Tags

v0.16.0

opencode-memory 0.16.0

v0.15.0

opencode-memory 0.15.0

Ships the weak/free-model compaction mitigation (#158, MRs !140+!141):
- Model-class classification (cost.input===0) gating all mitigations
- Anti-confabulation compaction guard for weak models
- Post-compaction ground-truth re-injection from memory
- get_ground_truth server tool (decisions independent of boot toggle)

v0.14.0

Release v0.14.0

v0.13.0

Release v0.13.0

Watchers: gitlab_entity kind for event-conditional queue deferment (#141).
Reconciles the missing tag for the 0.13.0 CHANGELOG entry.

v0.13.1

Release v0.13.1

Opt-in auto-updates with tiered consent, MCP tool-surface change detection,
and rollback (#152).

v0.12.0

v0.12.0

Reconciled release folding the never-tagged 0.11.1-0.11.5 versions and the
previously unreleased 0.12.0 feature set into a single tagged release. See
CHANGELOG.md for the full 0.12.0 entry.

v0.11.0

v0.11.0

Secret-guard hybrid reversible redaction, background job visibility,
OpenCode DB lifecycle maintenance (#120), memory_status operational
stats (#121), runaway-I/O fixes (#117 #118 #119), datetime tz
normalization, and /stats performance (catches PyPI up from 0.10.0).

v0.10.0

v0.10.0 - health-gated background-worker autoscaler (#112) + queue due-dates, lazy startup, bg-subagent write-capability fixes

v0.9.2

v0.9.2 - comprehensive event-loop offload fix + regression guard tests (ruff ASYNC + event-loop safety tests)

v0.9.1

v0.9.1 - reliability patch: get_context SAME_ENTITY links (!29), /stats fast+non-blocking (!24), /stats+/metrics off event loop (!30)

v0.9.0

Release v0.9.0

v0.8.5

v0.8.4

v0.8.4: recall detuning + get_by_id tool

v0.8.3

v0.8.3: fix concurrent subagent MCP connection contention

v0.8.2

release: v0.8.2 - injection effectiveness tracking and workspace auto-init

- Injection effectiveness analysis module for tracking context impact
- Auto-register workspace on boot context (implicit workspace initialization)
- Relevance filtering with MIN_RELEVANCE_THRESHOLD (0.35)
- Heuristic and LLM-based reanalysis for injection quality

v0.8.1

release: v0.8.1 - injection effectiveness tracking

- Injection effectiveness analysis module for tracking context impact
- Auto-register workspace on boot context (implicit workspace initialization)
- Relevance filtering with MIN_RELEVANCE_THRESHOLD (0.35)
- Heuristic and LLM-based reanalysis for injection quality

v0.8.0

v0.8.0 - GitLab bulk ingest

v0.7.1

v0.7.1 - May 2026 audit fixes (all 75 findings resolved)

v0.7.0

v0.6.1

v0.6.1 - Documentation alignment and fixes

- Documentation alignment: 17 fixes across all docs
- Updated BOOTSTRAP.md for auto-injection vs manual workflow
- Added HTTP endpoints section to ARCHITECTURE.md
- Added missing tool tables (Reminders, Knowledge Graph, Code Indexing, GitLab Spider, Analytics)
- Documented unified context_events table migration
- Updated tool count from 25+ to 45+
- Various doc fixes for Python version, memory age format, SKIP_BOOTSTRAP env var

See v0.6.0 for the main feature release.