RFC-015: Unified Observability Protocol (UOP) — 2. Signal Types
AIGP Specification › RFC-015: Unified Observability Protocol (UOP) › 2. Signal Types
← 1. Purpose · Section index · 3. Metrics Schema →
2. Signal Types
UOP collects three signal types, each corresponding to a protocol layer in the system.
2.1 AIGP Traffic (AI Usage Signals)
AIGP signals capture every AI governance interaction requests, decisions, and usage records.
| Signal | Source | Description |
|---|---|---|
AIGP_REGISTER |
All amigos governance-server | Heartbeat registration |
AIGP_REQUEST |
All amigos governance-server | Pre-invocation policy check |
AIGP_DECISION |
governance-server requester | ALLOW / ALLOW_WITH_CONSTRAINTS / DENY |
AIGP_RECORD |
All amigos governance-server | Post-invocation usage record |
Metrics derived:
- Request volume (per-app, per-model, per-use-case)
- Decision distribution (ALLOW vs DENY vs ALLOW_WITH_CONSTRAINTS)
- Token consumption (input, output, total)
- Invocation latency (duration_ms from RECORD)
- Error rate (status != SUCCESS in RECORD)
2.2 ACP Traffic (Context Flow Signals)
ACP signals capture every context assembly and delivery operation.
| Signal | Source | Description |
|---|---|---|
ACP_FETCH |
Assembler Provider | Context fetch request |
ACP_DELIVER |
Assembler Receiver | Assembled context delivery |
ACP_SCHEMA_FAIL |
Assembler | Schema validation failure |
Metrics derived:
- Fetch volume (per-provider, per-context-type)
- Fetch latency (time to assemble full context)
- Schema validation failure rate
- Context payload size (bytes)
- Cache hit rate (TTL-based)
2.3 ALP Traffic (Artifact Lifecycle Signals)
ALP signals capture the lifecycle of manufactured artifacts scripts, configs, policies.
| Signal | Source | Description |
|---|---|---|
ALP_CREATE |
runbook-service | Artifact generated |
ALP_VALIDATE |
runbook-service | Artifact validation result |
ALP_APPROVE |
governance-server / User | Artifact approved for execution |
ALP_EXECUTE |
runbook-service | Artifact executed |
ALP_FAIL |
runbook-service | Artifact execution failure |
Metrics derived:
- Artifacts created per period
- Approval rate and approval latency
- Execution success/failure rate
- Time from creation to execution (lifecycle duration)