-
-
Notifications
You must be signed in to change notification settings - Fork 0
API Reference
Cross-cutting reference. Describes the service-wrapper interface referenced in the repository roadmap (docs/api/) that exposes the pipeline described in Architecture Overview to source adapters and to the founder-facing review interface.
flowchart LR
subgraph Inbound["Inbound (Source Adapters -> Sentinel)"]
A1[POST /v1/ingest]
end
subgraph Query["Query (Review Interface -> Sentinel)"]
A2[GET /v1/queue]
A3[GET /v1/messages/:id/analysis]
A4[GET /v1/messages/:id/draft]
A5[GET /v1/negotiation/:thread_id/brief]
A6[GET /v1/capacity/current]
end
subgraph Action["Action (Founder -> Sentinel)"]
A7[POST /v1/messages/:id/approve]
A8[POST /v1/messages/:id/edit]
A9[POST /v1/messages/:id/reject]
end
A1 --> Pipe[Ingestion -> Cortex -> Persona/Negotiation/Energy]
Pipe --> A2
Pipe --> A3
Pipe --> A4
Pipe --> A5
Pipe --> A6
Used by source adapters (see Cortex Ingestion Pipeline) to submit a raw message for processing.
Request body: channel-specific raw payload plus a source field identifying the adapter.
Response: 202 Accepted with a message_id once the message has passed fingerprinting — processing continues asynchronously through scoring, drafting, and gating.
Returns the founder's current review queue, ordered by priority tier and gated by current capacity state (see Energy Shield Biometric Integration). Supports filtering by tier and by whether a draft is attached.
Returns the full AnalysisResult (see Data Schemas Reference) for a given message, including every contributing factor behind the urgency and sentiment scores.
Returns the current DraftResponse, including which Persona Contextual Mask was applied and its style-consistency score.
Returns the NegotiationBrief for a commercial thread, per Negotiation Intelligence Algorithms — always including the full factor breakdown, never a bare recommendation.
Returns the founder's current CapacityState, useful for the review interface to explain why certain items are or aren't currently surfaced.
Approves a draft as-is for sending. Logged with a timestamp and, if applicable, the auto-send policy that authorized it (see Deployment Guide).
Submits a founder-edited version of a draft. The diff between original and edited draft is retained as calibration signal for Persona Stylometry Engine.
Discards a draft without sending. Logged for the same calibration purpose.
sequenceDiagram
participant Adapter as Source Adapter
participant API as Sentinel API
participant Pipe as Internal Pipeline
participant UI as Review Interface
participant Founder
Adapter->>API: POST /v1/ingest
API-->>Adapter: 202 Accepted, message_id
API->>Pipe: process (ingestion -> cortex -> persona/negotiation/energy)
UI->>API: GET /v1/queue (polling or subscribed)
API-->>UI: prioritized, capacity-gated queue
UI->>API: GET /v1/messages/:id/draft
API-->>UI: DraftResponse
Founder->>UI: reviews draft
UI->>API: POST /v1/messages/:id/approve
API->>Adapter: (via outbound channel) send
All endpoints require authenticated access scoped to a single founder's Sentinel instance — there is no multi-tenant cross-founder query surface by design, consistent with the sovereignty principle in Sovereign Path Data Sovereignty. Exact authentication mechanism (API key, OAuth, local-only socket) is a deployment-topology choice — see Deployment Guide.
This API surface reflects the target specification for docs/api/, tracked as a Phase 4 deliverable in Roadmap and Status. Implementations should treat endpoint names and shapes here as the contract to build against.
A reference architecture for emotionally-aware business automation.
Designed and maintained by Ciprian Ștefan Pleșca (Stefano D'Angelo) · LocalPulse Research
Home · Architecture · Design Philosophy · Five Pillars · Roadmap · Glossary & FAQ
"Sentinel does not lie." — engineering constraint, not just intention. See Design Philosophy for how this principle is enforced architecturally rather than merely stated.
Repository: github.com/Ciprian-LocalPulse/the-sentinel-protocol · License: MIT · © 2026 Ciprian Ștefan Pleșca