pi-smart-router 0.15.0 → 0.16.0
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/README.md +1 -1
- package/config/benchmark-profiles.json +2 -2
- package/package.json +2 -1
package/README.md
CHANGED
|
@@ -734,7 +734,7 @@ Dry-run behavior:
|
|
|
734
734
|
| `--include-excluded-in-fit` | Weak rows may warm-start the fit pool; `ece_eligible` / holdout ECE stay verifier-grade |
|
|
735
735
|
| Soft threshold | Advisory `0.25` calibrated ECE — **does not** change `config/release-gates.json` |
|
|
736
736
|
|
|
737
|
-
**#96 / `modernbert_k4` advisory:** when deciding whether to enable ModernBERT K=4 heads, use **pack holdout ECE / Top-1 error on verifier-grade packs** (SWE-Gym + FC-RewardBench), not fixture-only QR and **not** weak-fit ECE. Weak TwinRouterBench rows are warm-start only. This task does **not** flip `modernbert_k4` defaults. Go/no-go evidence (SP-204 / #113): [`spine-tasks/_authoring/release-v0.11.0/encoder-gonogo-artifact.md`](spine-tasks/_authoring/release-v0.11.0/encoder-gonogo-artifact.md).
|
|
737
|
+
**#96 / `modernbert_k4` advisory:** when deciding whether to enable ModernBERT K=4 heads, use **pack holdout ECE / Top-1 error on verifier-grade packs** (SWE-Gym + FC-RewardBench), not fixture-only QR and **not** weak-fit ECE. Weak TwinRouterBench rows are warm-start only. This task does **not** flip `modernbert_k4` defaults. Go/no-go evidence (SP-204 / #113): [`spine-tasks/_authoring/release-v0.11.0/encoder-gonogo-artifact.md`](spine-tasks/_authoring/release-v0.11.0/encoder-gonogo-artifact.md). K=4 Top-1 / offline A/B decision (SP-219 / #114): [`spine-tasks/_authoring/release-v0.16.0/modernbert-k4-top1-artifact.md`](spine-tasks/_authoring/release-v0.16.0/modernbert-k4-top1-artifact.md) — recommendation: **keep default** (gate not measurable without trained heads; #96 open).
|
|
738
738
|
|
|
739
739
|
### Operator tuning (frugality slider)
|
|
740
740
|
|
|
@@ -7,8 +7,8 @@
|
|
|
7
7
|
"livecodebench": "https://livecodebench.github.io/leaderboard.html",
|
|
8
8
|
"bfcl": "https://gorilla.cs.berkeley.edu/leaderboard.html"
|
|
9
9
|
},
|
|
10
|
-
"scrape_date": "2026-
|
|
11
|
-
"catalog_freeze_date": "2026-
|
|
10
|
+
"scrape_date": "2026-08-03",
|
|
11
|
+
"catalog_freeze_date": "2026-08-03"
|
|
12
12
|
},
|
|
13
13
|
"aliases": {
|
|
14
14
|
"anthropic/claude-opus-4": "claude-opus-4-5",
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "pi-smart-router",
|
|
3
|
-
"version": "0.
|
|
3
|
+
"version": "0.16.0",
|
|
4
4
|
"description": "Auto-model router middleware for the pi.dev coding agent",
|
|
5
5
|
"type": "module",
|
|
6
6
|
"license": "MIT",
|
|
@@ -69,6 +69,7 @@
|
|
|
69
69
|
"routing:verify-calibration": "bash -c 'if [[ \" $* \" == *\" --skip-embed \"* ]]; then vitest run tests/unit/train-routing-calibration.test.ts -t \"verifyRoutingCalibration passes|evaluates benchmark triage gates\"; else npm run build && tsx scripts/verify-routing-calibration.ts \"$@\"; fi' --",
|
|
70
70
|
"routing:calibration-dry-run": "tsx scripts/verify-routing-calibration.ts --dry-run-packs",
|
|
71
71
|
"routing:eval-replay": "tsx scripts/eval/counterfactual-replay.ts",
|
|
72
|
+
"routing:eval-k4-ab": "tsx scripts/eval/counterfactual-replay.ts --k4-ab --corpus tests/eval/corpus/twinrouterbench/ci-subset.json",
|
|
72
73
|
"routing:eval-harness": "tsx scripts/eval/run-harness.ts",
|
|
73
74
|
"routing:eval-harness:smoke": "tsx scripts/eval/run-harness.ts --summary-only",
|
|
74
75
|
"routing:eval-harness:corpus-smoke": "tsx scripts/eval/run-harness.ts --fixtures tests/eval/corpus/twinrouterbench --summary-only",
|