Logs tell you what happened. Code tells you why. The gap between them is where root cause hides, and where most automated systems fail.
For FRR #20366, the logs told us a route-map with set local-preference +10 produced a LOCAL_PREF of 10 instead of 110. What they didn't tell us was why the relative operation failed. For that, we had to reason about the code itself, its structure, its relationships, its logic, none of which shows up in a flat text dump.
In Part 6, we go under the hood of our code reasoning stack: how Tree-sitter parses a 185,000+ line C codebase into structured ASTs, how we build a Neo4j graph of function relationships, and how that graph let us search across 18,549 functions to find the two that actually needed fixing.
Download Part 6 to see how we turn raw source code into a queryable graph and connect it back to root cause.
