Azure/SGLang output-cap rejection handling and retry logic
Cluster 78405 · agent. Data as of
Nothing open on GitHub. Fix merged in #115897 on 19 Sep 2026. Check it covers your case.
The short answer
- Known?
- Yes. 3 related items grouped here, 1 issue already closed.
- Fix event
- A fix was merged to main in PR #115897. Check it covers your case. No release data is tracked.
- Tracker
- The maintainers' tracker marks this
resolved, deciding PR #115897. That is separate from the fix event. - Still open
- Nothing open on GitHub.
These items address how the agent handles output-token-cap rejections from Azure and SGLang, improving retry behavior to reduce the cap rather than entering a compression loop. A related bug reports that Hermes sends an oversized max_tokens value regardless of config, also triggering compression loops.
Open issues0
Open PRs0
Fixes on record
-
#115897 fix(agent): Azure and SGLang output-cap rejections retry with a smaller cap instead of a compression loop; failure copy stops naming the removed model.max_tokens key (#78405, #83521, salvage #81289, #83558)
files:
agent/model_metadata.py,agent/turn_overflow.py,tests/agent/test_output_cap_parsing.py
1 PR closed without merging
From the close comments: merged via another change 1
-
#81289 fix(agent): parse Azure output-cap rejection phrasing (#78405) · merged via another change →
6d8a8be,#115897,#78405,#83521,#81289,#83558
Closed issues · 1
- Closed#78405Bug: Hermes sends max_tokens=65536 ignoring config, causes compression loop
How this cluster was grouped
-
Link evidence (
structural links throughout): every member is connected by structural evidence: shared changed files, or a closing/superseded reference. -
Pairwise link density
0.67: the share of item pairs that carry a link, counting a structural link in full and a relevance-only link at 0.36.1.00means every pair is linked.
Pick up a Lab task
Check whether a fix covers the open issues here, with your own agent. Bring back evidence for maintainers.
Open the LabAt a glance
- Subsystem
- agent
- Merged PRs
- #115897
- Closed issues
- 1
- Attempts
- 1 PR closed without merging
- Fix link checked
- 19 Sep 2026