mshegolev · mshegolev · May 15, 2026 · May 15, 2026
diff --git a/.planning/MILESTONE-AUDIT-v1.0.md b/.planning/MILESTONE-AUDIT-v1.0.md
@@ -0,0 +1,37 @@
+# Milestone Audit: v1.0 Speedly Catcher + Speedly Sayer
+
+status: passed
+
+## Scope
+
+Milestone v1.0 targeted faster, more measurable live-call trigger detection and
+prepared-response playback, plus a local sample review loop.
+
+## Requirement Coverage
+
+- CATCH-01: catch-path latency is printed from `saymo auto`.
+- CATCH-02: live tuning is configurable globally and per meeting profile.
+- CATCH-03: `saymo trigger-eval` evaluates saved samples.
+- CATCH-04: `saymo trigger-eval --promote` learns a variant and re-runs.
+- SAY-01: `saymo auto-preflight` checks readiness before a call.
+- SAY-02: auto-mode prints playback-start latency.
+- SAY-03: forced playback returns structured blocked reasons.
+- SAY-04: cache misses route to prepared standup fallback by default.
+- TRAIN-01: `saymo trigger-samples list` inspects sample metadata.
+- TRAIN-02: `saymo trigger-samples replay` reclassifies and optionally plays.
+- TRAIN-03: `saymo trigger-samples report` exports sanitized markdown.
+
+## Verification
+
+- `.venv/bin/python -m pytest -q`
+  - 241 passed
+- `git diff --check`
+  - passed
+- `.venv/bin/saymo --help`
+  - command surface includes new CLI commands
+
+## Residual Notes
+
+The evaluator becomes more valuable as real call samples accumulate under
+`~/.saymo/trigger_samples/<profile>/`.
+
diff --git a/.planning/MILESTONES.md b/.planning/MILESTONES.md
@@ -5,11 +5,18 @@
 ### v1.0 Speedly Catcher + Speedly Sayer
 
 **Started:** 2026-05-15
-**Status:** Planned
+**Completed:** 2026-05-15
+**Status:** Complete
 
 Focus: make live-call trigger detection and prepared-response playback faster,
 measurable, and easier to tune from captured meeting samples.
 
+Delivered:
+- Live latency metrics and configurable catcher tuning.
+- Offline trigger sample evaluation with fuzzy variant promotion.
+- Auto preflight diagnostics and structured playback blocked reasons.
+- Trigger sample list/replay/report commands.
+
 ## Historical Baseline
 
 Before GSD planning was introduced, Saymo already had a working local voice

diff --git a/.planning/PROJECT.md b/.planning/PROJECT.md
@@ -18,6 +18,8 @@ enough that the call still feels live.
 **Goal:** Make live-call detection and prepared-response playback faster,
 measurable, and easier to tune from real meeting samples.
 
+**Status:** Complete on 2026-05-15.
+
 **Target features:**
 - Speedly Catcher: lower-latency trigger/addressing detection with profile
   tuning and offline sample evaluation.
@@ -35,12 +37,9 @@ measurable, and easier to tune from real meeting samples.
 - ✓ Manual hotkeys and takeover checks exist before this milestone.
 - ✓ `saymo trigger-capture` can save classified call windows as WAV plus JSON
   metadata before this milestone.
-
-### Active
-
-- [ ] Faster catch path is measurable and tunable per profile.
-- [ ] Faster say path is preflighted and measured before calls.
-- [ ] Captured trigger samples feed a repeatable tuning/evaluation workflow.
+- ✓ Faster catch path is measurable and tunable per profile.
+- ✓ Faster say path is preflighted and measured before calls.
+- ✓ Captured trigger samples feed a repeatable tuning/evaluation workflow.
 
 ### Out of Scope
 
@@ -77,8 +76,8 @@ measurable, and easier to tune from real meeting samples.
 | Decision | Rationale | Outcome |
 |----------|-----------|---------|
 | Keep live-call listening local-first | Avoid leaking meeting audio and preserve offline operation | ✓ Good |
-| Use classified trigger samples before deeper ML changes | Real misses/false positives are the best tuning data | — Pending |
-| Separate catch latency from say latency | Different bottlenecks require different measurements and tests | — Pending |
+| Use classified trigger samples before deeper ML changes | Real misses/false positives are the best tuning data | ✓ Implemented via `trigger-eval` |
+| Separate catch latency from say latency | Different bottlenecks require different measurements and tests | ✓ Implemented in `saymo auto` diagnostics |
 
 ---
-*Last updated: 2026-05-15 after starting milestone v1.0 Speedly Catcher + Speedly Sayer*
+*Last updated: 2026-05-15 after completing milestone v1.0 Speedly Catcher + Speedly Sayer*
diff --git a/.planning/REQUIREMENTS.md b/.planning/REQUIREMENTS.md
@@ -10,35 +10,35 @@ Requirements for milestone v1.0 Speedly Catcher + Speedly Sayer.
 
 ### Speedly Catcher
 
-- [ ] **CATCH-01**: User can see catch-path latency for each live window:
+- [x] **CATCH-01**: User can see catch-path latency for each live window:
   capture, transcription, trigger match, addressing decision, and final action.
-- [ ] **CATCH-02**: User can tune capture window length, overlap, cooldown, and
+- [x] **CATCH-02**: User can tune capture window length, overlap, cooldown, and
   silence handling per meeting profile without editing source code.
-- [ ] **CATCH-03**: User can run offline trigger evaluation against saved call
+- [x] **CATCH-03**: User can run offline trigger evaluation against saved call
   samples and see counts for `asked_to_speak`, `question`, `speech`, false
   positives, and misses.
-- [ ] **CATCH-04**: User can promote a heard trigger variant from an evaluated
+- [x] **CATCH-04**: User can promote a heard trigger variant from an evaluated
   sample into `vocabulary.fuzzy_expansions` and re-run the evaluation.
 
 ### Speedly Sayer
 
-- [ ] **SAY-01**: User can run one preflight command before a call that verifies
+- [x] **SAY-01**: User can run one preflight command before a call that verifies
   response cache coverage, output device routing, provider readiness, and
   configured profile triggers.
-- [ ] **SAY-02**: User can see say-path latency from trigger/hotkey to playback
+- [x] **SAY-02**: User can see say-path latency from trigger/hotkey to playback
   start for cached responses.
-- [ ] **SAY-03**: User can force prepared playback without waiting for trigger
+- [x] **SAY-03**: User can force prepared playback without waiting for trigger
   detection and receive a clear reason if playback is blocked.
-- [ ] **SAY-04**: User can keep auto-mode responsive when response-cache lookup
+- [x] **SAY-04**: User can keep auto-mode responsive when response-cache lookup
   misses by using a documented prepared-response fallback.
 
 ### Training Loop
 
-- [ ] **TRAIN-01**: User can list captured sample folders and inspect metadata
+- [x] **TRAIN-01**: User can list captured sample folders and inspect metadata
   without opening raw JSON files manually.
-- [ ] **TRAIN-02**: User can replay selected captured samples through a local
+- [x] **TRAIN-02**: User can replay selected captured samples through a local
   command to validate transcripts and classifications.
-- [ ] **TRAIN-03**: User can export a sanitized evaluation report that excludes
+- [x] **TRAIN-03**: User can export a sanitized evaluation report that excludes
   raw audio and private config values.
 
 ## v2 Requirements
@@ -67,17 +67,17 @@ Deferred to a later milestone.
 
 | Requirement | Phase | Status |
 |-------------|-------|--------|
-| CATCH-01 | Phase 1 | Pending |
-| SAY-02 | Phase 1 | Pending |
-| CATCH-02 | Phase 2 | Pending |
-| CATCH-03 | Phase 2 | Pending |
-| CATCH-04 | Phase 2 | Pending |
-| SAY-01 | Phase 3 | Pending |
-| SAY-03 | Phase 3 | Pending |
-| SAY-04 | Phase 3 | Pending |
-| TRAIN-01 | Phase 4 | Pending |
-| TRAIN-02 | Phase 4 | Pending |
-| TRAIN-03 | Phase 4 | Pending |
+| CATCH-01 | Phase 1 | Complete |
+| SAY-02 | Phase 1 | Complete |
+| CATCH-02 | Phase 2 | Complete |
+| CATCH-03 | Phase 2 | Complete |
+| CATCH-04 | Phase 2 | Complete |
+| SAY-01 | Phase 3 | Complete |
+| SAY-03 | Phase 3 | Complete |
+| SAY-04 | Phase 3 | Complete |
+| TRAIN-01 | Phase 4 | Complete |
+| TRAIN-02 | Phase 4 | Complete |
+| TRAIN-03 | Phase 4 | Complete |
 
 **Coverage:**
 - v1 requirements: 11 total
@@ -86,4 +86,4 @@ Deferred to a later milestone.
 
 ---
 *Requirements defined: 2026-05-15*
-*Last updated: 2026-05-15 after creating milestone v1.0 roadmap*
+*Last updated: 2026-05-15 after completing milestone v1.0*
diff --git a/.planning/ROADMAP.md b/.planning/ROADMAP.md
@@ -9,21 +9,21 @@ wraps the sample workflow in operator-facing commands and docs.
 
 ## Milestones
 
-- 📋 **v1.0 Speedly Catcher + Speedly Sayer** - Phases 1-4 (planned)
+- ✅ **v1.0 Speedly Catcher + Speedly Sayer** - Phases 1-4 (complete)
 
 ## Phases
 
 **Phase Numbering:**
 - Integer phases (1, 2, 3): Planned milestone work
 - Decimal phases (2.1, 2.2): Urgent insertions
 
-- [ ] **Phase 1: Latency Baseline** - Measure catch and say latency without
+- [x] **Phase 1: Latency Baseline** - Measure catch and say latency without
   changing behavior.
-- [ ] **Phase 2: Catcher Tuning Loop** - Add profile tuning and offline sample
+- [x] **Phase 2: Catcher Tuning Loop** - Add profile tuning and offline sample
   evaluation.
-- [ ] **Phase 3: Sayer Preflight Path** - Verify readiness and speed up
+- [x] **Phase 3: Sayer Preflight Path** - Verify readiness and speed up
   prepared playback decisions.
-- [ ] **Phase 4: Sample Review Workflow** - Make captured samples inspectable,
+- [x] **Phase 4: Sample Review Workflow** - Make captured samples inspectable,
   replayable, and reportable.
 
 ## Phase Details
@@ -41,8 +41,8 @@ future speed work is measured, not guessed.
 **Plans**: 2 plans
 
 Plans:
-- [ ] 01-01: Instrument catch-path timings
-- [ ] 01-02: Instrument say-path playback timings
+- [x] 01-01: Instrument catch-path timings
+- [x] 01-02: Instrument say-path playback timings
 
 ### Phase 2: Catcher Tuning Loop
 **Goal**: Let the user tune live detection from config and validate changes
@@ -59,9 +59,9 @@ against saved call samples.
 **Plans**: 3 plans
 
 Plans:
-- [ ] 02-01: Add profile-level catcher tuning config
-- [ ] 02-02: Build offline trigger-sample evaluator
-- [ ] 02-03: Add promote-and-rerun workflow for fuzzy variants
+- [x] 02-01: Add profile-level catcher tuning config
+- [x] 02-02: Build offline trigger-sample evaluator
+- [x] 02-03: Add promote-and-rerun workflow for fuzzy variants
 
 ### Phase 3: Sayer Preflight Path
 **Goal**: Make prepared-response playback readiness explicit before calls and
@@ -77,9 +77,9 @@ keep auto-mode responsive when playback cannot start.
 **Plans**: 3 plans
 
 Plans:
-- [ ] 03-01: Add live-call preflight command
-- [ ] 03-02: Normalize forced-playback blocked reasons
-- [ ] 03-03: Harden cache-miss fallback behavior
+- [x] 03-01: Add live-call preflight command
+- [x] 03-02: Normalize forced-playback blocked reasons
+- [x] 03-03: Harden cache-miss fallback behavior
 
 ### Phase 4: Sample Review Workflow
 **Goal**: Turn captured call windows into a practical local review loop for
@@ -96,9 +96,9 @@ training and debugging trigger behavior.
 **Plans**: 3 plans
 
 Plans:
-- [ ] 04-01: Add sample listing command
-- [ ] 04-02: Add sample replay and reclassify command
-- [ ] 04-03: Add sanitized evaluation report and docs
+- [x] 04-01: Add sample listing command
+- [x] 04-02: Add sample replay and reclassify command
+- [x] 04-03: Add sanitized evaluation report and docs
 
 ## Progress
 
@@ -107,7 +107,7 @@ Phases execute in numeric order: 1 → 2 → 3 → 4
 
 | Phase | Milestone | Plans Complete | Status | Completed |
 |-------|-----------|----------------|--------|-----------|
-| 1. Latency Baseline | v1.0 | 0/2 | Not started | - |
-| 2. Catcher Tuning Loop | v1.0 | 0/3 | Not started | - |
-| 3. Sayer Preflight Path | v1.0 | 0/3 | Not started | - |
-| 4. Sample Review Workflow | v1.0 | 0/3 | Not started | - |
+| 1. Latency Baseline | v1.0 | 2/2 | Complete | 2026-05-15 |
+| 2. Catcher Tuning Loop | v1.0 | 3/3 | Complete | 2026-05-15 |
+| 3. Sayer Preflight Path | v1.0 | 3/3 | Complete | 2026-05-15 |
+| 4. Sample Review Workflow | v1.0 | 3/3 | Complete | 2026-05-15 |
diff --git a/.planning/STATE.md b/.planning/STATE.md
@@ -6,33 +6,36 @@ See: .planning/PROJECT.md (updated 2026-05-15)
 
 **Core value:** Saymo must reliably catch when the user is expected to answer
 and respond fast enough that the call still feels live.
-**Current focus:** Phase 1: Latency Baseline
+**Current focus:** Milestone v1.0 complete; ready for next milestone planning
 
 ## Current Position
 
-Phase: 1 of 4 (Latency Baseline)
-Plan: —
-Status: Ready to plan
-Last activity: 2026-05-15 — Milestone v1.0 Speedly Catcher + Speedly Sayer initialized
+Phase: 4 of 4 (Sample Review Workflow)
+Plan: 04-03 complete
+Status: Milestone complete
+Last activity: 2026-05-15 — Completed Speedly Catcher + Speedly Sayer autonomous execution
 
-Progress: [░░░░░░░░░░] 0%
+Progress: [██████████] 100%
 
 ## Performance Metrics
 
 **Velocity:**
-- Total plans completed: 0
-- Average duration: —
-- Total execution time: 0.0 hours
+- Total plans completed: 11
+- Average duration: autonomous batch
+- Total execution time: one autonomous session
 
 **By Phase:**
 
 | Phase | Plans | Total | Avg/Plan |
 |-------|-------|-------|----------|
-| - | - | - | - |
+| 1. Latency Baseline | 2 | 2/2 | autonomous |
+| 2. Catcher Tuning Loop | 3 | 3/3 | autonomous |
+| 3. Sayer Preflight Path | 3 | 3/3 | autonomous |
+| 4. Sample Review Workflow | 3 | 3/3 | autonomous |
 
 **Recent Trend:**
-- Last 5 plans: —
-- Trend: —
+- Last 5 plans: 03-02, 03-03, 04-01, 04-02, 04-03
+- Trend: complete
 
 ## Accumulated Context
 
@@ -45,6 +48,10 @@ Recent decisions affecting current work:
   quality regressions can be isolated.
 - v1.0: Use captured call windows as the tuning substrate before adding any
   custom classifier.
+- v1.0: Keep the sample report sanitized by omitting transcript text and raw
+  audio payloads while still listing sample basenames and classification flags.
+- v1.0: Treat response cache coverage as a preflight warning, not a hard block,
+  because prepared standup fallback remains valid.
 
 ### Pending Todos
 
@@ -53,10 +60,11 @@ None yet.
 ### Blockers/Concerns
 
 - Need enough real captured samples in `~/.saymo/trigger_samples/` to make
-  offline evaluation meaningful.
+  offline evaluation statistically meaningful.
 
 ## Session Continuity
 
 Last session: 2026-05-15
-Stopped at: Milestone initialized; next step is `$gsd-plan-phase 1`.
+Stopped at: Milestone v1.0 complete; next step is `$gsd-new-milestone` if more
+work is needed.
 Resume file: None
diff --git a/.planning/phases/01-latency-baseline/01-01-PLAN.md b/.planning/phases/01-latency-baseline/01-01-PLAN.md
@@ -0,0 +1,21 @@
+# Plan 01-01: Instrument Catch-Path Timings
+
+## Requirements
+
+- CATCH-01: Show live-window latency split by capture, transcription, trigger
+  match, addressing, and action.
+
+## Tasks
+
+- Replace hardcoded capture/read/cooldown/silence values with resolved live
+  tuning values.
+- Measure capture read, Whisper transcription, and trigger match per processed
+  window.
+- Measure addressing and confirmation/action decisions when a trigger fires.
+- Keep trigger, addressing, and confirmation semantics unchanged.
+
+## Verification
+
+- Focused tests for live tuning config.
+- Full pytest to guard existing trigger behavior.
+
diff --git a/.planning/phases/01-latency-baseline/01-01-SUMMARY.md b/.planning/phases/01-latency-baseline/01-01-SUMMARY.md
@@ -0,0 +1,16 @@
+# Summary 01-01: Instrument Catch-Path Timings
+
+## Completed
+
+- Added `LiveConfig` and `resolve_live_tuning()` in `saymo/config.py`.
+- Updated `saymo auto` to use live tuning for chunk length, overlap, read
+  timeout, cooldown, silence threshold, and pre-speak delay.
+- Added catch latency output for capture, STT, trigger match, addressing, and
+  action decisions.
+
+## Files
+
+- `saymo/config.py`
+- `saymo/commands/core.py`
+- `tests/test_live_tuning_config.py`
+
diff --git a/.planning/phases/01-latency-baseline/01-02-PLAN.md b/.planning/phases/01-latency-baseline/01-02-PLAN.md
@@ -0,0 +1,18 @@
+# Plan 01-02: Instrument Say-Path Playback Timings
+
+## Requirements
+
+- SAY-02: Show time from trigger/hotkey to playback start for cached responses.
+
+## Tasks
+
+- Add an auto-response decision object with source and reason metadata.
+- Add a playback result object with success, blocked reason, and start state.
+- Allow `_play_cached_audio` to call a playback-start callback.
+- Print response routing and playback-start latency from `saymo auto`.
+
+## Verification
+
+- Existing Q&A response-resolution tests remain compatible.
+- New tests cover decision metadata and blocked playback reason.
+