Skip to content
AftergraphPublic

About

Operator and product experience for Aftergraph systems: mission status, evidence, approvals, needs-you flows and verified outcomes.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

97 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

OpenSSF Scorecard

Aftergraph Studio — Unified Intelligence Operating Environment

The sovereign operating environment for governed AI intelligence, durable work, and verified outcomes (Aftergraph Workspace v5 foundation, V6 Unified Intelligence OS).

V5 preserves the low-chrome Chat/Work experience from V4 and adds Space as a first-class spatial work mode for durable objects, active agents, evidence, replay and multimodal intent.

Objects persist. Surfaces adapt. Capabilities extend. Agents act. Humans retain control.

Primary product model

The permanent human-facing modes stay intentionally small:

  • Chat — conversation-first.
  • Work — outcome-first.
  • Space — composition-first.

Projects and Unified Recents provide continuity. The canonical ownership domains remain internal/contextual:

NOW · CHAT · WORK · AGENTS · BRAIN · OUTPUT · CONTROL · CONNECT · SYSTEM

V5 inhouse package boundary

The product API and visible interaction language are Aftergraph-owned:

  • packages/ui — semantic product components and control surfaces.
  • packages/motion — lifecycle-bound semantic motion and reduced-motion policy.
  • packages/tokens — semantic color, type, density, depth and motion tokens.
  • packages/icons — Aftergraph icon vocabulary.
  • packages/runtime-ui — contextual UI composition.
  • packages/spatial — AGSpace, AGRegion, dock/focus/peek/stack primitives and semantic zoom.
  • packages/presence — human/agent presence and follow-agent semantics.
  • packages/interaction — normalized interaction commands.
  • packages/visualization — trajectory, evidence and replay visualizations.
  • packages/composer — multimodal intent, capability and context composition.

Low-level/headless primitives may be used behind this boundary in production, but the visible product consumes Aftergraph semantics rather than exposing a third-party design system.

Spatial interaction

Space is durable state, not a decorative canvas:

  • dock typed Artifact / Agent / Evidence / Replay surfaces;
  • pointer drag between regions with a keyboard-accessible Move alternative;
  • focus and balanced/overview layouts;
  • semantic zoom: Mission → Workstream → Task → Agent → Action → Evidence;
  • follow active agents through presence;
  • replay durable event history;
  • restore spatial state from the local full-stack backend.

The V4 global Pulse Rail is intentionally suppressed in Space because Presence + Context + Time are the native peripheral intelligence there.

Full-stack runtime

npm start runs the first-party Node backend and web shell on 127.0.0.1:8000 by default.

The backend provides:

  • durable JSON workspace state;
  • conversation/message writes;
  • mission execution state and SSE live updates;
  • approvals and Needs You;
  • takeover / hand-back;
  • context, memory and artifacts;
  • canonical-domain resource reads;
  • persistent spatial state;
  • persistent replay cursor;
  • multimodal attachment metadata.

Default state path:

.runtime/workspace-state.json

Override with AFTERGRAPH_STATE_FILE when needed.

Run locally

Requires Node.js 22+.

./start.sh
# or
npm start

Open:

http://127.0.0.1:8000

No npm install is required for the core web/runtime reference.

Verification

npm test
npm run verify
npm run verify:browser
npm run verify:fullstack
npm run verify:upstreams
npm run verify:polyrepo

The release gate additionally runs the V4-in-V5 regression browser suite and archive integrity checks before packaging.

Current verified release gate

  • 635/635 Node tests passed.
  • Source/conformance and syntax checks passed.
  • Expo TypeScript/source contracts passed.
  • SwiftUI parse/source contracts passed.
  • V4 behavior regression browser suite passed inside the V5 shell.
  • V5 Space desktop/mobile browser suite passed, including pointer drag docking.
  • Browser → V5 API client → real Node backend → durable state bridge passed.
  • No uncaught browser page errors were observed in the release smoke flows.

See QA-REPORT-V6.md for the exact scope and environment boundary.

PWA / offline

The service worker caches the V4 Living Interface foundation plus all V5 first-party runtime packages and v5.css, while explicitly excluding live /api/* responses from the offline cache.

Native reference layers

  • platforms/expo/
  • platforms/swiftui/

Both expose Chat / Work / Space and V5 spatial semantics. They are source/type/parse verified only in this container because Expo Go, Android Emulator, Xcode and iOS Simulator are not available here. No native device-runtime claim is made.

Truth boundary

This is a deterministic local full-stack reference build aligned to the previously verified Aftergraph architecture. It is not represented as a byte-identical checkout of the upstream Aftergraph repositories and is not connected to production WORKS/AIE/provider credentials.

The local server/runtime is real and persistence-backed; external provider/tool integrations remain reference contracts rather than live production connections.

See:

  • SOURCE-OF-TRUTH-V6.md
  • V6-IMPLEMENTATION-STATUS.md
  • QA-REPORT-V6.md (pre-V6 report archived at docs/archive/QA-REPORT.md)
  • CHANGELOG-v5.md
  • docs/superpowers/specs/2026-09-06-aftergraph-v6-unified-intelligence-operating-environment-design.md
  • docs/superpowers/plans/2026-09-06-aftergraph-v6-unified-intelligence-operating-environment.md

V6.0 (Unified Intelligence OS) is the verified foundation. Milestones V6.1 through V8.3 are delivered (Live Federation, Universal Search, Capability Runtime, Research OS, Venture OS, Intent OS, Adaptive Workspace, Agent Society, Temporal Intelligence, Outcome Economy, Institutional Intelligence, Distributed Aftergraph, Personal Brain, Autonomous Operations); V9.0 and V10.0 are north-star horizons. See docs/ROADMAP.md and CHANGELOG-v6.md.

V5.1 polyrepo integration

V5.1 connects the workspace to independently verified source-truth boundaries for Trust Gateway, WORKS, AIE, Work Intelligence V2, After Graph Governance and ISR. See POLYREPO-INTEGRATION-V6.md (pre-V6 version archived at docs/archive/POLYREPO-INTEGRATION.md).

The available container could not resolve github.com, so a normal git clone could not complete. The authenticated GitHub connector was used to verify default-branch heads and inspect the canonical contracts/source seams. Exact contracts are materialized under the sibling upstreams/ directory, while reviewed interface notes are explicitly labelled as notes rather than source clones.

The runtime path is:

V5 UI
  -> local V5 API / durable state
  -> source-specific adapters
     -> Trust Gateway: approvals / enforcement / audit
     -> WORKS: durable work / events / evidence / Brain
     -> AIE: authority-owned A2A task operations
     -> Work Intelligence: observations / proposals / explicit review+promotion
  -> Governance: canonical contract validation

POST /api/v1/upstreams/sync is read-only. Consequential writes are explicit and delegated to the canonical owning service. Background upstream/SSE updates cannot steal the user's local Chat/Work/Space or domain navigation.

About

Operator and product experience for Aftergraph systems: mission status, evidence, approvals, needs-you flows and verified outcomes.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages