feat(seer): Add deliver_feature_result RPC for Seer agent features

[trevor-e]

Summary

Adds a generic RPC endpoint (deliver_feature_result) that Seer can call when an agent feature run completes. Features register handlers in DELIVERY_HANDLERS keyed by feature_id.

Also adds the night_shift delivery handler that:

• Looks up the SeerNightShiftRun by ref
• Marks SKIPs in the skip cache
• Triggers autofix for fixable verdicts
• Creates SeerNightShiftRunResult rows

This runs in parallel with the existing night shift code - nothing is removed or modified. Once the Seer-side PR lands and we verify it works, a follow-up PR will migrate night shift to use the new delivery path exclusively.

Test plan

• Verify imports work and linting passes
• Deploy and test with Seer PR feat(grouplink): GroupCommitResolution backfill to GroupLink #6499

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

There are 4 total unresolved issues (including 2 from previous reviews).

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 0c0baaa. Configure here.}

[cursor]

Autofix runs inside Seer RPC

Medium Severity

deliver_feature_result invokes the night-shift handler synchronously in the Seer RPC request, and that handler can call trigger_autofix_agent for every fixable verdict. The prior cron path runs the same autofix loop inside a background task with a long processing deadline.

 

^{Reviewed by Cursor Bugbot for commit 0c0baaa. Configure here.}

[trevor-e]

our result handling is very lightweight and should work fine within the RPC

[chromy]

lgtm

[sentry]

Bug: The lookup in deliver_feature_result uses seer_run__uuid, which fails if the nullable seer_run foreign key hasn't been populated yet, causing results from Seer to be silently dropped.
_{Severity: HIGH}

Suggested Fix

Modify the query to not rely on the seer_run foreign key, which may be null. One option is to add the run_uuid to the SeerNightShiftRun model directly and query on that field. This avoids the join on a nullable foreign key and prevents the race condition where the callback from Seer arrives before the seer_run association is made.

Prompt for AI Agent

Review the code at the location below. A potential bug has been identified by an AI
agent. Verify if this is a real issue. If it is, propose a fix; if not, explain why it's
not valid.

Location: src/sentry/seer/night_shift/delivery.py#L33-L36

Potential issue: The `deliver_feature_result` function looks up a `SeerNightShiftRun`
record by joining on the `seer_run` foreign key using `seer_run__uuid`. However, the
`seer_run` field is nullable and is populated asynchronously after an agent run
completes. A race condition exists where Seer might call this endpoint with a `run_uuid`
before the corresponding `seer_run` foreign key has been set on the `SeerNightShiftRun`
record. In this scenario, the database query will fail to find the record due to the
INNER JOIN, causing a `DoesNotExist` exception. The handler catches this, logs a
warning, and silently discards the result from Seer, leading to data loss.