blendto · Chinmay-KB · Jul 29, 2026 · Jul 29, 2026 · Jul 29, 2026
diff --git a/docs/README.md b/docs/README.md
@@ -28,7 +28,7 @@ verified in their own repositories.
 
 ## Current compatibility
 
-- package version: `0.4.12`;
+- package version: `0.4.13`;
 - session JSON schema: `7`;
 - fingerprint schema: `6`;
 - minimum Dart SDK: `3.9.2`;

diff --git a/docs/integration/production-replay-acceptance-0.4.13.md b/docs/integration/production-replay-acceptance-0.4.13.md
@@ -0,0 +1,217 @@
+# Production replay acceptance: 0.4.12 → 0.4.13
+
+Use this after shipping SDK **0.4.13** (and collector build passthrough) and
+running Blend through a similar flow to the baseline session.
+
+## Baseline (locked)
+
+| Field | Value |
+| --- | --- |
+| Session | `session-1785142623932166` |
+| SDK | `0.4.12` |
+| App | `to.blend.mobile_app` |
+| Build | `3.17.177+1472` |
+
+| Metric | Baseline |
+| --- | ---: |
+| Raw `tap` | 86 |
+| Swipe-consumed taps | 63 |
+| Settled taps | 23 |
+| Truly orphaned taps | 0 |
+| `tap_settled` `result=unknown` | 6 |
+| unknown `superseded_route_epoch` | 2 |
+| `missing_frame` with zero local/normalized | 1 |
+| `capture_diagnostic` `missing_context_graph_build_identity` | 37 / 37 |
+
+## Manual run checklist
+
+1. Install / point Blend at tugboat **0.4.13** (includes same-turn claims,
+   deferred taps, session-end pointer fence, duplicate-down coalesce).
+2. Confirm collector with event `build` passthrough is deployed (Gate 8).
+3. Drive: scroll/flick on home, Get Pro / paywall, StageIt / sheet / chooser,
+   one rapid double-tap during settle.
+4. Paste the new `sessionId` (+ Blend build) in chat for scoring.
+
+## Hard gates (all must PASS)
+
+1. **Identity** — `metadata.sdkVersion = '0.4.13'` on event groups.
+2. **No phantom taps** — `swipe_consumed_taps = 0`.
+3. **Settle coverage** — `settled / raw_taps >= 0.95`.
+4. **No same-position bursts** — no UTC second with `tap_count >= 5` and
+   `distinct_positions = 1`.
+5. **Unknown settles** — `unknown / tap_settled <= 0.10` **and**
+   `superseded_route_epoch` count = 0.
+6. **After frames** — 100% of `navigated`/`changed` settles have `afterFrame`.
+7. **Missing-frame geometry** — any `missing_frame` tap has
+   `normalizedX/Y ∈ [0,1]` and `boundaryWidth/Height > 0`.
+8. **Diagnostics identity** — zero
+   `missing_context_graph_build_identity` on `capture_diagnostic`
+   (BLOCKED if collector not deployed).
+9. **Swipe start geometry** — every `swipe` has `payload.startCaptureCoordinate`.
+
+Overall verdict: **ACCEPT** only if every gate PASSes; otherwise **REJECT** or
+**BLOCKED** with the first failing gate id.
+
+## ClickHouse queries
+
+Replace `{newSession}` with the new session id. Service: pmkit ClickHouse.
+
+### Gate 1 — SDK version
+
+```sql
+SELECT
+  argMax(metadata.sdkVersion::Nullable(String), receivedAt) AS sdkVersion,
+  count() AS rows
+FROM pmkit.raw_events
+WHERE sessionId = {newSession}
+GROUP BY eventType
+ORDER BY eventType
+```
+
+### Gates 2–3 — tap fate
+
+```sql
+WITH events AS (
+  SELECT id,
+    argMax(eventType, receivedAt) AS eventType,
+    argMax(metadata.relatedEventId::Nullable(String), receivedAt) AS related
+  FROM pmkit.raw_events
+  WHERE sessionId = {newSession}
+  GROUP BY id
+),
+taps AS (SELECT id FROM events WHERE eventType = 'tap'),
+swipeRefs AS (SELECT related FROM events WHERE eventType = 'swipe' AND related IS NOT NULL),
+settleRefs AS (SELECT related FROM events WHERE eventType = 'tap_settled' AND related IS NOT NULL)
+SELECT
+  count() AS totalTaps,
+  countIf(id IN (SELECT related FROM swipeRefs)) AS consumedBySwipe,
+  countIf(id IN (SELECT related FROM settleRefs)) AS settled,
+  countIf(
+    id NOT IN (SELECT related FROM swipeRefs)
+    AND id NOT IN (SELECT related FROM settleRefs)
+  ) AS orphaned
+FROM taps
+```
+
+### Gate 4 — bursts
+
+```sql
+WITH taps AS (
+  SELECT
+    argMax(triggeredAt, receivedAt) AS triggeredAt,
+    argMax(payload.x::Nullable(Float64), receivedAt) AS x,
+    argMax(payload.y::Nullable(Float64), receivedAt) AS y
+  FROM pmkit.raw_events
+  WHERE sessionId = {newSession} AND eventType = 'tap'
+  GROUP BY id
+)
+SELECT
+  toStartOfSecond(triggeredAt) AS sec,
+  count() AS tapCount,
+  uniqExact((round(x, 1), round(y, 1))) AS distinctPositions
+FROM taps
+GROUP BY sec
+HAVING tapCount >= 5 AND distinctPositions = 1
+ORDER BY sec
+```
+
+### Gate 5 — unknown settles
+
+```sql
+WITH settles AS (
+  SELECT
+    argMax(result, receivedAt) AS result,
+    argMax(toJSONString(payload), receivedAt) AS payloadJson
+  FROM pmkit.raw_events
+  WHERE sessionId = {newSession} AND eventType = 'tap_settled'
+  GROUP BY id
+)
+SELECT
+  count() AS totalSettled,
+  countIf(result = 'unknown') AS unknownSettles,
+  countIf(
+    JSONExtractString(payloadJson, 'settleObservation', 'captureFailure')
+      = 'superseded_route_epoch'
+  ) AS supersededRouteEpoch
+FROM settles
+```
+
+### Gate 6 — after frames on navigated/changed
+
+```sql
+WITH settles AS (
+  SELECT
+    argMax(result, receivedAt) AS result,
+    argMax(afterFrame, receivedAt) AS afterFrame
+  FROM pmkit.raw_events
+  WHERE sessionId = {newSession} AND eventType = 'tap_settled'
+  GROUP BY id
+)
+SELECT
+  countIf(result IN ('navigated', 'changed')) AS outcomeRows,
+  countIf(result IN ('navigated', 'changed') AND afterFrame IS NULL) AS missingAfter
+FROM settles
+```
+
+### Gate 7 — missing_frame geometry
+
+```sql
+WITH taps AS (
+  SELECT argMax(toJSONString(payload), receivedAt) AS payloadJson
+  FROM pmkit.raw_events
+  WHERE sessionId = {newSession} AND eventType = 'tap'
+  GROUP BY id
+)
+SELECT
+  JSONExtractString(payloadJson, 'captureCoordinate', 'unavailableReason') AS reason,
+  JSONExtractFloat(payloadJson, 'captureCoordinate', 'normalizedX') AS nx,
+  JSONExtractFloat(payloadJson, 'captureCoordinate', 'normalizedY') AS ny,
+  JSONExtractFloat(payloadJson, 'captureCoordinate', 'boundaryWidth') AS bw,
+  JSONExtractFloat(payloadJson, 'captureCoordinate', 'boundaryHeight') AS bh
+FROM taps
+WHERE JSONExtractString(payloadJson, 'captureCoordinate', 'unavailableReason')
+  = 'missing_frame'
+```
+
+### Gate 8 — diagnostics enrichment
+
+```sql
+WITH diags AS (
+  SELECT argMax(toJSONString(payload), receivedAt) AS payloadJson
+  FROM pmkit.raw_events
+  WHERE sessionId = {newSession} AND eventType = 'capture_diagnostic'
+  GROUP BY id
+)
+SELECT
+  count() AS diagnostics,
+  countIf(
+    JSONExtractString(payloadJson, 'contextEnrichment', 'reason')
+      = 'missing_context_graph_build_identity'
+  ) AS missingBuildIdentity
+FROM diags
+```
+
+### Gate 9 — swipe startCaptureCoordinate
+
+```sql
+WITH swipes AS (
+  SELECT argMax(toJSONString(payload), receivedAt) AS payloadJson
+  FROM pmkit.raw_events
+  WHERE sessionId = {newSession} AND eventType = 'swipe'
+  GROUP BY id
+)
+SELECT
+  count() AS swipes,
+  countIf(JSONHas(payloadJson, 'startCaptureCoordinate')) AS withStartCoord
+FROM swipes
+```
+
+## Deliverable
+
+After scoring, write
+`docs/integration/production-replay-compare-0.4.12-vs-0.4.13.md` with:
+
+- session ids + SDK versions
+- side-by-side metric table
+- per-gate PASS/FAIL with deciding counts
+- single overall `ACCEPT` / `REJECT` / `BLOCKED`
diff --git a/docs/integration/production-replay-acceptance-0.4.15.md b/docs/integration/production-replay-acceptance-0.4.15.md
@@ -0,0 +1,65 @@
+# Production replay acceptance: interaction consolidation (0.4.15)
+
+Use this after shipping SDK **0.4.15** (canonical interactions + delayed
+reconciliation) and running Blend through the acceptance flow.
+
+## Baseline (locked)
+
+Prefer the nearest prior Blend session against SDK **0.4.12 / 0.4.13** for
+side-by-side scoring. Record the new session id and Blend build before scoring.
+
+## What changed in the SDK
+
+| Concern | 0.4.13 behavior | 0.4.15 behavior |
+| --- | --- | --- |
+| Gesture identity | `tap` + `tap_settled` peers | one `interaction` (`stream: semantic`) + legacy projection |
+| Claim window | microtask same-turn only | default 1,250 ms delayed reconciliation |
+| Diagnostics | mixed into normal events | `stream: diagnostic` |
+| Origin | frozen on pending tap | immutable `InteractionOrigin` on the transaction |
+| Swipe state | refreshed at pointer-up | frozen to pointer-down origin |
+
+## Manual run checklist
+
+1. Point Blend at tugboat **0.4.15**.
+2. Drive: home scroll/flick, Get Pro / paywall, full-screen navigation, modal
+   bottom sheet, asynchronous onboarding transition, rapid double-tap,
+   automatic redirect after a settled tap.
+3. Paste the new `sessionId` (+ Blend build) for scoring.
+
+## Hard gates (all must PASS)
+
+1. **Identity** — `metadata.sdkVersion = '0.4.15'`.
+2. **Canonical coverage** — one `stream: semantic` `interaction` per completed
+   user gesture (tap / swipe / scroll / cancelled).
+3. **Origin correctness** — interaction `origin.route` / `origin.targetAnchor`
+   match the pointer-down screen/component, never the destination.
+4. **Delayed attribution** — delayed navigation / bottom sheet inside 1,250 ms
+   has `attribution.kind = delayed_likely` (or `direct`) and
+   `result.status = navigated|changed`, with matching
+   `route_change.causedByInteractionId`.
+5. **Automatic false-claim rate** — timer/auth redirects after the window, and
+   routes with competing pointers, stay `navigationOrigin =
+   automatic_or_unknown`.
+6. **No semantic tap for scrolls/swipes** — completed scroll/swipe produces no
+   `stream: semantic` tap; one `interaction` with `gesture=scroll|swipe`.
+7. **Diagnostic isolation** — enrichment selection of `stream: semantic`
+   excludes `capture_diagnostic`.
+8. **Rage-tap precision** — three no-result taps on the same origin target flag
+   once; three scrolls or three successful navigation taps do not.
+
+## Soft / observational
+
+- Semantic event count per completed gesture should drop vs 0.4.0 raw
+  `tap`+`tap_settled`+scroll peer inflation.
+- Legacy projection remains present until collector/graph cut over; do not
+  delete `tap`/`tap_settled` selection until two representative Blend flows pass
+  on canonical interactions alone.
+- Instrument pending-to-success latency; retune `interactionClaimWindow` from
+  production evidence if 1,250 ms is too short/long.
+
+## Consumer follow-ups (separate PRs)
+
+- Collector / Context Graph enrichment select `stream: semantic` `interaction`
+  and map components via `origin.targetAnchor`.
+- Build causal edges from `result` / `causedByInteractionId`.
+- Update dashboard rage-tap detectors to the definition above.
diff --git a/docs/integration/production-replay-acceptance.md b/docs/integration/production-replay-acceptance.md
@@ -12,14 +12,19 @@ database receipt alone as proof that a replay is correct.
 
 ## Current acceptance status
 
-Production acceptance #13/#14 remains open. The SDK's route-epoch and frame
-provenance behavior is an intended invariant, but rapid/nested modal chains and
-programmatic/automatic navigation can still be absent or degraded in a
-production replay. Record those observations as SDK capture gaps; do not infer
-route/action coherence from the intended contract or repair the evidence in the
-dashboard. Stored tap coordinates are global logical pixels and are not
-capture-boundary-normalized for playback, so fractional overlay drift is also a
-known limitation.
+Interaction consolidation shipped in SDK **0.4.15** (canonical `interaction`
+events, 1,250 ms delayed claim window, diagnostic stream isolation). Use
+[`production-replay-acceptance-0.4.15.md`](./production-replay-acceptance-0.4.15.md)
+for the Blend scoring gates. Collector/Context Graph migration onto
+`stream: semantic` interactions remains a follow-up before legacy
+`tap`/`tap_settled` projection can be removed.
+
+Production acceptance #13/#14 remains open for rapid/nested modal chains and
+programmatic/automatic navigation gaps. Record those observations as SDK
+capture gaps; do not infer route/action coherence from the intended contract or
+repair the evidence in the dashboard. Stored tap coordinates are global logical
+pixels and are not capture-boundary-normalized for playback, so fractional
+overlay drift is also a known limitation.
 
 ## Roles and evidence