--generations N] [--children N] [--concurrency N] [--seed N] [--bench <suite.json>] [--tie faster] [--selection quality-diversity|behavioral-diversity|niche-steering|clade|pareto] [--crossover] [--epistasis] [--risk-budget N] [--fdr Q] [--curriculum] [--sandbox real|mock|agent] [--mutator deterministic|ruvllm] [--ruvllm-url URL] [--ruvllm-model M]\n' +
|
|
99
100
|
' bench create <repo> [--out <suite.json>]\n' +
|
|
100
101
|
' bench verify <suite.json>\n');
|
|
101
102
|
process.exit(1);
|
|
@@ -125,6 +126,13 @@ async function main() {
|
|
|
125
126
|
const curriculum = process.argv.includes('--curriculum');
|
|
126
127
|
const sbRaw = flag('--sandbox', 'real');
|
|
127
128
|
const sandboxMode = sbRaw === 'mock' || sbRaw === 'agent' ? sbRaw : 'real';
|
|
129
|
+
// ADR-259: pluggable mutator backend. Default = deterministic (no network, no key).
|
|
130
|
+
// --mutator ruvllm routes mutations to a local `ruvllm serve` endpoint (fully local,
|
|
131
|
+
// air-gapped, zero API cost). The OpenRouter LLM mutator stays library-only.
|
|
132
|
+
const mutatorRaw = flag('--mutator', 'deterministic');
|
|
133
|
+
const generator = mutatorRaw === 'ruvllm'
|
|
134
|
+
? new RuvllmMutator({ baseUrl: flag('--ruvllm-url', '') || undefined, model: flag('--ruvllm-model', '') || undefined })
|
|
135
|
+
: undefined;
|
|
128
136
|
const result = await evolve({
|
|
129
137
|
repoRoot,
|
|
130
138
|
workRoot,
|
|
@@ -143,6 +151,7 @@ async function main() {
|
|
|
143
151
|
...(fdrQ !== undefined ? { fdrQ } : {}),
|
|
144
152
|
...(curriculum ? { curriculum } : {}),
|
|
145
153
|
...(sandboxMode !== 'real' ? { sandboxMode } : {}),
|
|
154
|
+
...(generator ? { generator } : {}),
|
|
146
155
|
tieBreaker,
|
|
147
156
|
selection,
|
|
148
157
|
crossover,
|
package/dist/cli.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"cli.js","sourceRoot":"","sources":["../src/cli.ts"],"names":[],"mappings":";AACA,+BAA+B;AAC/B,EAAE;AACF,0BAA0B;AAC1B,EAAE;AACF,sEAAsE;AACtE,kEAAkE;AAClE,2EAA2E;AAC3E,gEAAgE;AAChE,iDAAiD;AACjD,EAAE;AACF,iFAAiF;AACjF,uDAAuD;AAEvD,OAAO,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AACpC,OAAO,EAAE,MAAM,EAAE,MAAM,aAAa,CAAC;AACrC,OAAO,EAAE,WAAW,EAAE,MAAM,oBAAoB,CAAC;AACjD,OAAO,EAAE,SAAS,EAAE,SAAS,EAAE,SAAS,EAAE,WAAW,EAAE,MAAM,kBAAkB,CAAC;AAIhF,SAAS,IAAI,CAAC,IAAY,EAAE,QAAgB;IAC1C,MAAM,CAAC,GAAG,OAAO,CAAC,IAAI,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;IACrC,OAAO,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,IAAI,QAAQ,CAAC,CAAC;AACjE,CAAC;AAED,SAAS,GAAG,CAAC,IAAY,EAAE,QAAgB;IACzC,MAAM,CAAC,GAAG,MAAM,CAAC,IAAI,CAAC,IAAI,EAAE,MAAM,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAC;IAC/C,OAAO,MAAM,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC;AAC3C,CAAC;AAED,SAAS,WAAW,CAAC,MAAuB;IAC1C,MAAM,MAAM,GAAG,MAAM,CAAC,OAAO;SAC1B,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC;SACtB,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,KAAK,EAAE,UAAU,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,KAAK,EAAE,UAAU,IAAI,CAAC,CAAC,CAAC,CAAC;IAE3E,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,+BAA+B,CAAC,CAAC;IACtD,KAAK,MAAM,CAAC,IAAI,MAAM,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,EAAE,CAAC;QACpC,MAAM,CAAC,GAAG,CAAC,CAAC,KAAM,CAAC;QACnB,MAAM,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,EAAE,KAAK,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,EAAE,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC,CAAC,EAAE,CAAC;QAC1E,OAAO,CAAC,MAAM,CAAC,KAAK,CAClB,KAAK,CAAC,CAAC,UAAU,CAAC,OAAO,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,OAAO,CAAC,EAAE,EAAE;YAC7C,MAAM,CAAC,CAAC,OAAO,CAAC,eAAe,aAAa,CAAC,CAAC,WAAW,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE;YACtE,UAAU,CAAC,CAAC,YAAY,CAAC,OAAO,CAAC,CAAC,CAAC,GAAG,GAAG,IAAI,CAChD,CAAC;IACJ,CAAC;IAED,IAAI,MAAM,CAAC,MAAM,EAAE,CAAC;QAClB,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,aAAa,MAAM,CAAC,MAAM,CAAC,OAAO,CAAC,EAAE,IAAI,CAAC,CAAC;QAChE,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,YAAY,MAAM,CAAC,aAAa,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;QACvE,MAAM,IAAI,GAAG,MAAM,CAAC,QAAQ,CAAC,KAAK,EAAE,UAAU,IAAI,CAAC,CAAC;QACpD,MAAM,GAAG,GAAG,MAAM,CAAC,MAAM,CAAC,KAAK,EAAE,UAAU,IAAI,CAAC,CAAC;QACjD,OAAO,CAAC,MAAM,CAAC,KAAK,CAClB,wBAAwB,CAAC,GAAG,GAAG,IAAI,IAAI,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,GAAG,CAAC,GAAG,GAAG,IAAI,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,IAAI,CACnF,CAAC;IACJ,CAAC;SAAM,CAAC;QACN,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,yBAAyB,CAAC,CAAC;IAClD,CAAC;AACH,CAAC;AAED,qEAAqE;AACrE,KAAK,UAAU,QAAQ;IACrB,MAAM,GAAG,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAE5B,IAAI,GAAG,KAAK,QAAQ,EAAE,CAAC;QACrB,MAAM,QAAQ,GAAG,OAAO,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,OAAO,CAAC,GAAG,EAAE,CAAC,CAAC;QAC3D,MAAM,OAAO,GAAG,MAAM,WAAW,CAAC,QAAQ,CAAC,CAAC;QAC5C,MAAM,GAAG,GAAG,OAAO,CAAC,IAAI,CAAC,OAAO,EAAE,OAAO,CAAC,QAAQ,EAAE,yBAAyB,CAAC,CAAC,CAAC,CAAC;QACjF,2EAA2E;QAC3E,8EAA8E;QAC9E,MAAM,IAAI,GAAkB;YAC1B,EAAE,EAAE,WAAW;YACf,IAAI,EAAE,QAAQ;YACd,MAAM,EAAE,SAAS;YACjB,KAAK,EAAE,wBAAwB;YAC/B,MAAM,EAAE,uCAAuC;YAC/C,iBAAiB,EAAE,OAAO,CAAC,WAAW;YACtC,iBAAiB,EAAE,OAAO,CAAC,WAAW;YACtC,qBAAqB,EAAE,OAAO,CAAC,WAAW;YAC1C,SAAS,EAAE,MAAM;YACjB,UAAU,EAAE,CAAC;YACb,oBAAoB,EAAE,EAAE;YACxB,YAAY,EAAE,CAAC,MAAM,EAAE,mBAAmB,EAAE,gBAAgB,EAAE,WAAW,EAAE,mBAAmB,CAAC;YAC/F,eAAe,EAAE,CAAC,oBAAoB,EAAE,oBAAoB,EAAE,yBAAyB,CAAC;YACxF,UAAU,EAAE,CAAC;YACb,IAAI,EAAE,CAAC,OAAO,EAAE,aAAa,CAAC;SAC/B,CAAC;QACF,MAAM,KAAK,GAAG,SAAS,CAAC,aAAa,EAAE,OAAO,EAAE,CAAC,IAAI,CAAC,CAAC,CAAC;QACxD,MAAM,SAAS,CAAC,GAAG,EAAE,KAAK,CAAC,CAAC;QAC5B,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,gBAAgB,KAAK,CAAC,KAAK,CAAC,MAAM,eAAe,KAAK,CAAC,QAAQ,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,OAAO,GAAG,IAAI,CAAC,CAAC;QACjH,OAAO;IACT,CAAC;IAED,IAAI,GAAG,KAAK,QAAQ,EAAE,CAAC;QACrB,MAAM,IAAI,GAAG,OAAO,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC;QAC5C,MAAM,KAAK,GAAG,MAAM,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC,mBAAmB;QACxD,MAAM,KAAK,GAAG,WAAW,CAAC,KAAK,CAAC,CAAC;QACjC,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,SAAS,KAAK,CAAC,EAAE,IAAI,KAAK,CAAC,OAAO,KAAK,KAAK,CAAC,KAAK,CAAC,MAAM,gBAAgB,KAAK,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,UAAU,KAAK,KAAK,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,MAAM,CAAC,CAAC;QAChK,OAAO;IACT,CAAC;IAED,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,qDAAqD,CAAC,CAAC;IAC5E,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;AAClB,CAAC;AAED,KAAK,UAAU,IAAI;IACjB,MAAM,OAAO,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAEhC,IAAI,OAAO,KAAK,OAAO,EAAE,CAAC;QACxB,MAAM,QAAQ,EAAE,CAAC;QACjB,OAAO;IACT,CAAC;IAED,IAAI,OAAO,KAAK,QAAQ,EAAE,CAAC;QACzB,OAAO,CAAC,MAAM,CAAC,KAAK,CAClB,8CAA8C;YAC5C,
|
|
1
|
+
{"version":3,"file":"cli.js","sourceRoot":"","sources":["../src/cli.ts"],"names":[],"mappings":";AACA,+BAA+B;AAC/B,EAAE;AACF,0BAA0B;AAC1B,EAAE;AACF,sEAAsE;AACtE,kEAAkE;AAClE,2EAA2E;AAC3E,gEAAgE;AAChE,iDAAiD;AACjD,EAAE;AACF,iFAAiF;AACjF,uDAAuD;AAEvD,OAAO,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AACpC,OAAO,EAAE,MAAM,EAAE,MAAM,aAAa,CAAC;AACrC,OAAO,EAAE,aAAa,EAAE,MAAM,qBAAqB,CAAC;AACpD,OAAO,EAAE,WAAW,EAAE,MAAM,oBAAoB,CAAC;AACjD,OAAO,EAAE,SAAS,EAAE,SAAS,EAAE,SAAS,EAAE,WAAW,EAAE,MAAM,kBAAkB,CAAC;AAIhF,SAAS,IAAI,CAAC,IAAY,EAAE,QAAgB;IAC1C,MAAM,CAAC,GAAG,OAAO,CAAC,IAAI,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;IACrC,OAAO,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,IAAI,QAAQ,CAAC,CAAC;AACjE,CAAC;AAED,SAAS,GAAG,CAAC,IAAY,EAAE,QAAgB;IACzC,MAAM,CAAC,GAAG,MAAM,CAAC,IAAI,CAAC,IAAI,EAAE,MAAM,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAC;IAC/C,OAAO,MAAM,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC;AAC3C,CAAC;AAED,SAAS,WAAW,CAAC,MAAuB;IAC1C,MAAM,MAAM,GAAG,MAAM,CAAC,OAAO;SAC1B,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC;SACtB,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,KAAK,EAAE,UAAU,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,KAAK,EAAE,UAAU,IAAI,CAAC,CAAC,CAAC,CAAC;IAE3E,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,+BAA+B,CAAC,CAAC;IACtD,KAAK,MAAM,CAAC,IAAI,MAAM,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,EAAE,CAAC;QACpC,MAAM,CAAC,GAAG,CAAC,CAAC,KAAM,CAAC;QACnB,MAAM,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,EAAE,KAAK,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,EAAE,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC,CAAC,EAAE,CAAC;QAC1E,OAAO,CAAC,MAAM,CAAC,KAAK,CAClB,KAAK,CAAC,CAAC,UAAU,CAAC,OAAO,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,OAAO,CAAC,EAAE,EAAE;YAC7C,MAAM,CAAC,CAAC,OAAO,CAAC,eAAe,aAAa,CAAC,CAAC,WAAW,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE;YACtE,UAAU,CAAC,CAAC,YAAY,CAAC,OAAO,CAAC,CAAC,CAAC,GAAG,GAAG,IAAI,CAChD,CAAC;IACJ,CAAC;IAED,IAAI,MAAM,CAAC,MAAM,EAAE,CAAC;QAClB,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,aAAa,MAAM,CAAC,MAAM,CAAC,OAAO,CAAC,EAAE,IAAI,CAAC,CAAC;QAChE,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,YAAY,MAAM,CAAC,aAAa,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;QACvE,MAAM,IAAI,GAAG,MAAM,CAAC,QAAQ,CAAC,KAAK,EAAE,UAAU,IAAI,CAAC,CAAC;QACpD,MAAM,GAAG,GAAG,MAAM,CAAC,MAAM,CAAC,KAAK,EAAE,UAAU,IAAI,CAAC,CAAC;QACjD,OAAO,CAAC,MAAM,CAAC,KAAK,CAClB,wBAAwB,CAAC,GAAG,GAAG,IAAI,IAAI,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,GAAG,CAAC,GAAG,GAAG,IAAI,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,IAAI,CACnF,CAAC;IACJ,CAAC;SAAM,CAAC;QACN,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,yBAAyB,CAAC,CAAC;IAClD,CAAC;AACH,CAAC;AAED,qEAAqE;AACrE,KAAK,UAAU,QAAQ;IACrB,MAAM,GAAG,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAE5B,IAAI,GAAG,KAAK,QAAQ,EAAE,CAAC;QACrB,MAAM,QAAQ,GAAG,OAAO,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,OAAO,CAAC,GAAG,EAAE,CAAC,CAAC;QAC3D,MAAM,OAAO,GAAG,MAAM,WAAW,CAAC,QAAQ,CAAC,CAAC;QAC5C,MAAM,GAAG,GAAG,OAAO,CAAC,IAAI,CAAC,OAAO,EAAE,OAAO,CAAC,QAAQ,EAAE,yBAAyB,CAAC,CAAC,CAAC,CAAC;QACjF,2EAA2E;QAC3E,8EAA8E;QAC9E,MAAM,IAAI,GAAkB;YAC1B,EAAE,EAAE,WAAW;YACf,IAAI,EAAE,QAAQ;YACd,MAAM,EAAE,SAAS;YACjB,KAAK,EAAE,wBAAwB;YAC/B,MAAM,EAAE,uCAAuC;YAC/C,iBAAiB,EAAE,OAAO,CAAC,WAAW;YACtC,iBAAiB,EAAE,OAAO,CAAC,WAAW;YACtC,qBAAqB,EAAE,OAAO,CAAC,WAAW;YAC1C,SAAS,EAAE,MAAM;YACjB,UAAU,EAAE,CAAC;YACb,oBAAoB,EAAE,EAAE;YACxB,YAAY,EAAE,CAAC,MAAM,EAAE,mBAAmB,EAAE,gBAAgB,EAAE,WAAW,EAAE,mBAAmB,CAAC;YAC/F,eAAe,EAAE,CAAC,oBAAoB,EAAE,oBAAoB,EAAE,yBAAyB,CAAC;YACxF,UAAU,EAAE,CAAC;YACb,IAAI,EAAE,CAAC,OAAO,EAAE,aAAa,CAAC;SAC/B,CAAC;QACF,MAAM,KAAK,GAAG,SAAS,CAAC,aAAa,EAAE,OAAO,EAAE,CAAC,IAAI,CAAC,CAAC,CAAC;QACxD,MAAM,SAAS,CAAC,GAAG,EAAE,KAAK,CAAC,CAAC;QAC5B,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,gBAAgB,KAAK,CAAC,KAAK,CAAC,MAAM,eAAe,KAAK,CAAC,QAAQ,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,OAAO,GAAG,IAAI,CAAC,CAAC;QACjH,OAAO;IACT,CAAC;IAED,IAAI,GAAG,KAAK,QAAQ,EAAE,CAAC;QACrB,MAAM,IAAI,GAAG,OAAO,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC;QAC5C,MAAM,KAAK,GAAG,MAAM,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC,mBAAmB;QACxD,MAAM,KAAK,GAAG,WAAW,CAAC,KAAK,CAAC,CAAC;QACjC,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,SAAS,KAAK,CAAC,EAAE,IAAI,KAAK,CAAC,OAAO,KAAK,KAAK,CAAC,KAAK,CAAC,MAAM,gBAAgB,KAAK,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,UAAU,KAAK,KAAK,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,MAAM,CAAC,CAAC;QAChK,OAAO;IACT,CAAC;IAED,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,qDAAqD,CAAC,CAAC;IAC5E,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;AAClB,CAAC;AAED,KAAK,UAAU,IAAI;IACjB,MAAM,OAAO,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAEhC,IAAI,OAAO,KAAK,OAAO,EAAE,CAAC;QACxB,MAAM,QAAQ,EAAE,CAAC;QACjB,OAAO;IACT,CAAC;IAED,IAAI,OAAO,KAAK,QAAQ,EAAE,CAAC;QACzB,OAAO,CAAC,MAAM,CAAC,KAAK,CAClB,8CAA8C;YAC5C,kXAAkX;YAClX,8CAA8C;YAC9C,+BAA+B,CAClC,CAAC;QACF,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC;IAED,MAAM,QAAQ,GAAG,OAAO,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,OAAO,CAAC,GAAG,EAAE,CAAC,CAAC;IAC3D,MAAM,QAAQ,GAAG,OAAO,CAAC,QAAQ,EAAE,cAAc,CAAC,CAAC;IAEnD,sEAAsE;IACtE,0EAA0E;IAC1E,8EAA8E;IAC9E,MAAM,SAAS,GAAG,IAAI,CAAC,SAAS,EAAE,EAAE,CAAC,CAAC;IACtC,MAAM,UAAU,GAAG,SAAS,CAAC,CAAC,CAAC,MAAM,SAAS,CAAC,OAAO,CAAC,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;IAC/E,MAAM,UAAU,GAAG,IAAI,CAAC,OAAO,EAAE,WAAW,CAAC,KAAK,QAAQ,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,WAAW,CAAC;IACpF,MAAM,MAAM,GAAG,IAAI,CAAC,aAAa,EAAE,OAAO,CAAC,CAAC;IAC5C,MAAM,SAAS,GACb,MAAM,KAAK,mBAAmB;QAC9B,MAAM,KAAK,sBAAsB;QACjC,MAAM,KAAK,gBAAgB;QAC3B,MAAM,KAAK,OAAO;QAClB,MAAM,KAAK,QAAQ;QACjB,CAAC,CAAC,MAAM;QACR,CAAC,CAAC,OAAO,CAAC;IACd,MAAM,SAAS,GAAG,OAAO,CAAC,IAAI,CAAC,QAAQ,CAAC,aAAa,CAAC,CAAC;IACvD,MAAM,SAAS,GAAG,OAAO,CAAC,IAAI,CAAC,QAAQ,CAAC,aAAa,CAAC,CAAC;IACvD,MAAM,OAAO,GAAG,IAAI,CAAC,eAAe,EAAE,EAAE,CAAC,CAAC;IAC1C,MAAM,eAAe,GAAG,OAAO,KAAK,EAAE,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,GAAG,CAAC,eAAe,EAAE,CAAC,CAAC,CAAC;IAC7E,MAAM,MAAM,GAAG,IAAI,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC;IACjC,MAAM,IAAI,GAAG,MAAM,KAAK,EAAE,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,EAAE,IAAI,CAAC,CAAC;IAC5D,MAAM,UAAU,GAAG,OAAO,CAAC,IAAI,CAAC,QAAQ,CAAC,cAAc,CAAC,CAAC;IACzD,MAAM,KAAK,GAAG,IAAI,CAAC,WAAW,EAAE,MAAM,CAAC,CAAC;IACxC,MAAM,WAAW,GAAG,KAAK,KAAK,MAAM,IAAI,KAAK,KAAK,OAAO,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,CAAC;IAE3E,oFAAoF;IACpF,qFAAqF;IACrF,6EAA6E;IAC7E,MAAM,UAAU,GAAG,IAAI,CAAC,WAAW,EAAE,eAAe,CAAC,CAAC;IACtD,MAAM,SAAS,GACb,UAAU,KAAK,QAAQ;QACrB,CAAC,CAAC,IAAI,aAAa,CAAC,EAAE,OAAO,EAAE,IAAI,CAAC,cAAc,EAAE,EAAE,CAAC,IAAI,SAAS,EAAE,KAAK,EAAE,IAAI,CAAC,gBAAgB,EAAE,EAAE,CAAC,IAAI,SAAS,EAAE,CAAC;QACvH,CAAC,CAAC,SAAS,CAAC;IAEhB,MAAM,MAAM,GAAG,MAAM,MAAM,CAAC;QAC1B,QAAQ;QACR,QAAQ;QACR,WAAW,EAAE,GAAG,CAAC,eAAe,EAAE,CAAC,CAAC;QACpC,qBAAqB,EAAE,GAAG,CAAC,YAAY,EAAE,CAAC,CAAC;QAC3C,WAAW,EAAE,GAAG,CAAC,eAAe,EAAE,CAAC,CAAC;QACpC,IAAI,EAAE,GAAG,CAAC,QAAQ,EAAE,CAAC,CAAC;QACtB,cAAc,EAAE,IAAI;QACpB,KAAK,EAAE;YACL,2BAA2B;YAC3B,iCAAiC;YACjC,qBAAqB;SACtB;QACD,GAAG,CAAC,UAAU,CAAC,CAAC,CAAC,EAAE,UAAU,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACrC,GAAG,CAAC,eAAe,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,eAAe,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QAC7D,GAAG,CAAC,IAAI,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACvC,GAAG,CAAC,UAAU,CAAC,CAAC,CAAC,EAAE,UAAU,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACrC,GAAG,CAAC,WAAW,KAAK,MAAM,CAAC,CAAC,CAAC,EAAE,WAAW,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QAClD,GAAG,CAAC,SAAS,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACnC,UAAU;QACV,SAAS;QACT,SAAS;QACT,SAAS;KACV,CAAC,CAAC;IAEH,WAAW,CAAC,MAAM,CAAC,CAAC;IACpB,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,gBAAgB,QAAQ,IAAI,CAAC,CAAC;AACrD,CAAC;AAED,IAAI,EAAE,CAAC,KAAK,CAAC,CAAC,KAAK,EAAE,EAAE;IACrB,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,GAAG,KAAK,YAAY,KAAK,CAAC,CAAC,CAAC,KAAK,CAAC,KAAK,IAAI,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;IACnG,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;AAClB,CAAC,CAAC,CAAC"}
|
package/dist/index.d.ts
CHANGED
|
@@ -5,6 +5,7 @@ export * from './templates.js';
|
|
|
5
5
|
export * from './generator.js';
|
|
6
6
|
export * from './mutator.js';
|
|
7
7
|
export * from './openrouter-mutator.js';
|
|
8
|
+
export * from './ruvllm-mutator.js';
|
|
8
9
|
export * from './phenotype.js';
|
|
9
10
|
export * from './epistasis.js';
|
|
10
11
|
export * from './clade.js';
|
package/dist/index.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAyBA,cAAc,YAAY,CAAC;AAC3B,cAAc,aAAa,CAAC;AAC5B,cAAc,oBAAoB,CAAC;AACnC,cAAc,gBAAgB,CAAC;AAC/B,cAAc,gBAAgB,CAAC;AAC/B,cAAc,cAAc,CAAC;AAC7B,cAAc,yBAAyB,CAAC;AACxC,cAAc,gBAAgB,CAAC;AAC/B,cAAc,gBAAgB,CAAC;AAC/B,cAAc,YAAY,CAAC;AAC3B,cAAc,iBAAiB,CAAC;AAChC,cAAc,aAAa,CAAC;AAC5B,cAAc,mBAAmB,CAAC;AAClC,cAAc,oBAAoB,CAAC;AACnC,cAAc,cAAc,CAAC;AAC7B,cAAc,aAAa,CAAC;AAC5B,cAAc,cAAc,CAAC;AAC7B,cAAc,aAAa,CAAC;AAI5B,OAAO,KAAK,KAAK,MAAM,kBAAkB,CAAC"}
|
|
1
|
+
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAyBA,cAAc,YAAY,CAAC;AAC3B,cAAc,aAAa,CAAC;AAC5B,cAAc,oBAAoB,CAAC;AACnC,cAAc,gBAAgB,CAAC;AAC/B,cAAc,gBAAgB,CAAC;AAC/B,cAAc,cAAc,CAAC;AAC7B,cAAc,yBAAyB,CAAC;AACxC,cAAc,qBAAqB,CAAC;AACpC,cAAc,gBAAgB,CAAC;AAC/B,cAAc,gBAAgB,CAAC;AAC/B,cAAc,YAAY,CAAC;AAC3B,cAAc,iBAAiB,CAAC;AAChC,cAAc,aAAa,CAAC;AAC5B,cAAc,mBAAmB,CAAC;AAClC,cAAc,oBAAoB,CAAC;AACnC,cAAc,cAAc,CAAC;AAC7B,cAAc,aAAa,CAAC;AAC5B,cAAc,cAAc,CAAC;AAC7B,cAAc,aAAa,CAAC;AAI5B,OAAO,KAAK,KAAK,MAAM,kBAAkB,CAAC"}
|
package/dist/index.js
CHANGED
|
@@ -29,6 +29,7 @@ export * from './templates.js';
|
|
|
29
29
|
export * from './generator.js';
|
|
30
30
|
export * from './mutator.js';
|
|
31
31
|
export * from './openrouter-mutator.js';
|
|
32
|
+
export * from './ruvllm-mutator.js';
|
|
32
33
|
export * from './phenotype.js';
|
|
33
34
|
export * from './epistasis.js';
|
|
34
35
|
export * from './clade.js';
|
package/dist/index.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.js","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA,+BAA+B;AAC/B,EAAE;AACF,mDAAmD;AACnD,EAAE;AACF,6EAA6E;AAC7E,+EAA+E;AAC/E,iFAAiF;AACjF,8EAA8E;AAC9E,EAAE;AACF,iFAAiF;AACjF,2BAA2B;AAC3B,EAAE;AACF,WAAW;AACX,yCAAyC;AACzC,yEAAyE;AACzE,wCAAwC;AACxC,iEAAiE;AACjE,oDAAoD;AACpD,iFAAiF;AACjF,8EAA8E;AAC9E,sEAAsE;AACtE,kFAAkF;AAClF,4EAA4E;AAC5E,+DAA+D;AAE/D,cAAc,YAAY,CAAC;AAC3B,cAAc,aAAa,CAAC;AAC5B,cAAc,oBAAoB,CAAC;AACnC,cAAc,gBAAgB,CAAC;AAC/B,cAAc,gBAAgB,CAAC;AAC/B,cAAc,cAAc,CAAC;AAC7B,cAAc,yBAAyB,CAAC;AACxC,cAAc,gBAAgB,CAAC;AAC/B,cAAc,gBAAgB,CAAC;AAC/B,cAAc,YAAY,CAAC;AAC3B,cAAc,iBAAiB,CAAC;AAChC,cAAc,aAAa,CAAC;AAC5B,cAAc,mBAAmB,CAAC;AAClC,cAAc,oBAAoB,CAAC;AACnC,cAAc,cAAc,CAAC;AAC7B,cAAc,aAAa,CAAC;AAC5B,cAAc,cAAc,CAAC;AAC7B,cAAc,aAAa,CAAC;AAE5B,gFAAgF;AAChF,2EAA2E;AAC3E,OAAO,KAAK,KAAK,MAAM,kBAAkB,CAAC"}
|
|
1
|
+
{"version":3,"file":"index.js","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA,+BAA+B;AAC/B,EAAE;AACF,mDAAmD;AACnD,EAAE;AACF,6EAA6E;AAC7E,+EAA+E;AAC/E,iFAAiF;AACjF,8EAA8E;AAC9E,EAAE;AACF,iFAAiF;AACjF,2BAA2B;AAC3B,EAAE;AACF,WAAW;AACX,yCAAyC;AACzC,yEAAyE;AACzE,wCAAwC;AACxC,iEAAiE;AACjE,oDAAoD;AACpD,iFAAiF;AACjF,8EAA8E;AAC9E,sEAAsE;AACtE,kFAAkF;AAClF,4EAA4E;AAC5E,+DAA+D;AAE/D,cAAc,YAAY,CAAC;AAC3B,cAAc,aAAa,CAAC;AAC5B,cAAc,oBAAoB,CAAC;AACnC,cAAc,gBAAgB,CAAC;AAC/B,cAAc,gBAAgB,CAAC;AAC/B,cAAc,cAAc,CAAC;AAC7B,cAAc,yBAAyB,CAAC;AACxC,cAAc,qBAAqB,CAAC;AACpC,cAAc,gBAAgB,CAAC;AAC/B,cAAc,gBAAgB,CAAC;AAC/B,cAAc,YAAY,CAAC;AAC3B,cAAc,iBAAiB,CAAC;AAChC,cAAc,aAAa,CAAC;AAC5B,cAAc,mBAAmB,CAAC;AAClC,cAAc,oBAAoB,CAAC;AACnC,cAAc,cAAc,CAAC;AAC7B,cAAc,aAAa,CAAC;AAC5B,cAAc,cAAc,CAAC;AAC7B,cAAc,aAAa,CAAC;AAE5B,gFAAgF;AAChF,2EAA2E;AAC3E,OAAO,KAAK,KAAK,MAAM,kBAAkB,CAAC"}
|
|
@@ -0,0 +1,32 @@
|
|
|
1
|
+
import type { CodeGenerator } from './mutator.js';
|
|
2
|
+
import type { MutationSurface } from './types.js';
|
|
3
|
+
export interface RuvllmMutatorOptions {
|
|
4
|
+
/** Base URL of the `ruvllm serve` endpoint. Default: http://localhost:8080 (or RUVLLM_URL). */
|
|
5
|
+
baseUrl?: string;
|
|
6
|
+
/** Model name passed in the request body. Default: 'local' (or RUVLLM_MODEL). */
|
|
7
|
+
model?: string;
|
|
8
|
+
maxTokens?: number;
|
|
9
|
+
temperature?: number;
|
|
10
|
+
/** Request timeout in ms. Default: 30_000. */
|
|
11
|
+
timeoutMs?: number;
|
|
12
|
+
}
|
|
13
|
+
export declare class RuvllmMutator implements CodeGenerator {
|
|
14
|
+
private readonly baseUrl;
|
|
15
|
+
private readonly model;
|
|
16
|
+
private readonly maxTokens;
|
|
17
|
+
private readonly temperature;
|
|
18
|
+
private readonly timeoutMs;
|
|
19
|
+
constructor(opts?: RuvllmMutatorOptions);
|
|
20
|
+
generateMutation(input: {
|
|
21
|
+
parentCode: string;
|
|
22
|
+
surface: MutationSurface;
|
|
23
|
+
repoSummary: string;
|
|
24
|
+
parentScore: number;
|
|
25
|
+
failedTraces: string[];
|
|
26
|
+
nonce?: number;
|
|
27
|
+
}): Promise<{
|
|
28
|
+
code: string;
|
|
29
|
+
summary: string;
|
|
30
|
+
}>;
|
|
31
|
+
}
|
|
32
|
+
//# sourceMappingURL=ruvllm-mutator.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"ruvllm-mutator.d.ts","sourceRoot":"","sources":["../src/ruvllm-mutator.ts"],"names":[],"mappings":"AAiBA,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,cAAc,CAAC;AAClD,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,YAAY,CAAC;AAElD,MAAM,WAAW,oBAAoB;IACnC,+FAA+F;IAC/F,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,iFAAiF;IACjF,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,8CAA8C;IAC9C,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB;AAQD,qBAAa,aAAc,YAAW,aAAa;IACjD,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAS;IACjC,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAS;IAC/B,OAAO,CAAC,QAAQ,CAAC,SAAS,CAAS;IACnC,OAAO,CAAC,QAAQ,CAAC,WAAW,CAAS;IACrC,OAAO,CAAC,QAAQ,CAAC,SAAS,CAAS;gBAEvB,IAAI,GAAE,oBAAyB;IAQrC,gBAAgB,CAAC,KAAK,EAAE;QAC5B,UAAU,EAAE,MAAM,CAAC;QACnB,OAAO,EAAE,eAAe,CAAC;QACzB,WAAW,EAAE,MAAM,CAAC;QACpB,WAAW,EAAE,MAAM,CAAC;QACpB,YAAY,EAAE,MAAM,EAAE,CAAC;QACvB,KAAK,CAAC,EAAE,MAAM,CAAC;KAChB,GAAG,OAAO,CAAC;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,OAAO,EAAE,MAAM,CAAA;KAAE,CAAC;CAuC/C"}
|
|
@@ -0,0 +1,73 @@
|
|
|
1
|
+
// SPDX-License-Identifier: MIT
|
|
2
|
+
//
|
|
3
|
+
// RuvllmMutator — local ruvllm server backend for Darwin Mode (ADR-259).
|
|
4
|
+
//
|
|
5
|
+
// Implements the `CodeGenerator` interface against a local `ruvllm serve` endpoint
|
|
6
|
+
// (OpenAI-compatible `POST /v1/chat/completions`). This makes evolution **fully local,
|
|
7
|
+
// air-gapped, and zero-API-cost** — no `OPENROUTER_API_KEY`, no public network. On-thesis
|
|
8
|
+
// for the "runs anywhere / cost optimizer" positioning.
|
|
9
|
+
//
|
|
10
|
+
// NOTE on value (honest, per ADR-087): the mutator is NOT the quality lever — the
|
|
11
|
+
// deterministic and frontier-LLM mutators both hit the 0.985 scorer ceiling. So
|
|
12
|
+
// RuvllmMutator's benefit is *operational* (local/free/private), not higher scores.
|
|
13
|
+
//
|
|
14
|
+
// Zero runtime dependencies; uses Node's built-in fetch (Node ≥ 18). Falls back to a
|
|
15
|
+
// safe no-op (returns the parent code unchanged) if the server is unreachable — the
|
|
16
|
+
// same contract as OpenRouterMutator, so a down server never breaks the evolution loop.
|
|
17
|
+
/** Strip a single ```fenced``` block if the model wrapped its output. */
|
|
18
|
+
function unfence(text) {
|
|
19
|
+
const m = text.match(/```(?:[a-zA-Z0-9]+)?\n([\s\S]*?)\n```/);
|
|
20
|
+
return (m ? m[1] : text).trim() + '\n';
|
|
21
|
+
}
|
|
22
|
+
export class RuvllmMutator {
|
|
23
|
+
baseUrl;
|
|
24
|
+
model;
|
|
25
|
+
maxTokens;
|
|
26
|
+
temperature;
|
|
27
|
+
timeoutMs;
|
|
28
|
+
constructor(opts = {}) {
|
|
29
|
+
this.baseUrl = (opts.baseUrl ?? process.env.RUVLLM_URL ?? 'http://localhost:8080').replace(/\/$/, '');
|
|
30
|
+
this.model = opts.model ?? process.env.RUVLLM_MODEL ?? 'local';
|
|
31
|
+
this.maxTokens = opts.maxTokens ?? 2000;
|
|
32
|
+
this.temperature = opts.temperature ?? 0.4;
|
|
33
|
+
this.timeoutMs = opts.timeoutMs ?? 30_000;
|
|
34
|
+
}
|
|
35
|
+
async generateMutation(input) {
|
|
36
|
+
const sys = 'You improve ONE file of an AI agent harness. Output ONLY the full replacement file — ' +
|
|
37
|
+
'no prose, no fences. HARD RULES: keep every exported name and signature identical; ' +
|
|
38
|
+
'introduce NO new capabilities, imports, network, filesystem, shell, or env access; ' +
|
|
39
|
+
'no new dependencies; pure refactor/tuning only. Make a small, plausibly score-improving ' +
|
|
40
|
+
`change to the "${input.surface}" surface.`;
|
|
41
|
+
const user = `Surface: ${input.surface}\nParent score: ${input.parentScore}\n` +
|
|
42
|
+
(input.repoSummary ? `Repo: ${input.repoSummary}\n` : '') +
|
|
43
|
+
(input.failedTraces.length ? `Recent failures:\n${input.failedTraces.slice(0, 5).join('\n')}\n` : '') +
|
|
44
|
+
`\n--- current file ---\n${input.parentCode}\n--- end ---\n` +
|
|
45
|
+
'Return the improved full file.';
|
|
46
|
+
let res;
|
|
47
|
+
try {
|
|
48
|
+
const controller = new AbortController();
|
|
49
|
+
const tid = setTimeout(() => controller.abort(), this.timeoutMs);
|
|
50
|
+
res = await fetch(`${this.baseUrl}/v1/chat/completions`, {
|
|
51
|
+
method: 'POST',
|
|
52
|
+
headers: { 'Content-Type': 'application/json' },
|
|
53
|
+
body: JSON.stringify({
|
|
54
|
+
model: this.model,
|
|
55
|
+
messages: [{ role: 'system', content: sys }, { role: 'user', content: user }],
|
|
56
|
+
max_tokens: this.maxTokens,
|
|
57
|
+
temperature: this.temperature,
|
|
58
|
+
}),
|
|
59
|
+
signal: controller.signal,
|
|
60
|
+
});
|
|
61
|
+
clearTimeout(tid);
|
|
62
|
+
}
|
|
63
|
+
catch (e) {
|
|
64
|
+
return { code: input.parentCode, summary: `ruvllm:${this.baseUrl} unreachable (${e.message}) — no-op` };
|
|
65
|
+
}
|
|
66
|
+
const j = (await res.json());
|
|
67
|
+
const content = j.choices?.[0]?.message?.content;
|
|
68
|
+
if (!content)
|
|
69
|
+
return { code: input.parentCode, summary: `ruvllm:${this.model} no content — no-op` };
|
|
70
|
+
return { code: unfence(content), summary: `ruvllm:${this.model} regenerated ${input.surface}` };
|
|
71
|
+
}
|
|
72
|
+
}
|
|
73
|
+
//# sourceMappingURL=ruvllm-mutator.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"ruvllm-mutator.js","sourceRoot":"","sources":["../src/ruvllm-mutator.ts"],"names":[],"mappings":"AAAA,+BAA+B;AAC/B,EAAE;AACF,yEAAyE;AACzE,EAAE;AACF,mFAAmF;AACnF,uFAAuF;AACvF,0FAA0F;AAC1F,wDAAwD;AACxD,EAAE;AACF,kFAAkF;AAClF,gFAAgF;AAChF,oFAAoF;AACpF,EAAE;AACF,qFAAqF;AACrF,oFAAoF;AACpF,wFAAwF;AAgBxF,yEAAyE;AACzE,SAAS,OAAO,CAAC,IAAY;IAC3B,MAAM,CAAC,GAAG,IAAI,CAAC,KAAK,CAAC,uCAAuC,CAAC,CAAC;IAC9D,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAE,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,GAAG,IAAI,CAAC;AAC1C,CAAC;AAED,MAAM,OAAO,aAAa;IACP,OAAO,CAAS;IAChB,KAAK,CAAS;IACd,SAAS,CAAS;IAClB,WAAW,CAAS;IACpB,SAAS,CAAS;IAEnC,YAAY,OAA6B,EAAE;QACzC,IAAI,CAAC,OAAO,GAAG,CAAC,IAAI,CAAC,OAAO,IAAI,OAAO,CAAC,GAAG,CAAC,UAAU,IAAI,uBAAuB,CAAC,CAAC,OAAO,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC;QACtG,IAAI,CAAC,KAAK,GAAG,IAAI,CAAC,KAAK,IAAI,OAAO,CAAC,GAAG,CAAC,YAAY,IAAI,OAAO,CAAC;QAC/D,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC,SAAS,IAAI,IAAI,CAAC;QACxC,IAAI,CAAC,WAAW,GAAG,IAAI,CAAC,WAAW,IAAI,GAAG,CAAC;QAC3C,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC,SAAS,IAAI,MAAM,CAAC;IAC5C,CAAC;IAED,KAAK,CAAC,gBAAgB,CAAC,KAOtB;QACC,MAAM,GAAG,GACP,uFAAuF;YACvF,qFAAqF;YACrF,qFAAqF;YACrF,0FAA0F;YAC1F,kBAAkB,KAAK,CAAC,OAAO,YAAY,CAAC;QAC9C,MAAM,IAAI,GACR,YAAY,KAAK,CAAC,OAAO,mBAAmB,KAAK,CAAC,WAAW,IAAI;YACjE,CAAC,KAAK,CAAC,WAAW,CAAC,CAAC,CAAC,SAAS,KAAK,CAAC,WAAW,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC;YACzD,CAAC,KAAK,CAAC,YAAY,CAAC,MAAM,CAAC,CAAC,CAAC,qBAAqB,KAAK,CAAC,YAAY,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC;YACrG,2BAA2B,KAAK,CAAC,UAAU,iBAAiB;YAC5D,gCAAgC,CAAC;QAEnC,IAAI,GAAa,CAAC;QAClB,IAAI,CAAC;YACH,MAAM,UAAU,GAAG,IAAI,eAAe,EAAE,CAAC;YACzC,MAAM,GAAG,GAAG,UAAU,CAAC,GAAG,EAAE,CAAC,UAAU,CAAC,KAAK,EAAE,EAAE,IAAI,CAAC,SAAS,CAAC,CAAC;YACjE,GAAG,GAAG,MAAM,KAAK,CAAC,GAAG,IAAI,CAAC,OAAO,sBAAsB,EAAE;gBACvD,MAAM,EAAE,MAAM;gBACd,OAAO,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE;gBAC/C,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC;oBACnB,KAAK,EAAE,IAAI,CAAC,KAAK;oBACjB,QAAQ,EAAE,CAAC,EAAE,IAAI,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,EAAE,EAAE,EAAE,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC;oBAC7E,UAAU,EAAE,IAAI,CAAC,SAAS;oBAC1B,WAAW,EAAE,IAAI,CAAC,WAAW;iBAC9B,CAAC;gBACF,MAAM,EAAE,UAAU,CAAC,MAAM;aAC1B,CAAC,CAAC;YACH,YAAY,CAAC,GAAG,CAAC,CAAC;QACpB,CAAC;QAAC,OAAO,CAAC,EAAE,CAAC;YACX,OAAO,EAAE,IAAI,EAAE,KAAK,CAAC,UAAU,EAAE,OAAO,EAAE,UAAU,IAAI,CAAC,OAAO,iBAAkB,CAAW,CAAC,OAAO,WAAW,EAAE,CAAC;QACrH,CAAC;QAED,MAAM,CAAC,GAAG,CAAC,MAAM,GAAG,CAAC,IAAI,EAAE,CAA4D,CAAC;QACxF,MAAM,OAAO,GAAG,CAAC,CAAC,OAAO,EAAE,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,OAAO,CAAC;QACjD,IAAI,CAAC,OAAO;YAAE,OAAO,EAAE,IAAI,EAAE,KAAK,CAAC,UAAU,EAAE,OAAO,EAAE,UAAU,IAAI,CAAC,KAAK,qBAAqB,EAAE,CAAC;QACpG,OAAO,EAAE,IAAI,EAAE,OAAO,CAAC,OAAO,CAAC,EAAE,OAAO,EAAE,UAAU,IAAI,CAAC,KAAK,gBAAgB,KAAK,CAAC,OAAO,EAAE,EAAE,CAAC;IAClG,CAAC;CACF"}
|
package/package.json
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@metaharness/darwin",
|
|
3
|
-
"version": "0.2.
|
|
4
|
-
"description": "An LLM supercharger and cost optimizer:
|
|
3
|
+
"version": "0.2.3",
|
|
4
|
+
"description": "An LLM supercharger and cost optimizer: freeze your model, evolve the harness around it. Measured on full SWE-bench Lite (300, official swebench Docker): open-loop 7.7% -> +repair 15.3% -> +Barbarian&Scholar hybrid 33.3%, at ~$0.01-$0.34/instance (vs $1-20 for frontier agents). Mutate->sandbox->score->archive. Dependency-free (Node built-ins).",
|
|
5
5
|
"type": "module",
|
|
6
6
|
"main": "./dist/index.js",
|
|
7
7
|
"types": "./dist/index.d.ts",
|
|
@@ -18,7 +18,9 @@
|
|
|
18
18
|
"dist/**",
|
|
19
19
|
"README.md",
|
|
20
20
|
"SECURITY.md",
|
|
21
|
-
"LICENSE"
|
|
21
|
+
"LICENSE",
|
|
22
|
+
"CHANGELOG.md",
|
|
23
|
+
"LEARNINGS.md"
|
|
22
24
|
],
|
|
23
25
|
"scripts": {
|
|
24
26
|
"build": "tsc",
|