{"repo":"NousResearch/hermes-agent","public_id":"513","display_label":"Observation masking and two-phase context compression proposals","summary":"These members propose reducing token usage in conversation history by masking or pruning older tool outputs before or instead of full context compaction. The ideas appear across multiple feature requests and implementation attempts addressing the same core mechanism.","subsystem":"agent","counts":{"members":5,"open_issues":0,"open_issues_distinct":0,"open_prs":0,"closed_issues":2,"merged_prs":0,"attempt_prs":3},"landed_fix":null,"resolution":{"state":"unresolved","by":null},"freshness":{"state":"never_verified","last_verified_at":null,"is_stale":false},"grouping":{"backbone":"STRUCTURAL_PARTIAL","link_density":0.33599999999999997},"fix_edges":[],"members":[{"kind":"issue","number":513,"title":"Feature: Two-Phase Context Management — Prune Tool Outputs Before Full Compaction (inspired by Kilocode)","url":null,"state":"closed"},{"kind":"issue","number":2046,"title":"[Feature]: Observation masking for older turns: reduce conversation history tokens without LLM summarization","url":null,"state":"closed"},{"kind":"pr","number":588,"title":"feat(agent): two-phase context compression — tool pruning + structured compaction","url":null,"state":"closed","close_reason":{"refs":[{"kind":"pr","value":"#1231","number":1231}],"category":"implemented_elsewhere","implication":"superseded"},"merged":false},{"kind":"pr","number":2093,"title":"feat(context): add observation masking to reduce token usage","url":null,"state":"closed","close_reason":{"category":"partial_fix","implication":"needs_llm"},"merged":false,"is_fix_pr":true,"fix_tier":"suggested","fix_detail_withheld":true},{"kind":"pr","number":2245,"title":"feat(context): observation masking — zero-cost context reduction for older tool outputs","url":null,"state":"closed","close_reason":{"refs":[{"kind":"pr","value":"#2323","number":2323}],"category":"superseded_by_pr","implication":"superseded"},"merged":false}],"trust":{"default_tier":"verified","includes_suggested":false}}