@miller-tech/uap 2.48.0 → 2.48.2

package/docs/plans/droid-client-performance-options.md

@@ -250,3 +250,59 @@ deliver run), which inflates its latency but not its gate outcomes. The
 frontier anatomy counts Droid-native tool results, not the UAP tool layer the
 monitoring round measured — the classes are aligned but the instruments
 differ; cross-comparison of rates between the two instruments is qualitative.
+
+---
+
+## Monitoring round 2 (2026-11-14 evening) — post-O-C/O-D/O-B state
+
+Re-mined the live corpus after the day's merges (v2.46.0 O-C hints+readGate,
+v2.47.0 O-D shim+O-E corpus, v2.48.0 turnpack ports; master `6dd4fc90`),
+with the O-B GLM-5.3 baseline now available as the frontier reference.
+
+### Finding 1 — THE CLIENT IS STALE (the round's headline)
+
+The running Droid qwen3.8 UAP client executes the **globally-installed
+`@miller-tech/uap` = v2.45.0** (NVM path, verified: the global dist lacks
+`delivery/path-hints.js`; the npm registry's `latest` is also 2.45.0 while
+master is 2.48.0). Every deliver/client session today therefore runs
+**pre-fix code** — none of O-C's hints, the readGate, or the turnpack
+listing-anchor ports are in any live client. The only fix that reached the
+client is O-D's proxy-side shim (the proxy serves from the repo tree and was
+restarted 19:14 and again 20:50 for v2.48.0). The npm-publish workflow
+(manual-trigger) was fired 09:58 UTC; until it lands and the global install
+is updated, **every measured "post-fix" window is actually shim-only.**
+
+### Finding 2 — the shim-only window (era B, n=209, directional)
+
+| Class | Era A (pre-shim, n=19,794) | Era B (shim live, n=209) |
+|---|---:|---:|
+| ok | 75.7% | **81.8%** |
+| stub-refusal | 5.4% | 7.2% |
+| edit-miss | 3.2% | **0** |
+| path-not-found | 2.6% | **0** |
+| no-op | 3.8% | 1.9% |
+
+Zero edit-miss and zero path-not-found in era B is encouraging but
+underpowered (one deliver run's task mix, n=209) — directional, not a
+finding. stub-refusal persists at 7.2% (all write_file gutting re-sends) —
+exactly the class the not-yet-client-live readGate targets, consistent with
+Finding 1.
+
+### The ranked levers now (against the measured O-B frontier reference)
+
+- **R2-A — Update the global client (free, minutes, the biggest immediate
+  lever):** once the npm publish lands, `npm i -g @miller-tech/uap@latest`
+  puts O-C + turnpack in every new client session. The next monitoring
+  window after that is the first true post-fix read; watch stub-refusal
+  (readGate) and edit-miss (turnpack) first.
+- **R2-B — Serve v31 for client sessions (free):** iq2_xs (floor tier) is
+  STILL serving all client work. The O-B suite measured what that tier
+  costs on identical tasks: same gates at ~24x tool volume and ~10x wall
+  time. This is the original O-A, still unexecuted.
+- **R2-C — Route frontier-class client sessions to glm-5.3 (one flag, now
+  measured):** `droid -m glm-5.3` / the session default (already set).
+  The O-B baseline quantifies the buy: same gates, 1/24 the tool activity,
+  1/10 the wall time, zero edit/write misses. This is the concrete
+  "Opus 5.5 / Fable 5.2 levels" lever on this box today.
+- **R2-D — Re-measure after R2-A:** the post-client-update window is the
+  honest post-fix read; era B's zeros become testable at population scale.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@miller-tech/uap",
-  "version": "2.48.0",
+  "version": "2.48.2",
   "description": "Autonomous AI agent memory system with CLAUDE.md protocol enforcement",
   "type": "module",
   "main": "dist/index.js",

package/scripts/lib/droid-sync-contracts.sh

@@ -66,6 +66,20 @@ _UAP_DROID_SYNC_CONTRACTS_LOADED=1
 # local model id — it must never carry a qwen id.
 : "${DROID_ANTHROPIC_BASE_MODEL:=claude-sonnet-4-6}"
 
+# DROID_CHAT_BASE_MODEL — the Factory BUILT-IN catalog id written into
+# `baseModelId` of the OpenAI-shaped local entries (the proxy chat /v1 route
+# and the direct Responses :8080 route). qwen3.8-max is the choice because the
+# served stack IS the Qwen3.8 family (Flash-Next 125B arms and its Edge0-145B
+# latent-MoE derivative): the entry inherits the built-in's Qwen3.8-class
+# prompts, tool-call conventions, reasoning levels, and Script-mode
+# eligibility instead of tripping the "does not map to a built-in model"
+# warning. Same family, same catalog — this is a BUILT-IN id, never a local
+# wire id: it must never be the served alias, a retired arm, or anything the
+# registry's advertises lists carry. (The prefix-era guard "never a qwen id"
+# predates the Factory catalog gaining Qwen3.8 built-ins; the exact local-id
+# check lives in the sandbox suite.)
+: "${DROID_CHAT_BASE_MODEL:=qwen3.8-max}"
+
 # DROID_SYNC_LOCK — lock file serialising concurrent Droid config writes
 # (session-start hook + manual installer run + a second Droid window).
 # Without it two writers read-modify-write the same JSON and one set of
@@ -95,15 +109,23 @@ fi
 #            in place or left alone. "0" = patch only when present.
 #   Reads from the caller's scope: MODEL_ALIAS, PROXY_URL, DIRECT_URL,
 #   CTX_SESSION, CTX_POOL, MAX_OUTPUT, DROID_ANTHROPIC_BASE_MODEL,
-#   PROXY_TOKEN (via UAP_SYNC_TOKEN env), DRY_RUN.
+#   DROID_CHAT_BASE_MODEL, PROXY_TOKEN (via UAP_SYNC_TOKEN env), DRY_RUN.
 #   Migration rules (idempotent; no change => file left byte-identical):
-#     * entries whose model is a RETIRED local id — or already the live
-#       alias, or model-less but pointed at :4000/:8080 — converge onto
-#       MODEL_ALIAS;
+#     * entries whose model is a RETIRED local id — already the live alias,
+#       a FLEET wire id (any registry advertises entry) pointed at this
+#       box's :4000/:8080, or model-less but pointed at :4000/:8080 —
+#       converge onto MODEL_ALIAS (fleet ids are gated on a local base:
+#       the compat names collide with real cloud ids, so a cloud-router
+#       entry carrying one survives untouched);
 #     * route by existing shape: baseUrl with :8080 -> direct strata route
 #       (pool-sized); :4000 WITHOUT /v1 -> Anthropic Messages root (proxy
-#       root + baseModelId + apiKey only if absent, session-sized); anything
-#       else local -> proxy chat route ($PROXY_URL/v1 + token);
+#       root + baseModelId=DROID_ANTHROPIC_BASE_MODEL + apiKey only if
+#       absent, session-sized); anything else local -> proxy chat route
+#       ($PROXY_URL/v1 + token); every camelCase local route also gains
+#       baseModelId (anthropic shape -> DROID_ANTHROPIC_BASE_MODEL, OpenAI
+#       shapes -> DROID_CHAT_BASE_MODEL) when absent, so Droid inherits the
+#       built-in's prompts/tools/reasoning instead of warning the entry
+#       "does not map to a built-in model";
 #     * cloud models (anything else, including the :8317 router entries) are
 #       deliberate user choices and are never repointed;
 #     * a settings.json with NO local entry after migration gains the