Open your Stalwart mailbox using the existing secure sign-in. No Google login or Control Panel OAuth required.
No inferred activity: WORKING requires a live native PID and recent native stream output. Dead PIDs and stale TaskStore states are shown explicitly.
| Task ID | Executor Agent | Target Node | State |
|---|---|---|---|
| No active tasks in stream. | |||
Task dispatch and edition review are available in the operator console above. Host power and service restart controls are intentionally not exposed here.
Operational memory and evidence authority
Auditable memory health, evidence retrieval, verified task writeback and attributable reasoning records.
Primary operations graph
Live departments, workers, services, routes, tasks and evidence from Control Panel and AI-AMS sources. Click a node to open its existing surface. Code graph stays a secondary drill-down.
Store health
Read-only retrieval
No query has been run. Results will include source and chunk provenance when supplied by AI AMS.
Verified outcomes
This surface is read-only. Verified result storage remains governed by the controlled-task writeback path.
Attributable summaries
Evidence-linked decisions and dissent only. Private chain-of-thought is never stored or displayed.
Secondary drill-down
AST-indexed files, routes and symbols. This is a drill-down under the primary operations topology, not the company graph.
| Method | Rule | Endpoint | Handler |
|---|---|---|---|
| UNAVAILABLE | |||
Unified entity view
Read-only file, symbol, task or service context from /api/ai-ams/entity-context. Same live API as Flask AI-AMS. No invented graph.
UNAVAILABLE
Session short-term memory
Read-only open sessions and compacted working facts from the canonical AI-AMS store. Promote to LTM stays gated. No second memory plane.
| Session | Agent | Started |
|---|---|---|
| UNAVAILABLE | ||
AEGIS Control Panel V3
Lead acquisition, qualification, L1 efficiency screening, savings intelligence and governed outreach in one operating surface. L1 is £500 fixed; L2 is 15% of verified annual savings only after actionable L1 evidence.
Upload a real machine/shift CSV. The installed open-source OEEForge engine calculates availability, performance, quality and OEE; no external email or lead mutation. Results are evidence, not a promised savings figure.
Columns: machine_id, shift_id, start, end, running_seconds, stopped_seconds, faulted_seconds, total_parts, reject_parts, ideal_cycle_seconds. The uploaded source file is discarded after analysis; only source hash and generated report are retained.
Run integrated LinePulse diagnostic on event intervals CSV.
Run integrated Libre MES reason review or work-order performance diagnostic on CSV export.
Run integrated OpenPMX machine-health (bearing RMS) diagnostic on CSV export.
Run integrated FactorySemantics I-MR QC assessment on CSV export.
Import a supplied machine/PLC telemetry JSON export through the licensed ThingsBoard v4.3.1.6 adapter into canonical AEGIS records. No MQTT connection, plant control or external send.
Expected format: {"device":"Press-1","telemetry":{"temperature_c":42,"running":true}}. Source upload is discarded; operator-only evidence and original SHA-256 are retained.
Consumes GET /api/database/leads only.
| ID | Company | Sector | Country | Status | Score | Savings |
|---|---|---|---|---|---|---|
| UNVERIFIED | ||||||
Read-only simulated sandbox · no outbound email · replies and staff responses are recorded to the internal sandbox only.
Live filtered lead data from the canonical database. CSV export uses the same current dataset/search/status filters.
Promote intake records directly into clean leads. Delete remains an explicit destructive confirmation.
Reviewed APPROVE candidates from existing CSV-backed API.
Proposal is read-only. Draft generation is preview-gated. Send stays disabled.
UNAVAILABLE
Single-record owner actions are live. Bulk destructive/send operations remain protected until a concrete selection is present.
Replies are shown only if present in current lead/audit payloads. Otherwise UNAVAILABLE.
Control Panel V3
Human-gated council review, visual approvals, evidence hashes, revision requests, controlled-task verification, and pending operator decisions.
UNAVAILABLE
UNVERIFIEDApproval locked: evidence UNVERIFIED.
UNAVAILABLE
UNAVAILABLE
UNAVAILABLE
Capacity facts are canonical hardware facts. Power, reachability and utilisation are displayed only from live internal APIs; unobserved devices remain UNAVAILABLE or SUSPENDED.
Every row is derived from a canonical controlled task. Click a row for acceptance, verifier, iFixAI and evidence summary; the Evidence link opens the canonical task API.
All departments. Ten repositories per batch. Source inventory status is not task acceptance; existing TaskStore execution contracts above remain authoritative for completion.
build_hardware_network_services_map(). Unknown service, route, endpoint, SMART, MCE or network state remains UNAVAILABLE.build_operations_map(). Controlled task, verifier, iFixAI, LOTO and AI-AMS writeback provenance required.Follow actual source-backed relationships, impact paths and upstream/downstream dependencies. Graph loads only when requested; no inferred or fabricated connections.
Autonomous multi-modal rendering, narration, and media packaging lifecycle.
Live Blackwell notebook for Council and Newsroom. Herdr manages the workers; MoLab is the GPU execution environment.
Only a direct QC PASS can unlock film review or Telegram. A reachable notebook does not mean a render is complete.
Loading packaging evidence from /api/productions…
Read-only review state from the governed website-production lane. No publish control is exposed.
Open governed video review page · Open governed productions surface
Operator preview backed by live repository evidence. The card can inspect the canonical media manifest and stage complete files for review; it does not create production jobs, queue renders or bypass existing AEGIS/Herdr execution routes.
Complete-file multipart ingest only. Files are stored under staging/imported/ and remain unapproved.
AEGIS INDUSTRIAL / MARKET INTELLIGENCE
Crypto + commodities / execution costs / paper performance. Research only. No account orders.
Unified research engine state unavailable until verified.
No verified free-model advisory yet.
Forward-testing observed wallet-buy activity against later market prices. No real trades or verified returns.
| Asset | Forecast 30m | After assumed costs | Chronological validation | Decision |
|---|---|---|---|---|
| No completed local model inference verified yet. | ||||
One AEGIS ledger: public DexScreener, GeckoTerminal trade/wallet activity, Solana contract state, RugCheck risk, Sentry scoring and local model analysis. Wallet activity is not proven wallet profit. No live trading orders.
| Token | Liquidity (USD) | 24h volume (USD) | Since prior scan | Sentry score | Pool wallet flow | On-chain / RugCheck | Decision |
|---|---|---|---|---|---|---|---|
| No third-party discovery yet. | |||||||
Commodity futures observations from this same $80-ledger research cycle. Quote delays and market closures apply. No commodity trading account or executable contract access is connected.
| Commodity | Class | Observed price | Market quote time | Execution |
|---|---|---|---|---|
| Loading verified observations… | ||||
Lower fees are not always lower execution cost. Order-book spread, liquidity, FX and slippage change the result. Personal exchange access and fee tier are not verified.
No maker-fill guarantees and no PAXG CoinJar pair assumption.
Evidence: canonical Herdr paper ledger · 7 spot crypto markets · 16 commodity futures benchmarks. Real trading disabled.
Autonomous RSS 2.0 / Atom feed validation, story synthesis, and automated narration drafting.
Script intake is read-only from the live newsroom feed. No publish action is exposed here.