Issues

v0.6.25: every index run aborts with a retryable snapshot conflict (exit 7) ~100 conversations in; identical pages across fresh-DB runs

#401

· possibilities opened

on Aug 15, 2026

v0.6.24: no cass index run can complete — watchdog exit 70 mid-persist on a giant transcript, or an unbounded post-publish spin (48.6 GB RSS)

#400

· postoso opened

on Aug 15, 2026

Feature request: embed the build commit in `--version` — a main build and the tagged release print identical version strings

#399

· StartupBros opened

on Aug 14, 2026

Feature request: a binary-only self-test — `health` exits 1 on a merely stale archive, so it cannot gate a binary

#398

· StartupBros opened

on Aug 14, 2026

Readiness probe reports `Search usable now: yes` and `db=available` while every search fails on an unopenable archive (v0.6.24)

#396

· ericziko opened

on Aug 14, 2026

TUI hangs on startup with large archive (15k sessions / 2.1M messages / 8.5GB DB)

#395

· shunkakinoki opened

on Aug 12, 2026

v0.6.24: `index --semantic` re-embeds the entire corpus on every run — watermark and manifest fingerprint are written but never checked (watch-once path does the same refresh in 25s)

#394

· burp opened

on Aug 12, 2026

v0.6.23 macOS: post-reboot st_dev change leaves namespace sidecars stale; archive-opening commands fail and doctor reports high data-loss risk

#393

· jonesmabea opened

on Aug 11, 2026

Recurring btree corruption in the conversations table (root 11): rowid-out-of-order written during normal indexing, twice in three weeks, across v0.6.13- and v0.6.22-written databases

#391

· aatosrutiini opened

on Aug 10, 2026

doctor integrity_check false positives: "page N is never used" on databases stock SQLite 3.51.2 calls ok — the ≥1 GiB fresh-build class is fixed by v0.6.23's checker, but a second class on a long-lived DB survives

#390

· aatosrutiini opened

on Aug 10, 2026

Unclean host shutdown leaves a dirty WAL that every readonly-first surface refuses (CANTOPEN); doctor --fix escalates to a full 39 GB reconstruct when a 6-second WAL checkpoint repairs the archive

#389

· craigvandotcom opened

on Aug 10, 2026

feat(connectors): index Prime Agent sessions from ~/.prime/agent/sessions

#388

· DanielsLoud opened

on Aug 9, 2026

Navigation Menu

All issues

v0.6.25: every index run aborts with a retryable snapshot conflict (exit 7) ~100 conversations in; identical pages across fresh-DB runs

v0.6.24: no cass index run can complete — watchdog exit 70 mid-persist on a giant transcript, or an unbounded post-publish spin (48.6 GB RSS)

Feature request: embed the build commit in `--version` — a main build and the tagged release print identical version strings

Feature request: a binary-only self-test — `health` exits 1 on a merely stale archive, so it cannot gate a binary

Readiness probe reports `Search usable now: yes` and `db=available` while every search fails on an unopenable archive (v0.6.24)

TUI hangs on startup with large archive (15k sessions / 2.1M messages / 8.5GB DB)

v0.6.24: `index --semantic` re-embeds the entire corpus on every run — watermark and manifest fingerprint are written but never checked (watch-once path does the same refresh in 25s)

v0.6.23 macOS: post-reboot st_dev change leaves namespace sidecars stale; archive-opening commands fail and doctor reports high data-loss risk

Recurring btree corruption in the conversations table (root 11): rowid-out-of-order written during normal indexing, twice in three weeks, across v0.6.13- and v0.6.22-written databases

doctor integrity_check false positives: "page N is never used" on databases stock SQLite 3.51.2 calls ok — the ≥1 GiB fresh-build class is fixed by v0.6.23's checker, but a second class on a long-lived DB survives

Unclean host shutdown leaves a dirty WAL that every readonly-first surface refuses (CANTOPEN); doctor --fix escalates to a full 39 GB reconstruct when a 6-second WAL checkpoint repairs the archive

feat(connectors): index Prime Agent sessions from ~/.prime/agent/sessions

Search results