Release/1.1.0

.claude/skills/cdk-infrastructure/SKILL.md

@@ -82,6 +82,29 @@ new AlbConstruct(this, 'Alb', { config, vpc: network.vpc });
 
 SSM parameters are published **only for runtime consumption** by ECS tasks and Lambdas — never for CDK-to-CDK references within the same stack.
 
+### Wire a resource's name to every compute that reads it
+
+When a construct exposes a table/bucket that backend code reads via
+`os.environ.get("X_NAME", "default")`, you must set `X_NAME` in the container
+environment of **every** compute that runs that code — thread the typed ref
+through that compute's env builder (e.g. `buildAppApiEnvironment` for app-api,
+the inference-agentcore construct's `environment` for inference-api). Wiring one
+does **not** wire the other.
+
+**Why this bites (silent 502):** the backend's default fallback hides the
+omission. If the env var is missing, the code queries the *default* name
+(e.g. `"memory-spaces"` instead of `{prefix}-memory-spaces`), the resource
+isn't found, boto3 raises `ResourceNotFoundException`, and the centralized
+handler (`apis/shared/security/error_handler.py`) maps it to a generic
+**502 `{"detail":"Upstream service error."}`**. Nothing in `cdk synth` or CI
+catches it — the stack is valid, the IAM grant may even exist; only a runtime
+read fails. (Real instance: PR #588 — memory-spaces names were wired to
+inference-api but not app-api, which owns the CRUD routes.)
+
+**Guard it:** add an env-map unit test asserting the key is emitted (see
+`test/app-api-environment.test.ts`). **Mind the boundary:** app-api owns
+user-facing CRUD; granting IAM or wiring inference-api does not cover it.
+
 ## DynamoDB Tables
 
 - Always use PK + SK for flexibility

.github/workflows/backend.yml

@@ -165,6 +165,150 @@ jobs:
           aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID }}
           aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
 
+  build-kb-sync:
+    name: Build kb-sync image
+    needs: test-backend
+    # Native ARM64 runner — both kb-sync Lambdas are arm64 (see the
+    # kb-sync CDK construct), matching the rag-ingestion pattern.
+    runs-on: ubuntu-24.04-arm
+    environment: ${{ (github.ref == 'refs/heads/main' && 'production') || 'development' }}
+
+    permissions:
+      id-token: write
+      contents: read
+
+    env:
+      CDK_AWS_REGION: ${{ vars.AWS_REGION }}
+      CDK_AWS_ACCOUNT: ${{ vars.CDK_AWS_ACCOUNT }}
+      CDK_PROJECT_PREFIX: ${{ vars.CDK_PROJECT_PREFIX }}
+      AWS_REGION: ${{ vars.AWS_REGION }}
+      AWS_ACCOUNT_ID: ${{ vars.CDK_AWS_ACCOUNT }}
+      AWS_ROLE_ARN: ${{ secrets.AWS_ROLE_ARN }}
+      AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
+      AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
+
+    outputs:
+      image_tag: ${{ steps.build.outputs.image_tag }}
+
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+      - uses: ./.github/actions/build-and-push-image
+        id: build
+        with:
+          image-name: kb-sync
+          aws-region: ${{ vars.AWS_REGION || 'us-west-2' }}
+          aws-role-arn: ${{ secrets.AWS_ROLE_ARN }}
+          aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID }}
+          aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
+
+  deploy-kb-sync-code:
+    name: Deploy kb-sync Lambda images
+    # One image, two functions: dispatcher + worker share the kb-sync
+    # image and differ only in ImageConfig.Command (CDK-owned), so a
+    # single job points both at the freshly-built tag.
+    needs: [build-kb-sync, test-backend]
+    runs-on: ubuntu-24.04
+    environment: ${{ (github.ref == 'refs/heads/main' && 'production') || 'development' }}
+
+    permissions:
+      id-token: write
+      contents: read
+
+    env:
+      CDK_AWS_REGION: ${{ vars.AWS_REGION }}
+      CDK_AWS_ACCOUNT: ${{ vars.CDK_AWS_ACCOUNT }}
+      CDK_PROJECT_PREFIX: ${{ vars.CDK_PROJECT_PREFIX }}
+      AWS_REGION: ${{ vars.AWS_REGION }}
+      AWS_ACCOUNT_ID: ${{ vars.CDK_AWS_ACCOUNT }}
+      AWS_ROLE_ARN: ${{ secrets.AWS_ROLE_ARN }}
+      AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
+      AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
+
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+      - uses: ./.github/actions/configure-aws-credentials
+        with:
+          aws-region: ${{ vars.AWS_REGION || 'us-west-2' }}
+          aws-role-arn: ${{ secrets.AWS_ROLE_ARN }}
+          aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID }}
+          aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
+      - name: Deploy kb-sync dispatcher image
+        run: bash scripts/build/deploy-image-lambda-one.sh kb-sync-dispatcher
+      - name: Deploy kb-sync worker image
+        run: bash scripts/build/deploy-image-lambda-one.sh kb-sync-worker
+
+  build-scheduled-runs:
+    name: Build scheduled-runs image
+    needs: test-backend
+    # Native ARM64 runner — both scheduled-runs Lambdas are arm64 (see the
+    # scheduled-runs CDK construct), matching the kb-sync pattern.
+    runs-on: ubuntu-24.04-arm
+    environment: ${{ (github.ref == 'refs/heads/main' && 'production') || 'development' }}
+
+    permissions:
+      id-token: write
+      contents: read
+
+    env:
+      CDK_AWS_REGION: ${{ vars.AWS_REGION }}
+      CDK_AWS_ACCOUNT: ${{ vars.CDK_AWS_ACCOUNT }}
+      CDK_PROJECT_PREFIX: ${{ vars.CDK_PROJECT_PREFIX }}
+      AWS_REGION: ${{ vars.AWS_REGION }}
+      AWS_ACCOUNT_ID: ${{ vars.CDK_AWS_ACCOUNT }}
+      AWS_ROLE_ARN: ${{ secrets.AWS_ROLE_ARN }}
+      AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
+      AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
+
+    outputs:
+      image_tag: ${{ steps.build.outputs.image_tag }}
+
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+      - uses: ./.github/actions/build-and-push-image
+        id: build
+        with:
+          image-name: scheduled-runs
+          aws-region: ${{ vars.AWS_REGION || 'us-west-2' }}
+          aws-role-arn: ${{ secrets.AWS_ROLE_ARN }}
+          aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID }}
+          aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
+
+  deploy-scheduled-runs-code:
+    name: Deploy scheduled-runs Lambda images
+    # One image, two functions: dispatcher + worker share the
+    # scheduled-runs image and differ only in ImageConfig.Command
+    # (CDK-owned), so a single job points both at the freshly-built tag.
+    needs: [build-scheduled-runs, test-backend]
+    runs-on: ubuntu-24.04
+    environment: ${{ (github.ref == 'refs/heads/main' && 'production') || 'development' }}
+
+    permissions:
+      id-token: write
+      contents: read
+
+    env:
+      CDK_AWS_REGION: ${{ vars.AWS_REGION }}
+      CDK_AWS_ACCOUNT: ${{ vars.CDK_AWS_ACCOUNT }}
+      CDK_PROJECT_PREFIX: ${{ vars.CDK_PROJECT_PREFIX }}
+      AWS_REGION: ${{ vars.AWS_REGION }}
+      AWS_ACCOUNT_ID: ${{ vars.CDK_AWS_ACCOUNT }}
+      AWS_ROLE_ARN: ${{ secrets.AWS_ROLE_ARN }}
+      AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
+      AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
+
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+      - uses: ./.github/actions/configure-aws-credentials
+        with:
+          aws-region: ${{ vars.AWS_REGION || 'us-west-2' }}
+          aws-role-arn: ${{ secrets.AWS_ROLE_ARN }}
+          aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID }}
+          aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
+      - name: Deploy scheduled-runs dispatcher image
+        run: bash scripts/build/deploy-image-lambda-one.sh scheduled-runs-dispatcher
+      - name: Deploy scheduled-runs worker image
+        run: bash scripts/build/deploy-image-lambda-one.sh scheduled-runs-worker
+
   deploy-artifact-render-code:
     name: Deploy artifact-render Lambda code
     # Lambda zip code deploy. Mirrors the build-* jobs but for a

.github/workflows/nightly.yml

@@ -469,6 +469,9 @@ jobs:
       - name: Build rag-ingestion image
         run: docker build -f backend/Dockerfile.rag-ingestion -t rag-ingestion:scan .
 
+      - name: Build kb-sync image
+        run: docker build -f backend/Dockerfile.kb-sync -t kb-sync:scan .
+
       - name: Trivy scan app-api
         uses: aquasecurity/trivy-action@ed142fd0673e97e23eac54620cfb913e5ce36c25 # v0.36.0
         with:
@@ -487,6 +490,15 @@ jobs:
           severity: 'CRITICAL,HIGH'
           output: trivy-inference-api.txt
 
+      - name: Trivy scan kb-sync
+        uses: aquasecurity/trivy-action@ed142fd0673e97e23eac54620cfb913e5ce36c25 # v0.36.0
+        with:
+          image-ref: 'kb-sync:scan'
+          format: 'table'
+          exit-code: '0'
+          severity: 'CRITICAL,HIGH'
+          output: trivy-kb-sync.txt
+
       - name: Trivy scan rag-ingestion
         uses: aquasecurity/trivy-action@ed142fd0673e97e23eac54620cfb913e5ce36c25 # v0.36.0
         with:

.github/workflows/platform.yml

@@ -89,6 +89,29 @@ jobs:
       # nonexistent host and MCP Apps fail to load. Cert MUST be in us-east-1.
       CDK_MCP_SANDBOX_CERTIFICATE_ARN: ${{ vars.CDK_MCP_SANDBOX_CERTIFICATE_ARN }}
       CDK_MCP_SANDBOX_EXTRA_FRAME_ANCESTORS: ${{ vars.CDK_MCP_SANDBOX_EXTRA_FRAME_ANCESTORS }}
+      # KB sync scheduled re-index. Default ON with a kill switch: CDK enables
+      # the EventBridge rate(15m) rule + KB_SYNC_ENABLED on both kb-sync
+      # Lambdas UNLESS this is set to "false". Unset resolves to empty string,
+      # which config.ts treats as the default (on). Set the
+      # `CDK_KB_SYNC_ENABLED` variable to "false" in an environment only to
+      # dark-stop the feature there.
+      CDK_KB_SYNC_ENABLED: ${{ vars.CDK_KB_SYNC_ENABLED }}
+      # Scheduled runs (headless "Run now" + grant routes, and the Phase-B
+      # scheduler when it lands). Default ON with a kill switch: unset
+      # resolves to empty string, which config.ts treats as the default
+      # (on). Set the `CDK_SCHEDULED_RUNS_ENABLED` variable to "false" in
+      # an environment only to dark-stop the feature there.
+      CDK_SCHEDULED_RUNS_ENABLED: ${{ vars.CDK_SCHEDULED_RUNS_ENABLED }}
+      # Memory Spaces (user-owned markdown "second brain" + agent binding).
+      # Default ON with a kill switch: unset resolves to empty string, which
+      # config.ts treats as the default (on). Set the `CDK_MEMORY_SPACES_ENABLED`
+      # variable to "false" in an environment only to dark-stop the feature there.
+      CDK_MEMORY_SPACES_ENABLED: ${{ vars.CDK_MEMORY_SPACES_ENABLED }}
+      # Agent Designer /agents surface. Default OFF (opt-in) until the Phase-4
+      # Designer UI ships — a headless API helps no forker. Set the
+      # `CDK_AGENTS_API_ENABLED` variable to "true" in an environment (e.g. dev)
+      # to enable it there for dogfooding.
+      CDK_AGENTS_API_ENABLED: ${{ vars.CDK_AGENTS_API_ENABLED }}
       # Secrets
       AWS_ROLE_ARN: ${{ secrets.AWS_ROLE_ARN }}
       AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}

CHANGELOG.md

@@ -4,6 +4,59 @@ All notable changes to this project are documented in this file. Format follows
 
 For narrative release notes written for operators and product owners, see [RELEASE_NOTES.md](RELEASE_NOTES.md).
 
+## [1.1.0] - 2026-07-08
+
+Feature release adding four new capabilities — **Scheduled Runs** (unattended/proactive agent runs), **Memory Spaces** (bindable, shareable markdown memory), the **Agent Designer** (primitive-binding authoring surface), and **Knowledge Base Sync** (scheduled re-index of assistant sources) — plus a chat/session UX overhaul (per-conversation streaming, interrupted-turn persistence, mid-stream session titles). Every new surface is flag-gated; there are no breaking changes and no migration. Scheduled Runs, Memory Spaces, and KB Sync default **on**; the Agent Designer defaults **off**. The three preview surfaces (Agents, Memory Spaces, Scheduled Runs) are nav-gated to system-admins and carry a "Preview" badge.
+
+### 🚀 Added
+
+- **Scheduled Runs** — run an agent prompt unattended on a cadence (daily/weekday/weekly/interval) or on demand, delivered as a session with an unread indicator. New `/schedules` CRUD, `/runs/now`, and `/runs/grant`; gated by `SCHEDULED_RUNS_ENABLED` (default on) plus a new `scheduled-runs` RBAC capability (#558, #560, #561, #562, #563, #564, #565, #578)
+- `run_agent_headless` entrypoint: per-owner Cognito bearer mint, server-side SSE drain, runtime session materialization, and an audit-only fail-closed governance floor (#560, #561)
+- Revocable headless-grant record (create-on-enable from an attended session, 30-day login TTL) backing unattended run-as-user auth (#561)
+- Interval cadence (every N minutes/hours) and "Run now" with app-wide background-task toasts (#578)
+- Unread dot on scheduled-run deliveries and a Mark-as-read / Mark-as-unread toggle in the session menus (`POST /sessions/{id}/read` and `/unread`) (#572, #577)
+- **Memory Spaces** — named, templated, shareable markdown "second brain" spaces with owner/editor/viewer sharing, loss-free `.zip` export, and a consolidation health pass. New `/memory/spaces` CRUD, `/shares`, `/export`, and `/consolidate`; gated by `MEMORY_SPACES_ENABLED` (default on) (#579, #582, #584, #585, #586, #589, #597)
+- Space templates (Blank, Chief-of-Staff, Research-Notebook) and a SPA Memory panel — list/detail/create/share/export/delete (#582, #587)
+- **Agent Designer** — an authoring surface that composes an Agent from RBAC-governed primitives (a governed single-select model + uniform `bindings[]` for tool / skill / knowledge_base / memory_space), evolving the assistant store in place. New `/agents/*` surface and `/agents/bindable` catalog; gated by `AGENTS_API_ENABLED` (default off). Legacy Assistants read as Agents via a compat mapping and `/assistants/*` is unchanged (#590, #591, #592, #598, #599)
+- Run-time Agent binding resolution, re-checked against the *invoking* user's RBAC with block-on-missing: `modelConfig` override, bound Memory Space prompt injection, and `memory_*` tools (#594, #596, #601)
+- **Knowledge Base Sync** — keep assistant KB sources (Google Drive files + web crawls) automatically re-indexed on a schedule (Daily/Weekly/Monthly), with pause/resume, run-now, and reconnect-on-reauth, surfaced as per-source controls on the assistant knowledge page. Gated by `KB_SYNC_ENABLED` (default on) (#542, #543, #544, #545, #546, #547)
+- Per-conversation chat streaming: streaming state is now keyed per session, so concurrent conversations no longer cross streams; adds per-conversation scroll restore and a sidebar in-progress dot (#535)
+- Interrupted-turn persistence: a stopped or dropped response now persists its partial text, context, and cost/usage metadata, with a reload chip to Continue (`POST /sessions/{id}/interrupt`) (#541, #548)
+- Mid-stream session titles via a new `session_title` SSE event, with a shimmer skeleton while the title generates (#540)
+- Session options menu (Rename / Share / Save / Delete) on the top-nav title, and the active assistant surfaced as a pill in the top nav (#538)
+
+### ✨ Improved
+
+- Conversation export now defaults to messages only — tool calls, images, and citations are opt-in (#537)
+- OAuth call sites forward the connector's admin-configured `customParameters` verbatim, dropping hardcoded vendor baselines so the AgentCore vault key stays consistent across consent and retrieval (#550)
+- KB-sync auto-sync control clarified: self-describing verbs, always-visible "Last synced", a "Saving…" indicator, unified skeleton loading, and web-source/document parity (#552, #555)
+- Assistant editor groups the Knowledge Base section into a contrasted inset panel; added Google connector (Drive/Docs/Gmail/Calendar) logos (#557)
+
+### 🐛 Fixed
+
+- OAuth consent state is cleared on session switch so an "Authorization needed" banner can't leak onto another conversation (#539)
+- Tool card is kept after an OAuth / tool-approval resume instead of being dropped when the resumed stream omits the original `tool_use` block (#532)
+- User-sync timestamps normalized to strict ISO 8601 (single trailing `Z`), so admin "Last login" / "Created" dates no longer render as "Never" in Safari (#556)
+- Schedule edits can now clear a schedule's assistant or tool restriction via explicit clear flags (a bare `null` was read as "leave unchanged") (#569)
+- Client-supplied `enabled_tools` is intersected with the caller's RBAC on the headless paths (schedule create/update, Run now), so a scheduled run can't persist a tool outside the owner's role (#568)
+- Scheduled-runs worker lean image made importable (missing `cryptography`/`cachetools`, top-level `agents`/`strands` imports on the delivery path) (#566, #567)
+
+### 🏗️ Infrastructure
+
+- New dedicated `memory-spaces` DynamoDB table (OwnerIndex + MemberIndex GSIs) and content-addressed S3 bucket via `MemorySpacesConstruct` (#582)
+- New sparse GSIs: `DueScheduleIndex` and `HeadlessGrantUserIndex` on the sessions-metadata table (scheduled runs), and `DueSyncIndex` on the assistants table (KB sync) (#542, #561, #563)
+- Two new lean Lambda images with EventBridge sweeps, both following the platform-as-bootstrap pattern (byte-stable stub + out-of-band image deploy): `Dockerfile.scheduled-runs` (dispatcher + worker, `rate(5m)`) and `Dockerfile.kb-sync` (dispatcher + worker, `rate(15m)`) (#543, #565)
+- New feature flags forwarded through `platform.yml`, all empty-string-safe kill switches: `SCHEDULED_RUNS_ENABLED`, `MEMORY_SPACES_ENABLED`, `KB_SYNC_ENABLED` (default on) and `AGENTS_API_ENABLED` (default off) (#553, #554, #561, #593, #597)
+- IAM: added `bedrock:ListFoundationModels` / `bedrock:GetFoundationModel` to the app-api task role (admin model list), and granted the kb-sync worker read on the vault's backing OAuth secrets (#549, #571)
+
+### 🔧 CI/CD
+
+- `backend.yml` gains build + API-driven code-deploy jobs for the kb-sync and scheduled-runs Lambda images; nightly image-scan and supply-chain pinning lists include both new Dockerfiles (#543, #565)
+
+### 📦 Dependencies
+
+- New pins in the lean Lambda images only (no core backend or frontend dependency changes): `beautifulsoup4` 4.13.5, `trafilatura` 2.0.0, `lxml` 6.1.1 (kb-sync web re-crawl); `cryptography` 48.0.1, `cachetools` 6.2.4 (scheduled-runs worker); `httpx` 0.28.1, `bedrock-agentcore` 1.9.1, `boto3` 1.43.9, `pydantic` 2.12.5 (shared image runtime)
+
 ## [1.0.4] - 2026-07-01
 
 IAM hotfix restoring AgentCore Memory. Both the App API task role and the AgentCore Runtime execution role were missing `bedrock-agentcore:GetMemory`, which `get_memory_strategies()` needs to resolve strategy IDs — breaking the memory dashboard (empty results) and long-term recall (retrieval silently disabled). Ships via the platform (CDK) pipeline; no migration.