{"repo":"NousResearch/hermes-agent","public_id":"101898","display_label":"Blank reply streaks bypass retries and logging","summary":"These items address a defect where empty or usage-absent responses skip the fast-fail path, exhaust the full retry ladder unnecessarily, and leave no per-attempt log entry. A related concern is that the empty sentinel can surface to the user as a blank message bubble.","subsystem":null,"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":102051,"merged_at":"2026-09-03T10:41:53+00:00"},"resolution":{"state":"resolved","by":"pr-102051"},"freshness":{"state":"verified","last_verified_at":"2026-09-03T10:41:53+00:00","is_stale":false},"grouping":{"backbone":"STRUCTURAL_SPANNING","link_density":0.6666666666666666},"fix_edges":[{"issue_number":101898,"pr_number":102051,"tier":"verified","basis":"closing_ref","confidence":0.95,"recoverable_from_timeline":true,"pr_merged_at":"2026-09-03T10:41:53+00:00"}],"members":[{"kind":"issue","number":101898,"title":"agent: usage-absent empty streaks skip deterministic fast-fail and log no per-attempt record; \"(empty)\" sentinel reaches the desktop as a blank bubble","url":null,"state":"closed"},{"kind":"pr","number":101921,"title":"fix(agent): stop blank replies from exhausting full retry ladders","url":null,"state":"closed","close_reason":{"refs":[{"kind":"commit","value":"c3e9defd18","number":0},{"kind":"pr","value":"#102051","number":102051},{"kind":"pr","value":"#101898","number":101898}],"category":"merged_via","implication":"superseded"},"merged":false},{"kind":"pr","number":102051,"title":"fix(agent): usage-less empty streaks stop after two attempts and stay logged (#101898, salvage #101921)","url":null,"state":"merged","close_reason":{"category":"merged"},"merged":true,"is_fix_pr":true,"fix_tier":"verified","files_changed":["agent/conversation_loop.py","agent/empty_response_guard.py","tests/agent/test_empty_response_guard.py","tests/run_agent/test_run_agent.py"]}],"trust":{"default_tier":"verified","includes_suggested":false}}