Incident Context: tracing production failures without guessing
This is a submission for the Sanity Challenge, Path One: Ship an Agent That Queries Real Content . What I Built Incident Context is an incident-investigation agent for questions that cannot afford a confident but unsupported answer. During an outage, evidence is rarely contained in one document. It is spread across services and dependencies, deployments and released versions, configuration…
Incident Context is an incident-investigation agent designed to query real content and provide supported answers during production outages. The agent models operational records as connected graph records in Sanity documents, including services, deployments, changes, runbooks, and incidents. It follows relationships between these records to separate confirmed evidence from inference and preserve source paths.
The problem lies in keyword searches, which cannot determine causation between incidents, configurations, and runbooks. The solution is a graph model that traverses relationships to present evidence trails, confirmed evidence, inferences, and recommended next steps. The agent avoids turning correlation into confirmed causation.
To use Incident Context, users can ask questions like "What changed before INC-208?" or "Which runbook applies to INC-142?" The agent returns a validated JSON report with an evidence trail, confirmed evidence, inferences, recommended next steps, and sources. In the demo application, users can explore questions related to a small production system and compare incidents based on the evidence paths.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
