State
Live operating context
Workstreams, agents and human decisions are visible in one current-state layer.
Helia Nexus
Command layer
Executive AI command layer
Make agentic work visible, governable, executable.
Helia Nexus coordinates agents, memory, decisions, tasks and execution state in one calm operating layer.
State
Live operating context
Workstreams, agents and human decisions are visible in one current-state layer.
Control
Governed agent output
Agent work is surfaced with source, confidence, risk and required approval gates.
Action
Execution rhythm
Every view pushes towards ownership, next action and a durable decision trail.
Product model
Not another AI dashboard. A state layer for intelligent execution.
Every surface must answer what is happening, what matters, what needs a decision, who is acting, and what context must persist.
Control
Governed agent output
Agent work is surfaced with source, confidence, risk and required approval gates.
Action
Execution rhythm
Every view pushes towards ownership, next action and a durable decision trail.
Live dashboard preview
A command view backed by the workspace API.
The preview now reads current state from the local API, posts a structured agent event, and refreshes the command surface without leaving the page.
Live command view
Execution state
Loading stage
Workstreams
0
Active execution lanes
Decisions
0
Open and recommended gates
Agent events
0
Visible event feed
Persisted
0
Stored agent events
Review queue
0
Needs human review
Actions
0
Open execution objects
Risks
0
Open risk register
System health
Storage modeUnknown
Persisted events0
Last eventNone yet
Ingest authUnknown
API statusLoading
State storeLocal file dev store
Policy / tool visibility
The dashboard now exposes the rules and callable interfaces agents are using.
Router policy
vunknown
Telemetry threshold0%
Review below0%
Priority reviewUnknown
Always reviewUnknown
Default targetsUnknown
Agent tool contract
unknown / vunknown
Tools
0
Manifest entries
Read
0
Read-only tools
Write
0
State-changing tools
Coverage
0%
OpenAPI mapped
Default write path / unknown
Operating map
MVP execution map
A lightweight project-galaxy layer: what is active, what needs review, and where agent work connects to execution.
active
Helia Nexus
Local dev state layer
0 events
active
Review Queue
No visible review pressure
0 pending
queued
Execution Registers
Actions, risks, memory candidates
0 open actions
active
Risk / Decision Layer
Decision gates and risk register
0 decisions / 0 risks
queued
Agent Interface
Router policy and tool contract
0 route logs
protected
Portfolio Frontier
AI Small Business, AgentEmployee, EPC work
manual links next
Map pressure
Stable
Review/risk signal
Agent linkage
0
Router decisions logged
Next layer
Manual links
Projects, ownership, dependencies
Growth visibility
Adaptive agent capability model
Growth is variable-first. A user agent can track hundreds or thousands of evidence-backed variables depending on the user case, usage pattern, domain, tools, and confirmed upgrades.
Agent profile frame
Class
Strategic
Cognition / execution agent
Mode
Adaptive
No fixed stat ceiling
Variables
Unbounded
Schema grows by use case
Evidence
7
Verified MVP capabilities
This is not a hard-coded avatar score. The visible panel is a compact projection of a larger variable graph that should later be agent-maintained and user-specific.
Current capability projection
Execution
Active
Verified builds, commits, dashboard increments, and working service checks.
Governance
Expanding
Router policy, review queue, tool manifest, and decision audit trail are visible.
Context
Adaptive
Signals are modelled as variable evidence, not a fixed character sheet.
Reliability
Evidence-led
Growth only counts when backed by real artifacts, tests, source, or operator confirmation.
Evidence stream
sample projection
Review Queue and hygiene controlsRouter policy and decision logOpenAPI and tool manifest contractProtected server-side CLI runbookOperating map / execution layer
Capability graph
Adaptive skill tree projection
The visible tree is only a projection. Real capability growth should add, merge, retire, and weight branches dynamically from evidence and user context.
live
Strategic Director
Frames direction, trade-offs, and next best action.
Evidence / 0 decisions visible
live
Execution Operator
Turns review items into actions, risks, decisions, and memory candidates.
Evidence / 0 open actions
live
Risk & Governance
Routes uncertain work to review with source, confidence, and audit trail.
Evidence / 0 route decisions
live
Codex Builder
Builds, verifies, and commits product increments with real tests.
Evidence / Verified app builds
developing
Memory Curator
Separates durable knowledge from temporary task state.
Evidence / 0 candidates
latent
Domain Extensions
User-specific branches should emerge from usage, not a fixed template.
Evidence / variable graph next
Graph doctrine
Branches are not permanent categories. Future agents should create domain branches, store raw variables separately, and show only the useful projection for the user’s current decision context.
Mission log
Evidence of growth over time
A growth log should capture meaningful capability changes and verified outcomes, not raw chat volume or decorative levelling.
01
capability
Capability graph added
Growth model now separates raw variables from visible skill-tree projection.
02
capability
Growth visibility added
Agent development is framed as unbounded, user-specific, and evidence-led.
03
governance
Governance spine established
Agent outputs now have policy, tool contract, review queue, and router audit trail.
04
execution
Execution objects flowing
Review items can become durable actions, risks, decisions, or memory candidates.
05
risk
Noise risk identified
Review hygiene separates operational items from smoke/demo/verification noise.
Signal type
Proof
Evidence before score
Cadence
Milestone
Meaningful changes only
Next layer
Persist
Store growth events via API
Open decisions / next quests
Turn state into the next move
This panel converts the operating map, growth proof, queue pressure, and execution registers into decision gates and next quests.
decide / Dean
Prioritise the next product surface
Decision: Confirm whether Next Quests becomes the main command layer before deeper UI polish.
Leverage: Prevents the dashboard becoming only observability; turns state into movement.
Evidence / MVP brief requires open decisions / next quests
watch / Helia
Triage real review items
Decision: Keep the queue clear and route only high-signal items.
Leverage: Protects trust in the human review surface.
Evidence / 0 pending review items
decide / Helia
Advance execution objects
Decision: Create the next execution object from a confirmed decision.
Leverage: Keeps the app connected to real execution, not dashboard decoration.
Evidence / 0 open actions / 0 open risks
decide / Agent
Prepare persistent growth events
Decision: Decide the minimal API/data model for mission-log events before adding more growth UI.
Leverage: Turns current static proof into durable agent-maintained history.
Evidence / 0 router decisions available as audit precedent
Current decision register
No open decision records are visible yet.
Router decision log
0 recorded
Audit trail showing how policy routed observations into telemetry or human review.
No router decisions recorded yet. Live router writes will now create this audit trail.
Loading workspace state...
Review queue hygiene
0 pending / 0 visible
Filters keep the human review surface focused. Test noise is hidden by default.
0 total0 pending0 test-noise pending0 hidden by limit
No review items match the current hygiene filters.
Operator runbook
Protected queue triage
The UI governs visibility. Protected cleanup and conversion should run server-side from the app directory so secrets stay out of the browser.
npm run nexus -- reviews list --status needs_review --limit 10npm run nexus -- reviews list --status needs_review --include-test-noisenpm run nexus -- reviews archive-test-noise --max-items 25npm run nexus -- reviews convert --id <review-item-id> --target actionRun from /root/projects/helia-nexus/app. The CLI loads .env.local by default and sends NEXUS_INGEST_API_KEY server-side.
Actions
0 open
No converted actions yet.
Risks
0 open
No converted risks yet.
Memory candidates
0 candidate
No memory candidates yet.