chore: close sprint 3, open sprint 4 with branching strategy

- Create sprint_2026_04.md with 4 priorities: branching/versioning,
run-time settings redesign, data narrative layer, 10k row test
- Add branching strategy to CLAUDE.md (main/develop/hotfix model)
- Update sprint pointer and history table in CLAUDE.md

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

CLAUDE.md

@@ -70,10 +70,56 @@ The user may be running a live demo on HF. Pushing without permission can break
 
 ## Current Sprint
 
-**ALWAYS READ FIRST**: `sprint_2026_03.md` (in root)
+**ALWAYS READ FIRST**: `sprint_2026_04.md` (in root)
 
 For overall project status, see `PROJECT_STATUS.md` in root.
 
+---
+
+## Branching Strategy
+
+**Three-branch model: `main` / `develop` / hotfix**
+
+| Branch | Purpose | Deploys to |
+|--------|---------|------------|
+| `main` | Production — always stable and shippable | `hf` (prod) |
+| `develop` | Sprint integration — current sprint work | `hf-test` |
+| `feature/*` | Individual features — branch off `develop` | local / `hf-test` |
+
+### Rules
+
+**Normal sprint work:**
+```bash
+git checkout develop
+git checkout -b feature/my-thing
+# ... work ...
+git checkout develop && git merge feature/my-thing
+git push hf-test develop:main   # deploy to test
+```
+
+**Shipping to prod (end of sprint or stable milestone):**
+```bash
+git checkout main && git merge develop
+git push hf main                # deploy to prod
+git push origin main
+git tag v1.x.x && git push origin --tags
+```
+
+**Hotfix to prod (urgent fix while sprint work is in develop):**
+```bash
+git checkout main
+git checkout -b hotfix/description
+# ... fix ...
+git checkout main && git merge hotfix/description
+git push hf main                # deploy to prod immediately
+git checkout develop && git merge hotfix/description   # keep develop in sync
+git push hf-test develop:main
+git push origin main develop
+```
+
+> **Never push sprint/feature work directly to `main`.  
+> Never push to `hf` (prod) from anything other than `main`.**
+
 ### Sprint Doc Workflow - IMPORTANT
 
 **This is how we stay organized. Follow this process consistently:**
@@ -375,7 +421,8 @@ All docs live in `dev_notes/` unless noted. When in doubt about how something wo
 ### Sprint History
 | Doc | Status |
 |-----|--------|
-| `sprint_2026_03.md` *(root)* | **Current sprint** — read this at the start of every session |
+| `sprint_2026_04.md` *(root)* | **Current sprint** — read this at the start of every session |
+| `sprint_2026_03.md` *(root)* | Closed — March/April |
 | `sprint_2026_02.md` *(root)* | Closed — February/March |
 | `sprint_2026_01.md` *(root)* | Closed — January |
 
@@ -384,5 +431,5 @@ All docs live in `dev_notes/` unless noted. When in doubt about how something wo
 
 ---
 
-*Last Updated: April 28, 2026*
+*Last Updated: April 29, 2026*
 *This is the source of truth - update rules here, not in .cursorrules*

sprint_2026_04.md

@@ -0,0 +1,144 @@
+# Sprint: May 2026
+
+*Started: April 29, 2026*
+*Previous sprint: `sprint_2026_03.md` (closed — March/April)*
+
+---
+
+## Context
+
+App is live in production. Core pipeline is stable (6–7/8 tests passing consistently).
+This sprint focuses on developer workflow, UX improvements, and data quality before
+expanding the user base further.
+
+---
+
+## Branch Strategy
+
+- **`main`** — always reflects production (`hf` remote). Only merge here when ready to ship.
+- **`develop`** — integration branch for this sprint. Deploys to `hf-test`.
+- **Feature branches** — branch off `develop`, merge back to `develop` when done.
+- **Hotfixes** — branch off `main`, fix, merge to `main` AND `develop`, push to `hf`.
+
+> See CLAUDE.md "Branching Strategy" section for full workflow rules.
+
+---
+
+## Sprint Priorities
+
+### Priority 1 — Branching / Release Strategy ✅ (set up at sprint start)
+
+- [x] Define `main` / `develop` / hotfix workflow — documented in CLAUDE.md
+- [ ] Create `develop` branch from current `main`
+- [ ] Update CI/push targets: sprint work → `hf-test` via `develop`; prod fixes → `hf` via `main`
+- [ ] Add git tag for current prod release (`v1.0.0` or similar) before new sprint work begins
+
+---
+
+### Priority 2 — Run-Time Settings Redesign
+
+**Goal:** All pipeline control knobs are live on the app screen. No save/reload cycle.
+Setting a value takes effect for the next run immediately. Defaults come from Supabase
+admin settings, but can be overridden at runtime without touching the database.
+
+**Right panel layout:**
+```
+┌─────────────────────────────┐
+│  TS Environment  [dropdown] │  ← always visible
+│  AI Model        [dropdown] │  ← always visible
+├─────────────────────────────┤
+│  ▶ Advanced Settings        │  ← accordion, collapsed by default
+│    Liveboard Name  [text]   │
+│    Data Size       [dropdown] (Small / Medium / Standard / Large)
+│    Geo Scope       [dropdown] (USA Only / International)
+│    [other knobs as added]   │
+└─────────────────────────────┘
+│  Pipeline Status / Progress │  ← below settings
+│  [stage indicators]         │
+└─────────────────────────────┘
+```
+
+**Rules:**
+- Values shown are loaded from user's saved settings (Supabase) at page load
+- Changing a value in the panel takes effect immediately for the current session — no Save button needed
+- "Save as default" option (optional, later) — writes back to Supabase
+- Eliminates the separate Settings tab for pipeline-control settings (those move here)
+- Settings tab becomes admin/account settings only (TS credentials, user preferences)
+
+**Tasks:**
+- [ ] Audit all Settings tab fields — categorize as (a) run-time knobs or (b) admin/account
+- [ ] Move run-time knobs to right panel with accordion
+- [ ] Wire values to controller at run time (not just at session init)
+- [ ] Remove save/reload dependency for pipeline settings
+- [ ] Update `fact_table_size`, `dim_table_size`, `geo_scope`, `liveboard_name` to read from panel state
+
+---
+
+### Priority 3 — Data Narrative Layer
+
+**Goal:** Generated data tells the right story. Numbers reflect KPI targets, business context,
+and the outlier patterns the liveboard is designed to highlight.
+
+**Problem:** The data generator knows column names and types but not what story the data
+should tell. KPI targets, growth trends, and anomaly patterns live in the matrix
+(`uc_config`) but never reach `populate_demo_data`.
+
+**Approach:**
+- [ ] Research: audit what's in `uc_config` that's relevant to data generation (KPI definitions, story questions, outlier targets)
+- [ ] Design: define a `DataNarrativeSpec` — per-column target ranges, trend direction, key outlier rows
+- [ ] Pass `DataNarrativeSpec` from matrix config into `populate_demo_data` / LegitData
+- [ ] Generator uses spec to set realistic value ranges for numeric columns (revenue, churn, margin, etc.)
+- [ ] Outlier injection: a small number of rows (3–5%) hit the "interesting" values the liveboard questions are designed to surface
+
+---
+
+### Priority 4 — 10k Row Test
+
+**Goal:** Confirm the "large" preset (10k fact rows) works end-to-end and produces better liveboard quality.
+
+**Known issues to solve in test:**
+- Settings cached at controller init — test must change setting → trigger fresh session → run pipeline
+- 10k runs will be ~3–4x slower — bump test timeout for large-preset tests
+- Validate that `is_fact_table` name-based detection correctly assigns 10k rows (not 500)
+
+**Tasks:**
+- [ ] Add `fact_table_size=10000` variant to test suite (separate test case or flag)
+- [ ] Implement session reset between setting change and run in test harness
+- [ ] Increase timeout for large-preset test cases
+- [ ] Compare data/liveboard quality scores: 1k vs 10k on same company/use case
+
+---
+
+## Next Priorities (from status report "Working On Next")
+
+- [ ] **Async pipeline** — data population + TS model creation currently sequential; run in parallel to cut total pipeline time
+- [ ] **Enablement** — HTML user guide (full page on HF doc space, not just the markdown draft); demo videos
+- [ ] **Matrix field input** — story definitions need field review; liveboard makeup and key outlier questions per vertical
+
+---
+
+## Backlog (leftover from Sprint 3, lower priority)
+
+- [ ] Unified Outlier System — quality not there yet
+- [ ] Demo Pack Generation
+- [ ] Chart titles / viz naming
+- [ ] Tag assignment to models (returns 404; works for tables)
+- [ ] Fix Research Cache Not Loading
+- [ ] Fix DAYSONHAND Generation (currently random, needs business logic)
+- [ ] MCP 500 retry logic
+- [ ] Model Generator: Chasm Trap Fix
+- [ ] Dead code cleanup (model TML generators)
+- [ ] Incorporate Snowflake row counts into quality test data grading score
+- [ ] AI Feedback verbosity
+- [ ] `viz_type` enforcement (matrix chart type hints → override TS defaults)
+- [ ] Actual data outliers → liveboard (`data_outliers` param ready, needs connection)
+
+---
+
+## Done
+
+*(items move here as completed)*
+
+---
+
+*Last updated: April 29, 2026*