Releases: NodeJSmith/claude-code-recall
Releases · NodeJSmith/claude-code-recall
Release list
v0.25.0
v0.24.2
v0.24.1
Hooks
- fix the SessionStart hook loading the full embedding stack (numpy/fastembed/onnxruntime) just to check background-job status, adding unnecessary startup latency (#200)
Backfill
- fix summary backfill getting stuck retrying the same bad conversation branch forever instead of skipping it (#202)
- fix embeddings backfill getting stuck on one bad conversation branch, blocking every branch behind it from ever being embedded (#204)
Bug Fixes
- fix a crash in task/teammate-notification detection when a transcript message's
textfield was present but null (#198) - fix a false-positive "unresolved question" alert triggered by a system-generated notice that happened to quote a command tag (#198)
- fix import silently reporting success while a disk or database error caused it to skip every remaining file in a project (#198)
- fix a race between a manual backfill and the automatic sync that could hide a real "embeddings unavailable" alert (#198)
- correct PyPI package metadata to list only the platforms ccrecall actually supports (Linux, macOS) (#198)
v0.24.0
Search
searchandsearch-messagesnow accept the query positionally (ccrecall search "auth bug"), alongside the existing--query/-q;--max-resultscan also go up to 20 instead of 10 (#191)
Sync
- fix an import crash where one malformed transcript file could permanently block every transcript sorting after it from ever being imported again (#194)
- fix crashes on transcripts containing a null message or a null text field, instead of producing garbage output (#193)
- fix conversation branch links silently failing to write, which could lose part of a conversation's history with no error (#188)
Hooks
- fix the session-start hook crashing with no warning, instead of showing its usual alert, when
~/.ccrecallis unwritable (#187) - fix a stale "embeddings unavailable" alert getting cleared by an unrelated sync, hiding a real embedding-model problem (#186)
- fix the session-start summary sometimes showing the same exchanges twice, or a bogus "N earlier exchanges" gap (#189)
- fix the "unresolved decision" warning still appearing after you already moved on by running a slash command (#192)
Performance Improvements
- sync no longer re-checks every already-synced message for missing tool content on every run, cutting redundant work on long sessions (#190)
Bug Fixes
- fix
ccrecall recent --sort-order ascreturning the oldest sessions in the database instead of the most recent sessions in oldest-first order (#185)
v0.23.0
Breaking Changes
- the opt-in LLM summary enrichment subsystem is removed entirely:
ccrecall backfill llm-summaries, theccrecall-llm-summariesconsole script, and its sixllm_summary_*config keys are gone, anddisplay_title/summary_previewno longer appear inccrecall --json search/search-messagesoutput. Unknown keys left over inconfig.jsonare ignored, so stale config entries are harmless (#134)
Bug Fixes
- surface previously-silent parsing, database, and hook failures to the per-process log files instead of failing silently (#154)
- fix a memory spike in sync-current that could freeze the machine (7.5 GB observed) by capping sync-path embedding at 4096 tokens, dropping the worst-case peak to ~4 GB (#166)
- fix
exclude_projectsfailing to match hyphenated project names on the lossy project-name fallback path (#135) - fix the next schema migration bricking every database that has ever produced embeddings, by loading the sqlite-vec extension before every migration instead of only the first one; also fix an infinite loop on a corrupted transcript, and a non-atomic handoff-file write that could lose the handoff on a fresh machine (#132)
- warn when a database's schema version is ahead of what the running code expects, instead of silently treating it as fully migrated (#167)
Refactoring
- drop the redundant
--nlong-form alias fromsearch/search-messages/tail(keep-n); renamerecent's--nto--limit; add--verbosetosearch-messagesand--dbto allbackfillsubcommands (#150)
v0.22.1
Bug Fixes
- fix false-positive capability check failures caused by unrelated, concurrent Claude Code sessions (#112)
v0.22.0
Features
- add opt-in Claude-powered Branch Resume Brief enrichment (#106)
v0.21.1
Bug Fixes
- recover scoped vector matches beyond the initial KNN window (#102)
Performance Improvements
- cache confirmed-OK ingestion checks for faster repeated status audits (#105)
2026-08-24
- fix a memory spike in sync-current that could freeze the machine (7.5 GB observed) by capping sync-path embedding at 4096 tokens, dropping the worst-case peak to ~4 GB (#166)
- warn when a database's schema version is ahead of what the running code expects, instead of silently treating it as fully migrated (#167)
2026-08-19
- surface previously-silent parsing, database, and hook failures to the per-process log files instead of failing silently (#154)
2026-08-14
- fix a crash on upgrading pre-v4 databases caused by an index referencing a column added by a later migration (#136)
2026-08-09
- fix false-positive capability check failures caused by unrelated, concurrent Claude Code sessions (#112)
2026-08-08
- add opt-in Claude-powered Branch Resume Brief enrichment (#106)
- harden transcript discovery and capability-gated enrichment fallback (#106)
2026-08-04
v0.21.0
Breaking Changes
- the deprecated
ccrecall statscommand is removed; its CLI wrapper never forwarded--json/--days/--check-ingestion(it always printed markdown regardless of the global--jsonflag), andccrecall statusalready provides everything it did. Useccrecall statusinstead (#97)
Features
- add
--before/--afterdate filters tosearch/search-messages(#99)
v0.20.0
Features
- add a consolidated ingestion status report (#93)