Task board
synthkit
Composable synthetic-telemetry generator. One YAML blueprint emits structurally-correct metrics, traces, logs and RUM to Prometheus, Loki and Grafana Cloud over OTLP.
81 open, 372 completed ·All tasks ·Completed work · Documents· Repository
No tasks match these filters. Clear them to see all tasks.
To Do26
SKT-0087Support the AI Factory use case end to endSKT-0087.15Spike: facility BMS cooling telemetry (BACnet, MQTT) for liquid-cooled AI racksSKT-0087.20Ship AI Factory cross-domain incident scenariosSKT-0087.22Add AI Factory dashboards over the reference blueprintSKT-0087.24Source deferred DCGM product envelopes and NVSwitch/Grace profilesSKT-0102Checkpoint cumulative state and backfill gaps across restartsSKT-0165Source system.io.rrqm_s and wrqm_s for the Datadog standalone-host laneSKT-0221Support scoped parameterized export of the process operational dashboardSKT-0222Add a synthetic-telemetry dashboard template for otlp-nativeSKT-0230Map private realism recommendations to engine featuresSKT-0231Share per-request draws across every line of one log bundleSKT-0232Give log bundle lines their own event time advancing by each line's dwellSKT-0233Model ledger-driven stocks that accumulate and drainSKT-0234Drive histogram observation counts from the same arrival countersSKT-0235Add collector self-telemetry value sourcesSKT-0236Support per-source time zone and batch-shaped arrivalsSKT-0237Document the realism engine features in field docs and signalsSKT-0238Review the composed realism landingSKT-0239Step a last-success timestamp on a declared scheduleSKT-0240Read telemetry DSL field comments into the generated blueprint schemaSKT-0241Seed item stocks from the arrival rate over the residence windowSKT-0242Stabilise the Fleet HA graceful restart test's lease lookup timeoutSKT-0243Bring the app package race run back under the hosted test timeoutSKT-0244Fix realism engine review minorsSKT-0245Stabilise the owned Fleet HA interruption and recovery test under loadSKT-0246Diagnose the non-HA Fleet ledger release failure in the owned CLI teardown test
In Progress5
SKT-0168.09Make the private pipeline demo production-plausible and its dashboards read as real operations viewsSKT-0224Bound dynamic legacy timing evaluation cost before broad live useSKT-0227Remove quadratic emitted-log lookup from app proofs without changing their coverageSKT-0229Reuse canonical parity metric bytes without weakening full wire proofSKT-0247Classify baseline public provisioning preview refusals
Parked50
SKT-0071Settle the two remaining Azure managed Service Bus error producer pairsSKT-0087.03Add an SNMP exporter construct for device health (IF-MIB plus vendor modules)SKT-0087.05Emit NVIDIA driver kernel events (Xid, SXid, NVRM) in GPU host logsSKT-0087.06Add NVIDIA GPU Operator, Network Operator and device plugin add-on telemetrySKT-0087.07Add a Redfish BMC hardware construct (inventory, health, power, thermal, events)SKT-0087.08Add Spectrum-X / Cumulus Linux fabric telemetry via NetQ and switch OTLP exportSKT-0087.09Add NVLink domain telemetry for NVL72 racks (NMX-Telemetry and NetQ NVLink)SKT-0087.10Add Base Command Manager CMDaemon exporter telemetrySKT-0087.11Add Mission Control platform state, autonomous recovery events and forwarded Alertmanager alertsSKT-0087.12Add Run:ai scheduler telemetry with workload-to-GPU allocationSKT-0087.13Add VAST Data storage telemetry (Prometheus exporter and events)SKT-0087.14Add rack power and liquid-cooling telemetry (PDU, power shelf, CDU, leak detection)SKT-0087.18Add Slurm scheduler telemetry and DCGM HPC job mappingSKT-0087.19Add GPU inference-serving engine telemetry (vLLM, Triton, NIM)SKT-0087.21Ship an ai-factory reference blueprintSKT-0117Remove private integration evidence paths from public task notesSKT-0128Classify and repair the Fleet HA fixture initial Lease timeout without weakening proofSKT-0153Model pharma regulatory document authoring on current Agent ObservabilitySKT-0153.03Provision Agent Observability evaluators, hook rules, evaluation rules and suites from blueprintsSKT-0153.04Call hooks:evaluate from synthetic agents with real provisioned rulesSKT-0153.05Model a multi-service agent pipeline with review loops and terminal outcomesSKT-0153.05.03Wire agent namespace claims and receipt-backed runtime admissionSKT-0153.06Run real Agent Observability experiments from synthkitSKT-0153.09Ship the pharma-regulatory-authoring blueprintSKT-0153.10Document, gate and live-verify the pharma regulatory authoring estateSKT-0156Diagnose bounded HA helper HTTP readiness failure under the full race gateSKT-0158Diagnose intermittent HA renewal-loss observation timeout under loaded gatesSKT-0168Model a generic CRM data-pipeline estate: AWS data-service CloudWatch families, coherent pipeline runs and SolaceSKT-0168.02Live-capture and add CloudWatch families for Lambda, DynamoDB, SQS, provisioned Redshift and CloudWatch Logs deliverySKT-0168.03Drive Glue, MWAA, Lambda, SQS and Redshift coherently from declared pipeline runs, with stall and run-failure modesSKT-0168.07Extend the private operational dashboard with sourced AWS observationsSKT-0168.08Route the private demo's alerts through on-call and incident toolingSKT-0182Adopt causal incidents across the shipped blueprintsSKT-0182.02Migrate shipped blueprints to declared causes, one family per laneSKT-0202Report a bound Lambda's failed executions at their real durationSKT-0203Join a synchronously invoked function's trace to its caller during the callSKT-0205Align the causal model's downstream reject reads and timeout resends with the appSKT-0206Lay baseline-failed causal requests on the engine timelineSKT-0207Report a bound agent runtime's failed invocations at their real latencySKT-0208Let several unreserved Lambda functions share one account-level causal concurrency poolSKT-0211Make causal span timestamps identical on amd64 and arm64SKT-0212Declare a per-resource failed-call hold instead of a shared pre-work constantSKT-0213Use the declared healthy reference before a healthy window is learnedSKT-0214Let causally bound function duration follow a declared load curveSKT-0215Represent handled data-quality rejection spans as UNSET with a rejection markerSKT-0216Decorate baseline failures only for faults on their own causal pathSKT-0217Evaluate retrieval rejection after the store call completesSKT-0218Bind all five declared document-store clients to their causal resourcesSKT-0219Correct stale causal timing comments and parentless-span exemption scopeSKT-0223Classify unchanged custody forecast race failure in composed telemetry gate