v0.4.1

Fixes

• existing archives that already report schema version 2 now self-heal missing embedding tables and columns before 0.4.x sync/update commands continue.

v0.4.0

Changes

• semantic message search now ranks across the full compatible local vector set instead of only the newest candidate window. (#36) Thanks @GaosCode.
• hybrid message search now fuses FTS with local semantic vectors while avoiding embedding-provider calls when no local vectors exist. (#37) Thanks @GaosCode.
• local embedding providers now support OpenAI-compatible endpoints, Ollama, and llama.cpp, and doctor can probe the configured provider before you queue vectors
• embed now drains the queued embedding backlog in bounded batches, requeues safely on provider throttling, and drops stale stored vectors when messages no longer have embeddable content
• Git snapshot publishing can now opt in to backing up generated embedding vectors with --with-embeddings while still keeping embedding queue state local.
• Git-backed snapshot imports are now much faster on large archives by using import-only SQLite pragmas and bulk-load FTS5 settings during search index rebuilds
• messages and mentions now use composite read-path indexes so larger archives spend less time sorting/filtering common guild, channel, and author queries

Fixes

• normalized message text is now sanitized before it reaches SQLite and FTS5, repairing malformed UTF-8 and stripping invisible/control-character noise that can poison search content
• Git-backed snapshots now keep embedding queue state and generated vectors local to each archive, so subscribers no longer inherit misleading embedding backlog metadata. (#38) Thanks @GaosCode.

Docs

• docs now cover semantic and hybrid search setup, embedding privacy, Git snapshot behavior, and local vector rebuilds. (#39) Thanks @GaosCode.

Tests

• Git embedding snapshot export/import now has CLI, share-package, and Docker E2E coverage.
• total Go test coverage now reaches the 85% line.

v0.3.0

• sync --all now bypasses default_guild_id so one run can fan out across every discovered guild without clearing the single-guild default first
• sync --full no longer aborts when forum thread discovery hits Discord 403 Missing Access; inaccessible channels are skipped and marked unavailable while accessible channels continue syncing
• startup now validates and stamps SQLite schema version via PRAGMA user_version, and fails fast if the local DB schema is newer than the running binary
• git-backed archive sharing can now export/import compressed JSONL snapshots with manifests, subscribe to a Git repo as the data source, and run in git-only mode without Discord credentials
• messages, search, and reports can automatically refresh stale git-backed data, preferring the Git snapshot before falling back to live Discord when both sources are configured
• the Discord backup publisher workflow now syncs latest messages, publishes the archive to a private GitHub repo, serializes concurrent runs, validates required secrets, and skips the member crawl for faster updates
• the backup report workflow now updates README activity stats from the backup action and keeps those queries bounded with process timeouts
• sync --latest-only adds a lightweight refresh path for checking recent Discord messages without doing a full historical crawl
• repository imports now skip expensive rebuilds when the snapshot manifest is already current, and GitHub Actions persist the warmed SQLite database across runs
• the Docker git-source smoke test now verifies that a fresh install can subscribe to a repository-only archive and query messages, SQL, and reports
• CI now uses Go 1.26.2, actions/setup-go 6.4.0, cache actions 5.0.5, Node 24 for report generation, and refreshed SQLite dependencies

v0.2.0

• much faster sync --full behavior on large archives: incomplete backfills are auto-batched, active-thread discovery is more precise, and steady-state refreshes avoid re-scanning every archived thread once history is already complete
• sync --since now reliably honors the cutoff during bootstrap and full-history backfill, while still allowing a later sync --full without --since to continue older history
• full-sync progress is more resilient: slow member crawls no longer hold message sync hostage, and stale unavailable-channel markers are cleared so recovered channels can sync again
• offline member-profile search is now much richer: members search matches archived profile fields in addition to names
• members show now accepts either Discord IDs or queries and can include recent messages plus message stats for the resolved member
• archived profile extraction now surfaces stored fields like bio, pronouns, location, website, x, github, and discovered URLs when present
• messages --sync can do a blocking pre-query refresh for the matching channel or guild scope before reading the local archive
• messages --hours adds recent-hour slices without manual RFC3339 timestamps
• messages --last returns the newest matching rows while still printing them oldest-to-newest

v0.1.0

• initial public release of discrawl
• multi-guild Discord crawler with single-guild default UX
• local SQLite archive with FTS5 search
• commands: init, sync, tail, search, messages, mentions, sql, members, channels, status, doctor
• env-based bot token discovery
• resumable full-history sync, live gateway tailing, repair sync loop, targeted channel sync
• attachment-text indexing for small text-like uploads
• structured user and role mention indexing/querying
• empty-message filtering based on real searchable/displayable content instead of raw body only
• CI with lint, tests, secret scanning, and coverage enforcement
• release plumbing via GoReleaser, GitHub Actions, and Homebrew tap packaging
• sync correctness fixes for empty channels, inaccessible channels, unknown channels, and large-channel resume behavior
• SQLite/FTS performance fixes for backfill throughput and lower write amplification