Releases: Opencode-DCP/opencode-dynamic-context-pruning
Releases · Opencode-DCP/opencode-dynamic-context-pruning
v2.1.7 - Prune origin tracking and sync improvements
What's Changed
- Track prune origins to improve undo reconciliation behavior
- Add sync message flow and rename reconcile to sync origins
- Prevent compress tool summary modification during prune
- Add write/edit to default protected tools and skip already-pruned IDs
Full Changelog: v2.1.6...v2.1.7
v2.1.6 - SQLite migration and prune hardening
What's Changed
- Migrated opencode helper scripts from API-backed reads to sqlite-backed access for session/message tooling.
- Added malformed-summary guards in pruning logic to prevent invalid compress operations.
- Bumped package version metadata to 2.1.6 for release.
Full Changelog: v2.1.5...v2.1.6
v2.1.5 - Compress permission gating update
What's Changed
- Gate message ID injection when compress tool permission is denied.
- Update message injection and prompt handling for compress permission checks.
- Bump package version to 2.1.5.
Full Changelog: v2.1.4...v2.1.5
v2.1.4 - Compress ID boundary migration and stability fixes
What's Changed
- Migrated
compressto use message ID boundaries and block placeholders, improving reliability while preserving context links during compression. - Added deterministic synthetic injection IDs to make compression targeting stable across turns.
- Bug fixes
Contributors
Full Changelog: v2.1.3...v2.1.4
v2.1.3 - compress summary handling fix
v2.1.2 - Bug fixes for compress tool
v2.1.1 - Prompt guidance update
What's Changed
- Improve prompt guidance for reliable compress boundary matching.
Full Changelog: v2.1.0...v2.1.1
v2.1.0 - Manual mode, model-specific limits, and compress fixes
What's Changed
New Features
- Manual Mode Controls - New manual mode for context management tools, giving users direct control over when DCP tools run. Includes
/dcpcommands with focus arguments, ephemeral trigger prompts via chat transform, and configurable manual mode settings - Model-Specific Context Limits - New
toolSettings.modelLimitsconfig allows per-model context limits as absolute token values or percentages, withcontextLimitas fallback (#366)
Fixes
- Compress boundary matching - Fix compress tool to correctly search transformed message views for start/end boundary strings (#375)
- Compress token calculation - Fix bugs related to compress tool calculating tokens incorrectly
- Compress multi-match error - Fix compress tool incorrectly finding multiple start and end strings then erroring out
- Map null assertions - Fix unsafe non-null assertions in Map operations (#369)
Improvements
- Refactor: use pre-computed token counts in prune state instead of re-scanning messages
- Refactor: simplify manual trigger return type for cleaner control interface
- Refactor: internalize manual-mode guards into strategies
- Docs: add manual mode config and commands to README
Contributors
Full Changelog: v2.0.2...v2.1.0
v2.0.2 - Fuzzy compress matching improvements
v2.0.1 - Percentage contextLimit support
What's Changed
Features
- Percentage contextLimit support - Configure
contextLimitwith percentage values like"60%"instead of just absolute token counts
Improvements
- Default
contextLimitnow set to60%for optimal context management - Default
compresspermission changed todeny, this tool is experimental for now
Breaking Changes
- Removed
"model"option fromcontextLimitconfiguration - use percentage or absolute values instead
Contributors
Full Changelog: v2.0.0...v2.0.1