openagentics

Flow Templates

Pre-built Flow templates seeded into the workflow engine, source-derived from the seed directory.

1 sections 5 items generated from services/openagentic-workflows/seed/templates/cost-anomaly.json, services/openagentic-workflows/seed/templates/failed-deploy-rca.json, services/openagentic-workflows/seed/templates/incident-triage.json, services/openagentic-workflows/seed/templates/rag-knowledge-qa.json, services/openagentic-workflows/seed/templates/research-and-publish.json

Seeded Templates

5 Flow templates shipped with the release

Cost Anomaly

Given a lookback window + service, queries AWS Cost Explorer (openagentic_aws.aws_cost_by_service for the per-service cost driver + openagentic_aws.aws_cost_summary for the total) IN PARALLEL with Prometheus usage over the same window (openagentic_prometheus.prometheus_query_range), converges them in a labeled merge keyed by source-node label (cost/usage), pre-computes the per-service cost table +

Failed Deploy RCA

Given a deployment + namespace, pulls Kubernetes rollout state IN PARALLEL — openagentic_kubernetes.k8s_rollout_status (is it progressing/stuck?), openagentic_kubernetes.k8s_list_events (recent namespace events) — alongside openagentic_loki.loki_search_errors (pod logs / errors for the namespace), converges them in a labeled merge keyed by source-node label (rollout/events/logs), pre-computes a ro

Incident Triage

Hero AIOps flow. Given an alert/symptom + namespace + time window, fans the trigger out to THREE parallel built-in MCP calls — openagentic_prometheus.prometheus_query (metrics), openagentic_loki.loki_search_errors (logs), openagentic_kubernetes.k8s_list_pods (cluster state) — converges them in a labeled merge keyed by source-node label (metrics/logs/kube), pre-computes per-source evidence excerpts

RAG Knowledge-Base Q&A

Grounded, cited question answering over the platform knowledge base. Expands the user's question into multiple retrieval queries (multi_query), embeds it, semantically searches the shared_knowledge collection (knowledge_search), re-ranks the retrieved chunks for relevance (rerank), fact-checks the drafted answer against the retrieved evidence (grounding_check), synthesizes a cited HTML answer (llm

Research and Publish

End-to-end data-layer proof: pulls real content from the web for a topic, ingests it into the platform knowledge base (Milvus shared_knowledge), RAGs it back against the user's question, and renders an interactive HTML report as an openable artifact in the artifact rail. Default topic is configurable. No mocks anywhere in the pipeline.