ReAct · Governed · Audited DataHub Hackathon
View:
Watch Sentinel think — then act, governed.
Inject a DataHub assertion-failure signal. Sentinel's ReAct loop investigates — fetches the asset, traverses lineage, reads prior post-mortems — then opens a GitHub issue in the demo pipeline repo and posts a Slack triage card. A code-level guardrail refuses writes to PII-tagged assets and surfaces an approval gate for ownership and glossary proposals. Every action is logged, audited, and rendered live below.
Select an injected signal to surface the on-call persona + failing asset.
LLM:llama-3.3-70b-versatile·Provider:groq
Inject a DataHub signal
Runs the full ReAct loop end-to-end. GitHub + Slack actions are logged by default (toggle below).
Inject a signal to render the asset's lineage graph.
ReAct Loop
🔴Observe
🧠Think
⚡Act
Reasoning stream
0 steps● monitoring
Select a signal above and click Inject & run Sentinel to watch the ReAct loop investigate in real time.
Enter to run⌘K command palette? shortcuts
Live metrics
Incidents
0
Reasoning steps
0
Prompt tokens
awaiting…
Completion tokens
awaiting…
Total tokens
awaiting…
LLM model
awaiting…
Cost & Efficiency
Current run
Est. time saved
awaiting…
LLM cost
awaiting…
Cumulative
Total time saved—
Incidents resolved0/0
Avg agent time—
Time saved vs. manual human triage (fetch asset, traverse lineage, read post-mortems, draft issue, post Slack, write post-mortem). Costs based on public provider pricing.
Agent Performance
No incidents yet
Connectors
DataHub Health
Incident history
Write-backs detail
No write-backs yet. Run a signal to see write-back actions here.