Optimize Tiny Stories end-to-end target-mode flow

.github/workflows/ci.yml

@@ -11,18 +11,18 @@ jobs:
     name: backend (pytest + import smoke)
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v7
 
       - name: Set up Python 3.11
-        uses: actions/setup-python@v5
+        uses: actions/setup-python@v6
         with:
           python-version: "3.11"
           cache: pip
 
-      - name: Set up Node 20 for backend source guards
-        uses: actions/setup-node@v4
+      - name: Set up Node 24 for backend source guards
+        uses: actions/setup-node@v7
         with:
-          node-version: "20"
+          node-version: "24"
           cache: npm
           cache-dependency-path: frontend2/package-lock.json
 
@@ -50,12 +50,12 @@ jobs:
       run:
         working-directory: frontend2
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v7
 
-      - name: Set up Node 20
-        uses: actions/setup-node@v4
+      - name: Set up Node 24
+        uses: actions/setup-node@v7
         with:
-          node-version: "20"
+          node-version: "24"
           cache: npm
           cache-dependency-path: frontend2/package-lock.json
 

artifacts/portfolio/tiny-stories-engineering-evidence-summary.json

@@ -2,6 +2,21 @@
   "schema_version": "tiny_stories_engineering_evidence_summary.v1",
   "project": "Tiny Stories",
   "framing": "Productized LLM / applied AI systems engineering, not HCI research.",
+  "current_live_evidence_anchor": {
+    "date": "2026-07-16",
+    "branch": "codex/tiny-stories-target-mode",
+    "template_id": "tmpl_159a158cc547",
+    "session_id": "sess_aafa73537725",
+    "completed_turns": 12,
+    "failure_count": 0,
+    "step_judge_passes": 12,
+    "contract_judge_passes": 12,
+    "quality_schema_version": "tiny_stories_golden_path_quality.v2",
+    "quality_status": "pass",
+    "quality_note": "Mode-aware character evidence passed with 12 responsive turns, 12 shifted turns, and three distinct NPC ids. This deterministic gate is not population-level fun or retention evidence.",
+    "visibility": "local_live_validation",
+    "public_link": false
+  },
   "historical_live_evidence_anchor": {
     "commit": "4382874 fix: keep opening live for eval gate",
     "snapshot": "snapshot/story-brief-opening-live-reliability-2026-06-08",
@@ -21,56 +36,102 @@
   },
   "live_gate": {
     "status": "pass",
-    "template_id": "tmpl_e6a132445fe8",
-    "session_id": "sess_61d85a921fda",
+    "template_id": "tmpl_159a158cc547",
+    "session_id": "sess_aafa73537725",
+    "completed_turns": 12,
     "failure_count": 0,
     "required_operations": [
       {
         "stage": "Story Butler",
         "operation": "create.story_butler_turn",
+        "call_count": 1,
         "source": "live",
         "status": "success",
-        "latency_ms": 1568,
-        "input_tokens": 2697,
+        "latency_ms": 2464,
+        "latency_ms_min": 2464,
+        "latency_ms_max": 2464,
+        "input_tokens": 2392,
         "cached_input_tokens": 384,
-        "output_tokens": 32,
-        "total_tokens": 2729,
+        "output_tokens": 39,
+        "total_tokens": 2431,
+        "retry_count_max": 0,
         "fallback": null
       },
       {
         "stage": "Story Brief",
         "operation": "narrative.story_brief",
+        "call_count": 1,
         "source": "live",
         "status": "success",
-        "latency_ms": 2479,
-        "input_tokens": 1408,
-        "cached_input_tokens": 1280,
-        "output_tokens": 156,
-        "total_tokens": 1564,
+        "latency_ms": 2063,
+        "latency_ms_min": 2063,
+        "latency_ms_max": 2063,
+        "input_tokens": 1477,
+        "cached_input_tokens": 1408,
+        "output_tokens": 148,
+        "total_tokens": 1625,
+        "retry_count_max": 0,
         "fallback": null
       },
       {
         "stage": "Opening",
         "operation": "narrative.opening",
+        "call_count": 1,
         "source": "live",
         "status": "success",
-        "latency_ms": 13100,
-        "input_tokens": 2630,
-        "cached_input_tokens": 0,
-        "output_tokens": 832,
-        "total_tokens": 3462,
+        "latency_ms": 10942,
+        "latency_ms_min": 10942,
+        "latency_ms_max": 10942,
+        "input_tokens": 2678,
+        "cached_input_tokens": 1408,
+        "output_tokens": 693,
+        "total_tokens": 3371,
+        "retry_count_max": 0,
         "fallback": null
       },
       {
-        "stage": "Play turn",
+        "stage": "Play turns",
         "operation": "narrative.advance_turn",
+        "call_count": 12,
+        "source": "live",
+        "status": "success",
+        "latency_ms": 14032,
+        "latency_ms_min": 8845,
+        "latency_ms_max": 16247,
+        "input_tokens": 98750,
+        "cached_input_tokens": 71680,
+        "output_tokens": 9264,
+        "total_tokens": 108014,
+        "retry_count_max": 0,
+        "fallback": null
+      }
+    ],
+    "completion_operations": [
+      {
+        "operation": "narrative.ending",
+        "source": "live",
+        "status": "success",
+        "latency_ms": 17471,
+        "total_tokens": 5639,
+        "retry_count": 0,
+        "fallback": null
+      },
+      {
+        "operation": "narrative.highlights",
+        "source": "live",
+        "status": "success",
+        "latency_ms": 14190,
+        "total_tokens": 6957,
+        "retry_count": 0,
+        "fallback": null
+      },
+      {
+        "operation": "narrative.branches",
         "source": "live",
         "status": "success",
-        "latency_ms": 7701,
-        "input_tokens": 6464,
-        "cached_input_tokens": 4864,
-        "output_tokens": 420,
-        "total_tokens": 6884,
+        "latency_ms": 17252,
+        "total_tokens": 7777,
+        "retry_count": 0,
         "fallback": null
       }
     ]

docs/tiny-stories-engineering-evidence-packet.md

@@ -9,6 +9,29 @@ observability, and repeatable validation gates.
 
 Framing: Productized LLM / applied AI systems engineering, not HCI research.
 
+Current live evidence anchor:
+
+On 2026-07-16, the strict live golden-path harness completed a fresh isolated
+Create -> Story Brief -> Opening -> 12-turn Play -> Ending run against the
+configured gateway. Required calls were `live/success`, no required call used
+fallback, and the persisted retry count was zero for every successful call.
+
+- Branch under review: `codex/tiny-stories-target-mode`
+- Template/session: `tmpl_159a158cc547` / `sess_aafa73537725`
+- Result: 12/12 turns and a generated ending; failure count 0
+- Reviewer evidence: 12 Step Judge passes and 12 Contract Judge passes
+- Deterministic quality packaging v2: `pass`. Consequence clarity, choice
+  diversity, escalation, mode-aware character response, brief payoff, and
+  playable options passed. In `story` mode, character quality is measured from
+  persisted NPC responses and state shifts rather than the active-agenda
+  contract reserved for `gauntlet`: 12/12 responsive turns, 12/12 shifted
+  turns, and three distinct NPC ids in this run.
+- Player-surface check: the generated ending rendered with no horizontal
+  overflow and zero browser console warnings/errors. A bounded copy scan found
+  no technical provider/debug/schema/token language; substring matches inside
+  ordinary words such as `escaped` and `traceable` were discarded as false
+  positives.
+
 Historical live evidence anchor:
 
 This is a dated 2026-06-08 live-gate snapshot kept for admissions and
@@ -101,36 +124,48 @@ The current acceptance gate is live-backed. It is not satisfied by `/health`
 alone and does not accept deterministic fallback for required product text
 paths.
 
-Live run exercised:
+The 2026-07-16 live run exercised:
 
 1. `/health` configured status for `text_llm`, `create_story_butler`,
    `story_brief`, `opening`, and `play_turns`;
 2. Story Butler guide turn;
 3. Story Brief shaping;
 4. template/opening generation;
 5. Play story fetch with reviewer trace enabled;
-6. one Play turn;
-7. reviewer-only telemetry query.
+6. twelve Play turns through the configured live text path;
+7. live ending, highlight, and branch generation;
+8. reviewer-only telemetry and judge evidence.
 
 Run result:
 
 - Status: pass
-- Template/session: `tmpl_e6a132445fe8` / `sess_61d85a921fda`
+- Template/session: `tmpl_159a158cc547` / `sess_aafa73537725`
 - Failure count: 0
-- Story Butler latency: 6216ms
-- Story Brief latency: 2487ms
-- template/opening latency: 13108ms
-- Play turn latency: 9649ms
-- telemetry endpoint latency: 3ms
+- Completed turns: 12/12
+- Step Judge / Contract Judge: 12 pass / 12 pass
+- Story Butler latency: 2464ms
+- Story Brief latency: 2063ms
+- opening latency: 10942ms
+- Play turn latency: 8845-16247ms; median 14032ms
+- Required-operation fallback count: 0
+- Maximum persisted retry count: 0
 
 Required live telemetry table:
 
-| Stage | Operation | Source | Status | Latency | Input | Cached input | Output | Total | Fallback |
-| --- | --- | --- | --- | ---: | ---: | ---: | ---: | ---: | --- |
-| Story Butler | `create.story_butler_turn` | `live` | `success` | 1568ms | 2697 | 384 | 32 | 2729 | none |
-| Story Brief | `narrative.story_brief` | `live` | `success` | 2479ms | 1408 | 1280 | 156 | 1564 | none |
-| Opening | `narrative.opening` | `live` | `success` | 13100ms | 2630 | 0 | 832 | 3462 | none |
-| Play turn | `narrative.advance_turn` | `live` | `success` | 7701ms | 6464 | 4864 | 420 | 6884 | none |
+| Stage | Operation | Calls | Source/status | Latency (min/median/max) | Input | Cached input | Output | Total | Retry max | Fallback |
+| --- | --- | ---: | --- | --- | ---: | ---: | ---: | ---: | ---: | --- |
+| Story Butler | `create.story_butler_turn` | 1 | `live/success` | 2464/2464/2464ms | 2392 | 384 | 39 | 2431 | 0 | none |
+| Story Brief | `narrative.story_brief` | 1 | `live/success` | 2063/2063/2063ms | 1477 | 1408 | 148 | 1625 | 0 | none |
+| Opening | `narrative.opening` | 1 | `live/success` | 10942/10942/10942ms | 2678 | 1408 | 693 | 3371 | 0 | none |
+| Play turns | `narrative.advance_turn` | 12 | `live/success` | 8845/14032/16247ms | 98750 | 71680 | 9264 | 108014 | 0 | none |
+| Ending | `narrative.ending` | 1 | `live/success` | 17471/17471/17471ms | 5311 | 256 | 328 | 5639 | 0 | none |
+| Highlights | `narrative.highlights` | 1 | `live/success` | 14190/14190/14190ms | 5783 | 640 | 1174 | 6957 | 0 | none |
+| Branches | `narrative.branches` | 1 | `live/success` | 17252/17252/17252ms | 7393 | 640 | 384 | 7777 | 0 | none |
+
+These are system-validation observations from one bounded canonical run, not
+population-level latency, quality, retention, or market evidence. In particular,
+the 14.0s median Play-turn latency is a real provider-bound product risk even though the run
+completed without fallback.
 
 The gate rejects these rows for the required operations:
 

frontend2/src/api/contracts.ts

@@ -802,6 +802,7 @@ export type NarrativeStoryBrief = {
   display_title?: string | null
   display_intro?: string | null
   premise_summary: string
+  player_role?: string | null
   genre_tone: string
   tension_profile: NarrativeTensionProfile
   story_kernel: string
@@ -839,10 +840,21 @@ export type NarrativeStoryBriefConsistencyCheck = {
   should_retry: boolean
 }
 
+export type NarrativeStoryBriefGuideContext = {
+  scene_summary: string
+  player_role: string
+  cast_or_factions: string[]
+  pressure: string
+  constraints: string[]
+  tone: string
+  confirmed_facts: string[]
+}
+
 export type NarrativeStoryBriefAdvisorRequest = {
   seed: string
   language?: NarrativeTemplateLanguage
   desired_tension_profile?: NarrativeTensionProfile | null
+  guide_context?: NarrativeStoryBriefGuideContext | null
 }
 
 export type NarrativeStoryBriefAdvisorResponse = {

frontend2/src/pages/create/README.md

@@ -22,9 +22,10 @@
 1. Home `Write a new story` opens `#/create` without a login/name gate.
 2. The page is a Story Butler chat, not a SaaS form.
 3. The assistant asks for missing scene-plan ingredients and redirects unsafe/out-of-spec prompts while keeping pre-plan chat natural.
-4. When enough fields are present, the scene plan is shaped automatically and appears inside the transcript.
-5. `Generate and enter story` lives with the scene plan.
-6. Generate creates a template/session and routes to Play.
+4. When enough fields are present, the compact Story Butler context is sent with the seed so the scene plan preserves confirmed roles, cast, pressure, constraints, and clues instead of rediscovering them from raw transcript text.
+5. The scene plan opens with only the playable essentials. Inferred settings, full cast planning, revisions, and handoff mechanics remain available in a closed details section.
+6. `Generate and enter story` lives with the scene plan.
+7. Generate creates a template/session and routes to Play.
 
 ## Story-Shape Settings
 
@@ -40,7 +41,7 @@ Examples handled by `shared/lib/story-guide-settings.ts`, with loop orchestratio
 - `backstage`, `disappearance`, `public scandal` -> high drama.
 - `cozy`, `clues`, `small town` -> cozy mystery.
 
-The inferred read is kept internal during ordinary collection so pre-plan chat stays one-question-at-a-time. It is surfaced inside the final scene plan before Generate. If a user correction changes story-shape settings after a scene plan is ready, the flow must require reshaping before Generate.
+The inferred read is kept internal during ordinary collection so pre-plan chat stays one-question-at-a-time. It is available inside the scene plan's closed planning details before Generate. If a user correction changes story-shape settings after a scene plan is ready, the flow must require reshaping before Generate.
 
 ## Visibility