Knuckles-Team Telemetry Ingestion Workflow
CONCEPT:KG-2.12
Extracts anomalous execution traces and success rates from Langfuse, then ingests them into the Knowledge Graph as ExecutionSummary and PerformanceAnomaly nodes.
Steps
Step 0: Langfuse Trace Retrieval [skill: langfuse-mcp]
Agent: intake-agent
Tools: graph_query, nc_files
Fetch recent execution traces and filter for long-running or failed tasks based on baseline configurations.
Expected: trace, filter
Step 1: Langfuse Metrics Aggregation [skill: langfuse-mcp]
Agent: processor-agent
Tools: graph_analyze, document_tools
Calculate overall success rates and token consumption averages across workflows.
Expected: success, token
Step 2: Graph Os
Agent: validator-agent
Tools: graph_query
Write Cypher queries using kg_write to ingest ExecutionSummary and PerformanceAnomaly nodes into LadybugDB with relationships to their respective Workflow, Agent, and Tool nodes.
Expected: cypher, ingest
Step 3: KG Persistence [depends_on: graph-os]
Agent: validator-agent
Tools: graph_write
Persist workflow results as nodes and edges in the Knowledge Graph. Create appropriate typed nodes with metadata and link to existing domain entities.
Output
- Telemetry Ingestion results persisted in KG
- Structured report (MD/PDF)
- Audit trail with timestamps and agent attributions
Human Oversight Required
✅ Critical decisions require human review and approval.
Execution
Run this workflow as a dependency-ordered DAG. Steps with no unmet depends_on run in parallel; dependents run after their prerequisites complete.
- Run first (in parallel): Step 0 — Langfuse Trace Retrieval; Step 1 — Langfuse Metrics Aggregation; Step 2 — Graph Os
- After level 0: Step 3 — KG Persistence
Execution: If graph-os is reachable, offload the whole DAG via graph_orchestrate action=execute_workflow (or the kg-delegate skill) for true parallel/swarm execution. Otherwise execute the steps natively in dependency order: run steps with no unmet depends_on in parallel, then their dependents.