Feature/ap dm chat

.github/workflows/canonical-unified-proof.yml

@@ -0,0 +1,98 @@
+name: Canonical Unified Proof
+
+on:
+  workflow_dispatch:
+  pull_request:
+    paths:
+      - "api/**"
+      - ".github/workflows/canonical-unified-proof.yml"
+  push:
+    branches:
+      - main
+    paths:
+      - "api/**"
+      - ".github/workflows/canonical-unified-proof.yml"
+
+concurrency:
+  group: canonical-unified-proof-${{ github.ref }}
+  cancel-in-progress: true
+
+permissions:
+  contents: read
+
+jobs:
+  proof:
+    runs-on: ubuntu-latest
+    timeout-minutes: 20
+
+    services:
+      postgres:
+        image: postgres:16
+        env:
+          POSTGRES_USER: postgres
+          POSTGRES_PASSWORD: localpassword
+          POSTGRES_DB: memory
+        ports:
+          - 5432:5432
+        options: >-
+          --health-cmd "pg_isready -U postgres -d memory"
+          --health-interval 10s
+          --health-timeout 5s
+          --health-retries 10
+
+    env:
+      DB_URL: postgres://postgres:localpassword@localhost:5432/memory
+      API_PORT: 8794
+      FIREHOSE_BRIDGE_SECRET: ci-canonical-bridge-secret
+      JWT_SECRET: ci-jwt-secret
+      MEMORY_API_BASE_URL: http://localhost:8794
+      MEMORY_PROOF_DB_VERIFY: "true"
+
+    steps:
+      - name: Check Out Repository
+        uses: actions/checkout@v4
+
+      - name: Set Up Bun
+        uses: oven-sh/setup-bun@v2
+        with:
+          bun-version: 1.2.15
+
+      - name: Install API Dependencies
+        run: bun install --cwd api --frozen-lockfile
+
+      - name: Apply Database Schema
+        run: bun --cwd api drizzle:push
+
+      - name: Start Memory API
+        run: |
+          bun --cwd api src/index.ts > /tmp/memory-api.log 2>&1 &
+          echo $! > /tmp/memory-api.pid
+
+      - name: Wait For API Health
+        run: |
+          for i in $(seq 1 40); do
+            if curl -fsS "${MEMORY_API_BASE_URL}/health" >/dev/null; then
+              exit 0
+            fi
+            sleep 1
+          done
+          echo "Memory API failed to become healthy"
+          cat /tmp/memory-api.log || true
+          exit 1
+
+      - name: Run Canonical Unified Proof
+        run: bun --cwd api proof:canonical:unified
+
+      - name: Upload API Log On Failure
+        if: failure()
+        uses: actions/upload-artifact@v4
+        with:
+          name: memory-api-log
+          path: /tmp/memory-api.log
+
+      - name: Stop Memory API
+        if: always()
+        run: |
+          if [ -f /tmp/memory-api.pid ]; then
+            kill "$(cat /tmp/memory-api.pid)" >/dev/null 2>&1 || true
+          fi

.vscode/settings.json

@@ -1,3 +1,9 @@
 {
-  "eslint.workingDirectories": ["./vueFrontend"]
+  "eslint.workingDirectories": ["./vueFrontend"],
+  "python.terminal.useEnvFile": true,
+  "files.associations": {
+    "**/api/drizzle/*.sql": "plaintext",
+    "**/drizzle/*.sql": "plaintext"
+  },
+  "mssql.intelliSense.enableErrorChecking": false
 }

Makefile

@@ -1,5 +1,5 @@
 .DEFAULT_GOAL := help
-.PHONY: docker-build docker-up build start log stop restart
+.PHONY: docker-build docker-up build start log stop restart db-push db-generate api-typecheck
 
 DOCKER_COMPOSE_DEV=docker compose -f docker-compose-dev.yml  --env-file .env
 DOCKER_COMPOSE_PROD=docker compose -f docker-compose-prod.yml --env-file ./api/.env --env-file ./frontend/.env
@@ -83,3 +83,27 @@ publish-backend-latest:
 	export TAG=latest
 	$(DOCKER_COMPOSE_PROD) build app-backend
 	$(DOCKER_COMPOSE_PROD) push app-backend
+
+# ---------------------------------------------------------------------------
+# Database helpers
+# Requires the dev containers (make start) to be running so the 'pg' service
+# is accepting connections on localhost:5432.
+# ---------------------------------------------------------------------------
+
+# Push schema changes directly to the dev DB (fast, for local development).
+# Equivalent to: "apply whatever the Drizzle schema files describe right now".
+db-push:
+	cd api && DB_URL=postgres://postgres:$(POSTGRES_PASSWORD)@localhost:5432/postgres bun run drizzle:push
+
+# Generate a checked-in Drizzle migration SQL file from the current schema.
+# Commit the generated file in api/drizzle/ before deploying to production.
+db-generate:
+	cd api && bun run drizzle:generate
+
+# Apply all pending checked-in migration files (use in CI / production).
+db-migrate:
+	cd api && DB_URL=postgres://postgres:$(POSTGRES_PASSWORD)@localhost:5432/postgres bun run drizzle:migrate
+
+# Type-check Memory API TypeScript without emitting build artifacts.
+api-typecheck:
+	cd api && bun run typecheck

README.md

@@ -1,6 +1,6 @@
 # Memory
 
-A Microblog-App that saves all data on your Pod.
+A Responsive Microblog-App and PWA that saves all data on your Pod.
 
 Built on the [ActivityPods](https://activitypods.org) framework.
 
@@ -38,10 +38,19 @@ Currently optimised mobile devices are:
 - go into the directory `cd api`
 - run `bun install` to install the dependencies
 - run `bun dev` to start the api
+- run `bun run typecheck` to perform a dedicated TypeScript type-check for the API
+
+From the repository root, you can also run:
+
+- `bun run typecheck:api` to type-check the API
+- `bun run check:api` to run API type-check + API tests
+- `make api-typecheck` for a Makefile-based type-check entrypoint
 
 #### Running the frontend
 - go into the directory `cd frontend`
 - run `bun install` to install the dependencies
+- copy env template: `cp .env.example .env`
+- set `VITE_KLIPY_APP_KEY` in `frontend/.env` to enable Klipy GIF search/trending/recent
 - run `bun dev` to start the frontend
 
 ### Building the app for android

api/.env.example

@@ -1,4 +1,25 @@
-DB_URL=postgres://postgres:postgres@pg:5432/postgres
+DB_URL=postgres://postgres:localpassword@localhost:5432/memory
 JWT_SECRET=SuperSecretJWTSecret
 API_PORT=8794
-FRONTEND_URL=http://localhost:4000
+API_URL=http://localhost:8794
+FRONTEND_URL=http://localhost:4000
+
+# Optional: when true, Memory may remove and re-register app grants if
+# upgradeNeeded remains stuck and no active inbox webhook exists.
+ACTIVITYPODS_ENABLE_REGISTRATION_RESET_FALLBACK=false
+
+# Optional: expected frontend session policy surfaced by
+# /activitypods/notifications/status for ops visibility.
+EXPECTED_FRONTEND_SESSION_MAX_AGE_MS=28800000
+EXPECTED_FRONTEND_POD_REAUTH_DEFER_MS=14400000
+
+# Optional: bounded AT bridge retention. Keep disabled until you are ready to
+# enforce finite local history windows for AT projections.
+AT_BRIDGE_RETENTION_ENABLED=false
+AT_BRIDGE_RETENTION_INTERVAL_MINUTES=60
+AT_BRIDGE_RETENTION_BATCH_SIZE=5000
+AT_BRIDGE_RETENTION_MAX_BATCHES_PER_RUN=8
+AT_BRIDGE_RETENTION_AT_POSTS_DAYS=30
+AT_BRIDGE_RETENTION_AT_RECORDS_DAYS=14
+AT_BRIDGE_RETENTION_INACTIVE_RECORDS_DAYS=3
+AT_BRIDGE_RETENTION_DRY_RUN=false

api/AT_BRIDGE_RETENTION.md

@@ -0,0 +1,104 @@
+# AT Bridge Retention
+
+Memory's AT bridge stores two large local datasets:
+
+- `at_posts`: feed-friendly projected post rows used by the unified feed and thread hydration.
+- `at_records`: raw supported lexicon records used for record inspection and some live derivations such as followed-author resolution.
+
+## Retention policy
+
+The cleanup policy is intentionally collection-aware.
+
+- Keep active follow records indefinitely:
+  - `app.bsky.graph.follow`
+  - `canonical.follow`
+- Delete inactive `at_records` after a shorter window.
+- Delete active non-follow `at_records` after a bounded window.
+- Delete old `at_posts` after a bounded window.
+
+Recommended starting values:
+
+- `AT_BRIDGE_RETENTION_INACTIVE_RECORDS_DAYS=3`
+- `AT_BRIDGE_RETENTION_AT_RECORDS_DAYS=14`
+- `AT_BRIDGE_RETENTION_AT_POSTS_DAYS=30`
+- `AT_BRIDGE_RETENTION_BATCH_SIZE=5000`
+- `AT_BRIDGE_RETENTION_MAX_BATCHES_PER_RUN=8`
+- `AT_BRIDGE_RETENTION_INTERVAL_MINUTES=60`
+
+## Environment variables
+
+- `AT_BRIDGE_RETENTION_ENABLED`: enable the cleanup service. Default `false`.
+- `AT_BRIDGE_RETENTION_INTERVAL_MINUTES`: run interval in minutes. Default `60`.
+- `AT_BRIDGE_RETENTION_BATCH_SIZE`: max rows deleted per batch. Default `5000`.
+- `AT_BRIDGE_RETENTION_MAX_BATCHES_PER_RUN`: max batches per table per run. Default `8`.
+- `AT_BRIDGE_RETENTION_AT_POSTS_DAYS`: retention window for `at_posts`. Default `30`.
+- `AT_BRIDGE_RETENTION_AT_RECORDS_DAYS`: retention window for active non-follow `at_records`. Default `14`.
+- `AT_BRIDGE_RETENTION_INACTIVE_RECORDS_DAYS`: retention window for inactive `at_records`. Default `3`.
+- `AT_BRIDGE_RETENTION_DRY_RUN`: log what would be deleted without deleting. Default `false`.
+
+## Safe recovery SQL
+
+For an already bloated local database, reclaim space in this order.
+
+1. Remove inactive raw records in batches:
+
+```sql
+WITH doomed AS (
+  SELECT ctid
+  FROM at_records
+  WHERE is_active = false
+    AND ingested_at < now() - interval '3 days'
+  LIMIT 5000
+)
+DELETE FROM at_records
+WHERE ctid IN (SELECT ctid FROM doomed);
+```
+
+2. Remove old active non-follow raw records in batches:
+
+```sql
+WITH doomed AS (
+  SELECT ctid
+  FROM at_records
+  WHERE is_active = true
+    AND ingested_at < now() - interval '14 days'
+    AND collection NOT IN ('app.bsky.graph.follow', 'canonical.follow')
+  LIMIT 5000
+)
+DELETE FROM at_records
+WHERE ctid IN (SELECT ctid FROM doomed);
+```
+
+3. Remove old projected AT posts in batches:
+
+```sql
+WITH doomed AS (
+  SELECT ctid
+  FROM at_posts
+  WHERE ingested_at < now() - interval '30 days'
+  LIMIT 5000
+)
+DELETE FROM at_posts
+WHERE ctid IN (SELECT ctid FROM doomed);
+```
+
+4. Refresh planner stats after batched deletes:
+
+```sql
+ANALYZE at_records;
+ANALYZE at_posts;
+```
+
+5. During a maintenance window, reclaim table file space:
+
+```sql
+VACUUM (VERBOSE, ANALYZE) at_records;
+VACUUM (VERBOSE, ANALYZE) at_posts;
+```
+
+6. Only after the database has headroom again, rerun thread graph migrations.
+
+## Notes
+
+- Do not blindly delete old follow records. Memory currently derives followed-author thread bumping from active follow records.
+- `VACUUM FULL` is intentionally not the default recommendation because it requires additional free space and stronger locking.

api/README.md

@@ -13,3 +13,59 @@ bun run index.ts
 ```
 
 This project was created using `bun init` in bun v1.1.45. [Bun](https://bun.sh) is a fast all-in-one JavaScript runtime.
+
+## ActivityPods Notification Interop
+
+The API exposes ActivityPods interoperability resources and notification webhook endpoints used by Memory app registration and inbox notification delivery.
+
+### Relevant environment variables
+
+- `API_URL`: public URL of the Memory API (used in app metadata and webhook callback URLs).
+- `FRONTEND_URL`: public URL of the Memory frontend (used for authorization callback endpoint metadata).
+- `JWT_SECRET`: used to sign local Memory JWTs and, by default, webhook callback signatures.
+- `ACTIVITYPODS_WEBHOOK_SECRET` (optional, recommended): dedicated secret for webhook callback signatures.
+- `ACTIVITYPODS_ENABLE_REGISTRATION_RESET_FALLBACK` (optional, default `false`): guarded self-healing fallback for persistent app-registration staleness.
+- `EXPECTED_FRONTEND_SESSION_MAX_AGE_MS` (optional, default `28800000`): expected frontend trusted-session max age surfaced by notification status.
+- `EXPECTED_FRONTEND_POD_REAUTH_DEFER_MS` (optional, default `14400000`): expected frontend reauth-defer window surfaced by notification status.
+- `AT_BRIDGE_RETENTION_ENABLED` and related `AT_BRIDGE_RETENTION_*` variables: bounded cleanup for local AT bridge history. See `AT_BRIDGE_RETENTION.md`.
+
+### Default-safe behavior
+
+By default (`ACTIVITYPODS_ENABLE_REGISTRATION_RESET_FALLBACK=false`), Memory will:
+
+- reconcile registration with `register-app` / `upgrade-app` when needed,
+- clean stale or duplicate webhook channels,
+- never remove an existing registration automatically.
+
+### Optional registration reset fallback
+
+When `ACTIVITYPODS_ENABLE_REGISTRATION_RESET_FALLBACK=true`, Memory may remove and re-register app grants, but only if all of these conditions are met:
+
+1. provider status remains `upgradeNeeded` after normal reconciliation,
+2. no valid inbox webhook is active,
+3. fallback is explicitly enabled.
+
+This guardrail minimizes user-impacting changes while still allowing recovery from stuck provider state.
+
+### Production rollout recommendation
+
+1. Start with fallback disabled.
+2. Observe bootstrap responses (`registrationReconciled`, `cleanedChannels`, `usedRegistrationResetFallback`) in logs/telemetry.
+3. Enable fallback only for environments where `upgradeNeeded` persists and blocks webhook activation.
+4. Keep `ACTIVITYPODS_WEBHOOK_SECRET` distinct from `JWT_SECRET` in production.
+
+### Frontend session policy defaults
+
+Recommended baseline defaults for Memory frontend:
+
+- `VITE_SESSION_MAX_AGE_MS=28800000` (8 hours): maximum trusted browser session age before requiring sign-in.
+- `VITE_POD_REAUTH_DEFER_MS=14400000` (4 hours): maximum window to defer pod re-authorization prompts while webhook delivery remains healthy.
+
+These values keep re-authentication bounded for security while reducing unnecessary prompt churn during temporary provider staleness.
+
+The notifications status response also includes:
+
+- `expectedFrontendPolicy.sessionMaxAgeMs`
+- `expectedFrontendPolicy.podReauthDeferMs`
+
+This allows operators to compare API-side expected policy with frontend runtime logs (`[SessionPolicy] Effective frontend policy`) during rollout validation.

api/drizzle.config.ts

@@ -2,7 +2,7 @@ import { defineConfig } from 'drizzle-kit'
 
 export default defineConfig({
   out: './drizzle',
-  schema: ['./src/db/schema.ts', './src/db/atBridgeSchema.ts'],
+  schema: ['./src/db/schema.ts', './src/db/atBridgeSchema.ts', './src/db/chatSchema.ts'],
   dialect: 'postgresql',
   dbCredentials: {
     url: process.env.DB_URL || 'postgres://postgres:postgres@pg:5432/postgres'