feat(kugetsu): add KUGETSU_VERBOSITY for PM agent output control #95
Closed
shoko
wants to merge 3 commits from
feat/issue-46-verbosity-v2 into main
pull from: feat/issue-46-verbosity-v2
merge into: shoko:main
shoko:main
shoko:fix/issue-254
shoko:fix/issue-252
shoko:fix/issue-248
shoko:fix/issue-246
shoko:fix/issue-165
shoko:fix/issue-158
shoko:fix/issue-244
shoko:fix/issue-121
shoko:fix/issue-119
shoko:fix/issue-166
shoko:fix/issue-229-pr-conflict-check
shoko:fix/issue-212
shoko:fix/issue-229-user-message-with-base-workflow
shoko:fix/issue-229-ensure-session-worktree-bug
shoko:fix/issue-168
shoko:fix/issue-225-cmd-continue-message-truncation
shoko:fix/issue-223-pr-creation-instructions
shoko:fix/issue-220-pm-context-enhancement
shoko:fix/issue-210-msg-file-race-condition
shoko:fix/issue-207-queue-daemon-set-debug-mode
shoko:fix/issue-message-encoding
shoko:fix/issue-118
shoko:fix/issue-daemon-worktree-session-handling
shoko:fix/issue-write-index-quoting
shoko:fix/issue-cmd-destroy-unbound-var
shoko:fix/issue-daemon-worktree-path-fix
shoko:fix/issue-syntax-error-372
shoko:fix/issue-187-start-forks-agent
shoko:fix/issue-185-worktree-wrong-directory
shoko:fix/issue-183-destroy-base-requires-target
shoko:fix/issue-181-get-repo-url-strips-user-org
shoko:fix/issue-179-worktree-path-doubled
shoko:fix/issue-176-extract-issue-ref
shoko:fix/issue-174-queue-daemon-crashes
shoko:fix/issue-172-init-script-wrong-session-ids
shoko:fix/issue-170-duplicate-update-queue
shoko:fix/issue-167-notification-bash
shoko:fix/issue-156-queue-fixes
shoko:fix/issue-155-queue-list-json
shoko:fix/issue-160-gitea-token-from-pm-agent
shoko:fix/issue-156
shoko:fix/issue-150
shoko:fix/issue-116-modularize-script
shoko:fix/issue-120
shoko:fix/issue-144-parse-issue-ref-format-v2
shoko:fix/issue-queue-daemon-excess-agents
shoko:fix/issue-142-process-queue-missing-parens
Labels
Clear labels
0.2.x-only
0.3.x-only
backport
blocked
bug
critical
feature
foundation
high
meta
need-review
parallelization
ready
ready-to-merge
task-delegation
Available only in 0.2.x, not backportable
Available only in 0.3.x, not backportable
Backport compatible - can merge to 0.1.x
Issue needs clarification/discussion before it can be worked on
Bug report
Critical - blocking parallel work
New feature
Foundation work for 0.2.x
High priority - workflow improvements
Meta issue tracking multiple issues
PR needs review before merge
Issues related to parallelization
Issue is clear and ready to be implemented/delegated
PR approved and ready to merge
Issues related to task-delegation
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: shoko/kugetsu#95
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Delete Branch "feat/issue-46-verbosity-v2"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Summary
Fix for #46 - PM Agent verbosity control.
Changes
Add
KUGETSU_VERBOSITYenvironment variable with three modes:total(default)verbosehybridImplementation
KUGETSU_VERBOSITYvariable with defaulttotalKUGETSU_VERBOSITYincmd_delegatewhen running opencodeUsage
PM agent skill should check
KUGETSU_VERBOSITYenv var and adjust posting behavior accordingly.Fixes #46
I think rather than letting the agent to decide what kind of behavior it should do, we should prompt it in pm agent markdown as the prior context to the agent.
maybe we have "verbosity" markdown file as context that we can always append to the agent context. so any delegate/start/continue will read current "verbosity", then pick verbosity markdown file accordingly and attached it to the agent context. what do you think? is that makes sense?
That makes sense! The current implementation is a bit naive - it just exports KUGETSU_VERBOSITY as an env var for the PM agent but doesnt actually use it to control behavior.
A better approach:
This way the agent gets explicit instructions based on verbosity level, rather than having to decide itself.
Ill update the PR with this approach. What verbosity levels would you suggest? e.g.:
Or perhaps: verbose, normal, minimal?
verbose, default, quiet. what do you think?
I like that naming scheme! Its more intuitive:
Currently the implementation uses total/medium/minimal. Want me to rename them to verbose/default/quiet?
I already pushed the markdown context approach implementation. Let me know if you want me to rename the levels.
lgtm
Pull request closed