IRIS Platform Operations
Performance & Optimization Control Plane
One scorecard for pages, APIs, agents, workflows, integrations, FinOps, SAP call reduction, evidence load, replay coverage, and graceful failure. Quiet telemetry windows are shown as capture-enabled or no-sample states, not as static health claims.
Measured Window: 7 days · Source: ledger and browser telemetry
Runtime samples
3813
Browser/API action events in the selected window
Agent executions
4
AgentExecution rows
LLM calls
0
LLMAuditRecord rows
Evidence packs
665
EvidencePack rows
Confirmed SOR writes
0
SorWriteback confirmed documents
Cache hit rate
0.00 %
Measured when UI/API telemetry reports cache hits
Runtime Observability Console
0.00 ms API p95
4 agent executions
Token & Compute FinOps Console
0 USD
Cost per decision 0.00 USD
SAP Call Reduction Monitor
0
0 confirmed writebacks
Page/API Performance Audit Console
0.00 ms page p95
0 failed actions
Decision Runtime Reliability Dashboard
100 % replay
4 decisions executed
Optimization Scorecard
| Area | Metric | Value | Target | State | Samples | Source / Next action |
|---|---|---|---|---|---|---|
| UI | Core page load p95 | 0.00 ms | < 2000 ms | Capture enabled | 0 | RuntimePerformanceTelemetry.page_load_ms Keep summary-first rendering and lazy-load drawers. |
| UI | Drawer load p95 | 2,361 ms | < 3000 ms | Measured | 2 | RuntimePerformanceTelemetry.drawer_load_ms Evidence and replay drawers should fetch on open. |
| API | Read API latency p95 | 0.00 ms | < 500 ms | Capture enabled | 0 | RuntimePerformanceTelemetry.api_latency_ms Use mission summary/queue/case/evidence/replay endpoints instead of over-fetching. |
| Data | DB latency p95 | 0.00 ms | < 100 ms | Capture enabled | 0 | RuntimePerformanceTelemetry.db_latency_ms Index hot paths by case, mission, status, priority, counterparty, and created_at. |
| Agent | Agent response p95 | 12,517 ms | under 15000 ms | Measured | 1 | AgentExecutionTelemetry.latency_ms / AgentExecution.execution_time_ms Use deterministic-first routing and cached context packs. |
| Evidence | Evidence replay coverage | 100 % | 100% | Measured | 665 | EvidencePack.is_replayable Governed decisions should seal replayable evidence packs. |
| Decision | Failed user actions | 0 | 0 | Measured | 3813 | RuntimePerformanceTelemetry.outcome Every failed action should retain the case state and expose retry/replay. |
| Integration | SAP calls avoided | 0 | maximize | Capture enabled | 0 | LLMAuditRecord.evidence_json.finops_counters Prefer UDP/cache reads; reserve SAP calls for governed writes and authoritative confirmations. |
| Cost | AI spend | 0 USD | budget governed | Capture enabled | 0 | LLMAuditRecord token ledger Track cost per case, mission, user, and business outcome. |
| Value | Measured business value | 0.00 USD | trace to value events | Measured | 1 | ValueEventRecord Aggregate case value to team, tower, FB&T, and CFO views. |
Layer Audit
| Layer | State | Samples | Checks | Next action |
|---|---|---|---|---|
| UI | Measured | 2 | summary-first page load, lazy drawer load, failed browser actions | Beacon is active on the optimization page; extend to mission pages as they are touched. |
| API | Measured | 0 | read p95, duplicate calls, payload split by summary/queue/case/evidence/replay | Track read API p95 by route and split heavy payloads. |
| Data | Measured | 0 | hot-path indexes, precomputed aggregates, DB latency | Keep mission summary APIs and indexed queue endpoints as the preferred access path. |
| Agent | Measured | 1 | deterministic-first routing, token cost, agent latency | Route deterministic checks before LLM reasoning and cache context packs. |
| Integration | Measured | 0 | SAP call reduction, cache/UDP reads, governed writebacks | Start recording SOR read/write counters on each agent execution. |
Mission-Level Performance Audits
| Surface | Mission | State | Page samples | Agent calls | Value |
|---|---|---|---|---|---|
| P2P Analyst page | P2P | Needs attention | 2 | 0 | 0.00 USD |
| O2C canonical mission surface | O2C | Capture enabled | 0 | 0 | 0.00 USD |
| R2R canonical mission surface | R2R | Capture enabled | 0 | 0 | 0.00 USD |
| Treasury canonical mission surface | Treasury | Capture enabled | 0 | 0 | 0.00 USD |
| FP&A canonical mission surface | FPA | Capture enabled | 0 | 0 | 0.00 USD |
| Tax canonical mission surface | Tax | Capture enabled | 0 | 0 | 0.00 USD |
Production Telemetry Contract
| Field | Purpose |
|---|---|
| user_action_id | correlation id for every UI action |
| page | route or surface emitting telemetry |
| persona | role/persona shown in the UI |
| mission | P2P, O2C, R2R, FP&A, Treasury, Controls, or platform |
| case_id | business case or transaction id where applicable |
| api_latency_ms | read/write API latency |
| agent_latency_ms | agent execution latency |
| db_latency_ms | database latency when reported by API |
| token_count | token usage from LLM ledger or browser action metadata |
| model_used | model selected for the action |
| sap_calls | SOR calls made for the action |
| cache_hit | whether context/cache served the action |
| decision_id | decision/replay id |
| evidence_pack_id | evidence pack id |
| outcome | captured, completed, failed, escalated, or retried |
| error_code | bounded error identifier |