Skip to content

talkops-ai/talkops-loki

v0.2.0Apache-2.0

Operational Agent Skills and MCP execution server for Grafana Loki 3.x and Loki Operator (loki.grafana.com/v1): discovery-first LogQL authoring, structured metadata & Bloom filter optimization, service error triage, release regression comparison, cross-pillar trace correlation, FinOps volume attribution, Loki Ruler alerting/recording rules, Compactor log deletion, and Kubernetes LokiStack management.

authoring-logql-and-analyzing-logs

Authors, lints, cost-checks, optimizes, and executes Grafana Loki LogQL log and metric queries using the 4-stage pipeline order (stream selector, line filter, parser, metadata/unwrap), index scan preflight checks, CIDR/pattern filters, and interactive A2UI views. Use when writing or optimizing LogQL queries, computing log error ratios or unwrap latency percentiles, filtering Loki 3.x structured metadata, fixing SampleExtractionErr or slow scans, or rendering A2UI log tables and charts. Do not use for Ruler/AlertingRule alert creation (use authoring-loki-alerting-and-recording-rules), schema/pattern discovery on unknown logs (use discovering-loki-schema-and-log-structure), multi-service incident triage (use triaging-loki-incidents-and-errors), or Prometheus PromQL queries (use authoring-promql-and-analyzing-metrics).

authoring-loki-alerting-and-recording-rules

Designs, validates, backtests, and deploys Grafana Loki Ruler alerting rules, recording rules, and Kubernetes Loki Operator AlertingRule/RecordingRule CRDs with dry-run and confirmation guardrails. Use when creating or deleting Loki Ruler rule groups, authoring AlertingRule or RecordingRule CRDs, pre-aggregating expensive LogQL queries into Prometheus metrics, tuning noisy/high-series log alerts, or running /setup-monitoring for log alerts. Do not use for Prometheus TSDB PrometheusRule CRDs (use authoring-and-testing-prometheus-rules), ad-hoc LogQL queries (use authoring-logql-and-analyzing-logs), LokiStack cluster provisioning (use operating-loki-deployments-and-ingestion), or Alertmanager routing simulation (use validating-alertmanager-routing).

discovering-loki-schema-and-log-structure

Discovers Grafana Loki indexed stream labels, scoped label values, active stream combinations, Loki 3.x detected fields/structured metadata, OTLP log mappings, Promtail unpack envelopes, and repeating log patterns to infer optimal LogQL parsers. Use when exploring an unfamiliar Loki cluster or tenant, onboarding services, locating OTLP resource/log attributes, or inferring pattern/logfmt/json/unpack/regexp parsers for raw log lines. Do not use for executing LogQL analytics or A2UI charts (use authoring-logql-and-analyzing-logs), auditing stream cardinality FinOps (use optimizing-loki-cardinality-and-finops), incident triage (use triaging-loki-incidents-and-errors), or Tempo trace attribute discovery (use authoring-traceql-and-metrics-queries).

operating-loki-deployments-and-ingestion

Operates Grafana Loki 3.x clusters, diagnoses loki_discarded_samples_total ingestion drops, manages Compactor GDPR/PII log deletion requests, tunes Grafana Alloy/OTel Collector OTLP ingestion pipelines, and sizes/applies LokiStack Kubernetes Operator CRDs. Use when troubleshooting Loki readiness or 429/too-old/line-too-long ingestion drops, creating or cancelling Compactor delete requests, choosing Monolithic vs Simple Scalable vs Microservices sizing, or deploying/scaling LokiStack CRs. Do not use for stream cardinality FinOps audits (use optimizing-loki-cardinality-and-finops), AlertingRule/RecordingRule authoring (use authoring-loki-alerting-and-recording-rules), TempoStack operations (use operating-tempo-deployments-and-tenants), or OTel Collector OTTL redaction processors (use controlling-telemetry-cardinality-and-noise).

optimizing-loki-cardinality-and-finops

Audits Grafana Loki log ingestion byte volume, stream cardinality explosions, tiny-chunk S3 PUT storms, TSDB v13 migrations, per-tenant limits, and Bloom filter FinOps optimizations. Use when investigating high Loki storage/S3 bills, attributing ingestion volume by namespace or service, calculating theoretical max streams, demoting pod/session_id/trace_id/client_ip to Loki 3.x structured metadata, dropping noisy healthcheck/debug logs, or sizing TSDB v13 Bloom filters. Do not use for Prometheus metric TSDB cardinality (use optimizing-tsdb-cardinality-and-finops), live OTel Collector OTTL patching (use controlling-telemetry-cardinality-and-noise), log parser format inference (use discovering-loki-schema-and-log-structure), or Compactor PII deletions (use operating-loki-deployments-and-ingestion).

triaging-loki-incidents-and-errors

Triages production incidents, 5xx spikes, CrashLoopBackOff/OOM pods, silent log streams, canary regressions, latency-without-5xx slowdowns, and cross-service failures using Loki N-to-1 summaries, period/cohort comparisons, error signature clustering, and trace/order ID log correlation. Use when investigating outages in Loki, comparing current vs baseline or canary vs stable logs, correlating trace_id or order_id timelines across services, clustering raw error lines, or running /triage. Do not use for Tempo span-waterfall analysis (use triaging-tempo-service-errors), Prometheus SLO burn math (use triaging-prometheus-incidents-and-slos), Alertmanager alert-storm grouping (use triaging-alertmanager-alerts), or LogQL syntax tutorials (use authoring-logql-and-analyzing-logs).