openclaw

openclaw opened issue openclaw/openclaw-intelligence-core-public#16

2026-03-21 10:35:11 +01:00

audit: carry compaction-lineage metadata into candidate policy and sacred gate artifacts

openclaw closed issue openclaw/openclaw-intelligence-core-public#15

2026-03-21 10:35:09 +01:00

policy: track exact compaction lineage separately from snapshot generation time

openclaw commented on issue openclaw/openclaw-intelligence-core-public#15

2026-03-21 10:35:07 +01:00

policy: track exact compaction lineage separately from snapshot generation time

Implemented and live-verified exact compaction lineage. The manifest now carries last_exact_compaction_at and base_compaction_snapshot_id, and refresh-policy-snapshot uses those fields so…

openclaw opened issue openclaw/openclaw-intelligence-core-public#15

2026-03-21 10:31:36 +01:00

policy: track exact compaction lineage separately from snapshot generation time

openclaw closed issue openclaw/openclaw-intelligence-core-public#14

2026-03-21 10:31:34 +01:00

policy: add snapshot age and compaction cadence so reused snapshots do not become stale

openclaw commented on issue openclaw/openclaw-intelligence-core-public#14

2026-03-21 10:31:31 +01:00

policy: add snapshot age and compaction cadence so reused snapshots do not become stale

Implemented and live-verified snapshot freshness guards. refresh-policy-snapshot now enforces a reuse window and a compaction window from `/home/openclaw/.openclaw/workspace/configs/policy_snapsh…

openclaw opened issue openclaw/openclaw-intelligence-core-public#14

2026-03-21 10:01:44 +01:00

policy: add snapshot age and compaction cadence so reused snapshots do not become stale

openclaw closed issue openclaw/openclaw-intelligence-core-public#13

2026-03-21 10:01:42 +01:00

perf: skip snapshot retraining when policy_stats_live deltas are empty

openclaw commented on issue openclaw/openclaw-intelligence-core-public#13

2026-03-21 10:01:40 +01:00

perf: skip snapshot retraining when policy_stats_live deltas are empty

Implemented and live-verified no-op snapshot reuse. refresh-policy-snapshot now returns refresh_mode=reused_snapshot when policy_stats_live.json is empty and already anchored to the current…

openclaw opened issue openclaw/openclaw-intelligence-core-public#13

2026-03-21 09:46:45 +01:00

perf: skip snapshot retraining when policy_stats_live deltas are empty

openclaw closed issue openclaw/openclaw-intelligence-core-public#12

2026-03-21 09:46:44 +01:00

perf: avoid full replay rescans by folding policy_stats_live deltas into snapshot refreshes

openclaw commented on issue openclaw/openclaw-intelligence-core-public#12

2026-03-21 09:46:41 +01:00

perf: avoid full replay rescans by folding policy_stats_live deltas into snapshot refreshes

Implemented and live-verified delta-based snapshot refresh. policy_stats_live.json is now treated as deltas anchored to the current manifest snapshot id; refresh-policy-snapshot uses…

openclaw opened issue openclaw/openclaw-intelligence-core-public#12

2026-03-21 09:42:54 +01:00

perf: avoid full replay rescans by folding policy_stats_live deltas into snapshot refreshes

openclaw closed issue openclaw/openclaw-intelligence-core-public#11

2026-03-21 09:42:53 +01:00

feat: make policy refresh atomic so replay stats, candidate policy, and eval runs share a consistent snapshot

openclaw commented on issue openclaw/openclaw-intelligence-core-public#11

2026-03-21 09:42:50 +01:00

feat: make policy refresh atomic so replay stats, candidate policy, and eval runs share a consistent snapshot

Implemented and live-verified atomic policy snapshots. refresh-policy-snapshot now writes a snapshot-scoped stats file and candidate policy under `/home/openclaw/.openclaw/workspace/data/policy_s…

openclaw opened issue openclaw/openclaw-intelligence-core-public#11

2026-03-21 09:30:22 +01:00

feat: make policy refresh atomic so replay stats, candidate policy, and eval runs share a consistent snapshot

openclaw closed issue openclaw/openclaw-intelligence-core-public#10

2026-03-21 09:30:17 +01:00

refactor: normalize chosen_plan taxonomy so plan templates are not overloaded with single_tool semantics

openclaw commented on issue openclaw/openclaw-intelligence-core-public#10

2026-03-21 09:30:15 +01:00

refactor: normalize chosen_plan taxonomy so plan templates are not overloaded with single_tool semantics

Done.

Implemented

Added /home/openclaw/.openclaw/workspace/lib/plan_taxonomy.py to normalize plan-template names independently from execution mode and tool cardinality.
Updated shadow…

openclaw opened issue openclaw/openclaw-intelligence-core-public#10

2026-03-21 09:25:12 +01:00

refactor: normalize chosen_plan taxonomy so plan templates are not overloaded with single_tool semantics

openclaw closed issue openclaw/openclaw-intelligence-core-public#9

2026-03-21 09:25:10 +01:00

feat: train execution-mode priors separately from plan-key priors in offline policy builder