GitHub

When a robot logs an error, the agent finds where that message comes from in your code, so support can explain the fault.

Setup: you connect GitHub and choose the repositories.

What to use it for

• Find where a robot error message is raised in your code, and which threshold fired.
• Compare a config file at two tags to see what changed in a release.
• Check limits in your code against what the robot actually did.
• Give the code owner a specific file, line and version to start from.

Example

Fictional example.

Ask: "Since 7.3, our robots log costmap update timeout and pause. What changed?"

Agent: 0 times on 7.2.x, 214 times on 7.3.0, on 9 of 11 robots. The message is raised in planner/costmap_watchdog.cpp. Its timeout, costmap.update_timeout_ms in config/planner.yaml, changed from 250 ms at v7.2.4 to 120 ms at v7.3.0.

How to use it

• Connect GitHub in Alloy under Integrations. You choose which repositories the agent can see.
• Ask in the Alloy app or in Slack. The agent goes to the code when the question needs it.
• The connection is read-only. The agent reads code at a branch, tag or release, not pull requests or commit history. Tell it which one your fleet runs.

Related