2025-03-21 - 2026-03-21
Overview
8 issues closed from 1 user
Closed
#15 policy: track exact compaction lineage separately from snapshot generation time
Closed
#14 policy: add snapshot age and compaction cadence so reused snapshots do not become stale
Closed
#13 perf: skip snapshot retraining when policy_stats_live deltas are empty
Closed
#12 perf: avoid full replay rescans by folding policy_stats_live deltas into snapshot refreshes
Closed
#11 feat: make policy refresh atomic so replay stats, candidate policy, and eval runs share a consistent snapshot
Closed
#10 refactor: normalize chosen_plan taxonomy so plan templates are not overloaded with single_tool semantics
Closed
#9 feat: train execution-mode priors separately from plan-key priors in offline policy builder
Closed
#1 fix: align shadow controller execution-mode semantics for single_tool vs run_plan
16 issues created by 1 user
Opened
#1 fix: align shadow controller execution-mode semantics for single_tool vs run_plan
Opened
#2 refactor: split the orchestrator into policy-router, effect-executor, and observation-reducer modules
Opened
#3 feat: add cross-system causal edges to the typed tool graph for root-cause traversal
Opened
#4 feat: implement hierarchical memory tiers for semantic, episodic, procedural, and causal system knowledge
Opened
#5 feat: build an online-to-offline distillation loop for policy improvement beyond static sacred gating
Opened
#6 chore: emit structured counterfactual decision traces from the shadow meta-controller
Opened
#7 perf: replace repeated analysis.update / return-analysis loops with a single idempotent observation commit
Opened
#8 feat: bootstrap the system topology graph from infrastructure manifests and live inventory
Opened
#9 feat: train execution-mode priors separately from plan-key priors in offline policy builder
Opened
#10 refactor: normalize chosen_plan taxonomy so plan templates are not overloaded with single_tool semantics
Opened
#11 feat: make policy refresh atomic so replay stats, candidate policy, and eval runs share a consistent snapshot
Opened
#12 perf: avoid full replay rescans by folding policy_stats_live deltas into snapshot refreshes
Opened
#13 perf: skip snapshot retraining when policy_stats_live deltas are empty
Opened
#14 policy: add snapshot age and compaction cadence so reused snapshots do not become stale
Opened
#15 policy: track exact compaction lineage separately from snapshot generation time
Opened
#16 audit: carry compaction-lineage metadata into candidate policy and sacred gate artifacts