yonlu/yellowjacket
1
1
Fork 0
Code Issues 48 Pull Requests Actions Packages Projects Releases 12 Wiki Activity

loop: model fallback chain and foreground launches #243

New Issue
Closed
opened 2026-09-04 03:20:11 +00:00 by logan · 1 comment
logan commented 2026-09-04 03:20:11 +00:00
Collaborator
Copy Link
Copy Source

Two findings from the #31 tick, both of which would strand an unattended run:

  1. Model quota is a leg-level failure, and the fallback chain was undocumented. The worker (pinned qwen/deepseek-v4-pro-0813) hit the qwen weekly 429 mid-tick. The obvious fallback deepseek/deepseek-v4-pro is wrong: the deepseek provider has no models, only catalog overrides, and fails silently (empty artifact, no session). The model actually lives on the go gateway as go/deepseek-v4-pro (PAYG, what the orchestrator session itself runs on); go/glm-5.3-flash is the next rung down and is what completed the tick. The skill must state: qwen 429 → go/deepseek-v4-pro → go/glm-5.3-flash, with the substitution journaled, and the T3 escalation model never spent on a quota substitution.
  2. Foreground subagent launches are the reliable mode here. The async runner died once without persisting a child session (nothing to resume), and emitted spurious "needs attention" nudges on runs that were already complete. The skill should say: launch legs in the foreground; a worker that dies mid-leg leaves uncommitted work, so recover by inspecting the tree and relaunching to complete, never to re-implement.
Two findings from the #31 tick, both of which would strand an unattended run: 1. **Model quota is a leg-level failure, and the fallback chain was undocumented.** The worker (pinned `qwen/deepseek-v4-pro-0813`) hit the qwen weekly 429 mid-tick. The obvious fallback `deepseek/deepseek-v4-pro` is wrong: the `deepseek` provider has no models, only catalog overrides, and fails silently (empty artifact, no session). The model actually lives on the `go` gateway as `go/deepseek-v4-pro` (PAYG, what the orchestrator session itself runs on); `go/glm-5.3-flash` is the next rung down and is what completed the tick. The skill must state: qwen 429 → `go/deepseek-v4-pro` → `go/glm-5.3-flash`, with the substitution journaled, and the T3 escalation model never spent on a quota substitution. 2. **Foreground subagent launches are the reliable mode here.** The async runner died once without persisting a child session (nothing to resume), and emitted spurious "needs attention" nudges on runs that were already complete. The skill should say: launch legs in the foreground; a worker that dies mid-leg leaves uncommitted work, so recover by inspecting the tree and relaunching to *complete*, never to re-implement.
logan added the Kind/Bug
Priority
Low
4
 labels 2026-09-04 03:20:11 +00:00
logan self-assigned this 2026-09-04 03:20:17 +00:00
logan added the
Status
In Progress
 label 2026-09-04 03:20:17 +00:00
logan commented 2026-09-04 03:20:17 +00:00
Author
Collaborator
Copy Link
Copy Source

⟦loop⟧ P1 finding, taking immediately. Branch fix/243-model-fallback.

Approach: two notes in .pi/skills/yj-loop/SKILL.md — a model fallback chain under the legs table, and a foreground-launch rule. No code.

⟦loop⟧ P1 finding, taking immediately. Branch `fix/243-model-fallback`. Approach: two notes in `.pi/skills/yj-loop/SKILL.md` — a model fallback chain under the legs table, and a foreground-launch rule. No code.
logan closed this issue 2026-09-04 03:37:29 +00:00
gitea-actions bot removed the
Status
In Progress
 label 2026-09-04 03:37:40 +00:00
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
Area/Design
Area/Downloads
Area/Explore
Area/Library-UI
Area/Metadata
Area/Packaging
Area/Player
Area/Queue
Area/Settings
Area/Shell-Nav
Compat/Breaking
Breaking change that won't be backward compatible
 Kind/Bug
Something is not working
 Kind/Documentation
Documentation changes
 Kind/Enhancement
Improve existing functionality
 Kind/Feature
New functionality
 Kind/Security
This is security issue
 Kind/Testing
Issue or pull request related to testing
 Platform/Android
Platform/Desktop
Priority
Critical
1
The priority is critical
Priority
High
2
The priority is high
Priority
Medium
3
The priority is medium
Priority
Low
4
The priority is low
Reviewed
Confirmed
1
Issue has been confirmed
Reviewed
Duplicate
2
This issue or pull request already exists
Reviewed
Invalid
3
Invalid issue
Reviewed
Won't Fix
3
This issue won't be fixed
Status
Blocked
1
Something is blocking this issue or pull request
Status
Need More Info
2
Feedback is required to reproduce issue or to continue work
Status
Abandoned
3
Somebody has started to work on this but abandoned work
Status
In Progress
Somebody is actively working on this right now
No labels Kind/Bug
Priority
Low
4
Milestone
No items
No Milestone
Projects
Clear projects
No projects
Assignees
Clear assignees
logan yonlu
No Assignees logan
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: yonlu/yellowjacket#243