{"repo":"NousResearch/hermes-agent","public_id":"78405","display_label":"Azure/SGLang output-cap rejection handling and retry logic","summary":"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.","subsystem":"agent","counts":{"members":3,"open_issues":0,"open_issues_distinct":0,"open_prs":0,"closed_issues":1,"merged_prs":1,"attempt_prs":1},"landed_fix":{"status":"merged_to_main","basis":"structural","pr":115897,"merged_at":"2026-09-19T18:30:55+00:00"},"resolution":{"state":"resolved","by":"pr-115897"},"freshness":{"state":"verified","last_verified_at":"2026-09-19T18:30:55+00:00","is_stale":false},"grouping":{"backbone":"STRUCTURAL_SPANNING","link_density":0.6666666666666666},"fix_edges":[{"issue_number":78405,"pr_number":115897,"tier":"verified","basis":"closing_ref","confidence":0.95,"recoverable_from_timeline":true,"pr_merged_at":"2026-09-19T18:30:55+00:00"},{"issue_number":83521,"pr_number":115897,"tier":"verified","basis":"closing_ref","confidence":0.95,"recoverable_from_timeline":true,"pr_merged_at":"2026-09-19T18:30:55+00:00"}],"members":[{"kind":"issue","number":78405,"title":"Bug: Hermes sends max_tokens=65536 ignoring config, causes compression loop","url":null,"state":"closed"},{"kind":"pr","number":81289,"title":"fix(agent): parse Azure output-cap rejection phrasing (#78405)","url":null,"state":"closed","close_reason":{"refs":[{"kind":"commit","value":"6d8a8bebf70b","number":0},{"kind":"pr","value":"#115897","number":115897},{"kind":"pr","value":"#78405","number":78405},{"kind":"pr","value":"#83521","number":83521},{"kind":"pr","value":"#81289","number":81289},{"kind":"pr","value":"#83558","number":83558}],"category":"merged_via","implication":"superseded"},"merged":false},{"kind":"pr","number":115897,"title":"fix(agent): Azure and SGLang output-cap rejections retry with a smaller cap instead of a compression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)","url":null,"state":"merged","close_reason":{"category":"merged"},"merged":true,"is_fix_pr":true,"fix_tier":"verified","files_changed":["agent/model_metadata.py","agent/turn_overflow.py","tests/agent/test_output_cap_parsing.py"]}],"trust":{"default_tier":"verified","includes_suggested":false}}