feat(runtime): Add live self diagnostic tool - #1964
Closed
sentry-junior[bot] wants to merge 2 commits into
Closed
sentry-junior[bot] wants to merge 2 commits into
sentry-junior[bot] wants to merge 2 commits into
Conversation
Expose current execution settings and available model profiles without relying on remembered configuration. Save diagnostic cards through the existing annotation path. Co-Authored-By: Daniel Griesser <dgriesser@sentry.io>
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
Co-Authored-By: daniel <dgriesser@sentry.io>
|
1 screenshot change — 1 changed · 0 added · 0 removed Review screenshots in Frameshift
|
Member
|
Let's table this for now, not sure if this tool adds any real value |
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Add
self_diagnosticso Junior can inspect its current execution instead of answering model questions from memory or repository defaults.The read-only tool returns the active model, profile, reasoning level, image-input support, and Conversation, Turn, and Run IDs. It lists configured profiles with effective reasoning levels and handoff availability. The callback reads current state after handoff and resume.
The diagnostic also shows the active-history context estimate, model window, effective input budget, and compaction threshold. It reports last-call cache counters and cached input share, plus token counters and estimated USD costs for completed main-model calls in the current execution slice. These counters survive history replacement but reset on resume. They exclude earlier slices, failed calls, and auxiliary models; they are not full Turn or Conversation totals. Missing counters stay unknown, including totals with incomplete reporting.
Each call saves a dated Message card through the existing Item annotation path. Later reads do not replace earlier facts. The result excludes secrets, raw configuration, and prompt content.
This PR is independent of #1946. That PR shows recorded event data; this tool reads live execution state.
via daniel.
--
View Junior Session [Sentry]