discrawl: vanhemmat versiot ja versiohistoria

0.17.0 - 2026-10-23
Highlights: Keep live Discord archives complete, protect public snapshots from excluded channels, and accept protected credential files for unattended services.
- Reconcile bounded exact-message and metadata failures during ordinary repair, expose failures and resumable text-repair progress in status, and preserve vectors when text is unchanged.
- Capture supported bulk message deletions, channel/thread tombstones, role changes and thread-list metadata; include voice/stage text in scoped REST repair.
- Open every archive and metrics writer connection with FULL synchronization and macOS full-fsync checkpoints, including pool replacements, while keeping read-only connections unchanged.
- Retain channel scope metadata independently of content, fail closed on unknown ancestry, and protect public snapshots from excluded ancestors; resolve allowed new channels without a restart.
- Honor channel and ancestor tombstones before page, exact-message and attachment requests even without configured exclusions, resolve every scope rejection as a policy outcome, and count only actionable attachment failures in status. Desktop cache imports no longer clear channel tombstones, and text repair keeps author and channel names in multilingual lexical indexes.
- Preserve attributed embed fields and whitespace boundaries in derived text; add durable attachment extraction status and retain successful text through transient failures.
- Settle channel message failures answered Unknown Channel only after an exact channel lookup corroborates them, as a provider outcome that keeps history and the retry marker; ordinary repair also settles earlier failures superseded by a recorded deletion or exclusion. Public-only snapshots withhold channels with an unavailable marker until a successful crawl re-admits them, and routine syncs no longer panic when only marked channels remain.
- Apply Gateway GUILD_CREATE and THREAD_LIST_SYNC channel catalogs in one transaction, so large guilds no longer exceed the tail handler deadline at startup under FULL synchronization or leave a partially applied catalog.
- Reconcile full guild members inside the running tail owner with separately bounded pagination waits, without treating cached authors as complete enumeration.
- Keep native Components V2 text as attributed parts and fence recovery pages so a strictly newer edit or deletion is never replaced by a delayed page.
- Accept protected owner-only credential files for the bot token and embedding key in unattended services, with no symlinks, writable ancestors or granting ACLs, and no Keychain fallback once selected.
- Add opt-in measurement of native OpenAI embedding attempts and usage.
- Fix sync --full --since repeatedly resuming incomplete stored history and never syncing the guild's other channels; windowed full syncs now scan the normal channel catalog. Thanks @rszrszrsz.
- Refresh Crawlkit, provider SDKs, telemetry, and the paired SQLite/libc runtime while retaining the Go minimum; let Crabbox hydrate the Go toolchain without an unsupported explicit cache option.

0.16.0 - 2026-10-20
- Give private checkpoint upload, full authenticated SQLite verification, and promotion independent step budgets, preserving every custody check and the promotion reserve when verification is slow.
- Security: Build source, CI, releases, Docker, and optional language helpers with Go 1.27.3 to fix the reachable certificate-validation advisory GO-2026-6809, retaining existing Go language minimums.
- Refresh the Anthropic SDK, Google API and RPC clients, and Go telemetry dependencies while retaining the Go minimum and SQLite/libc compatibility pins.
- Refresh the OpenAI SDK, YAML parser, and Go telemetry dependencies while preserving the Go minimum and SQLite/libc compatibility pins.
- Add a manual, selected-stream historical backfill mode with an explicit UTC-midnight bound less than 31 days old, retaining the ordinary 48-hour window and shared acquisition deadline.
- Move native and Kiwi macOS CI to Xcode 27, preserve the macOS 13 deployment target, and verify Intel release binaries explicitly through Rosetta.
- Include bounded numeric rate-limit response headers in checkpoint API errors without exposing private output or retrying failed requests.
- Use faster gzip compression for private checkpoints, trading larger transfers for less packing CPU while preserving full authenticated database validation.
- Avoid redundant checkpoint packing reads by hashing the existing byte streams, and separate transfer and restore verification timings while retaining full authenticated disk and SQLite checks.
- Schedule private acquisition every two hours outside the daily member-refresh slot, preserving one serialized publisher and the existing phase budgets.
- Pin the reviewed private acquisition policy and keep member refresh within the existing shared acquisition deadline.
- Let targeted Discord syncs opt into stored direct threads of selected text and announcement roots, so recurring bounded acquisition covers newly discovered threads without changing ordinary channel selection.
- Bound checkpoint saves by one shared deadline, report safe phase timings, and remove a redundant integrity scan while preserving authenticated full-state verification.
- Recover an unpublished publisher checkpoint from retained verified state, preserving native provenance and publishing its private archive once through a guarded parent-bound update without collecting again.
- Finalize an interrupted initial checkpoint from its existing private assets after two complete verified restores, preserving original provenance and holding uncertain publication outcomes for inspection.
- Add manual, privately scoped report-history acquisition after the normal guild refresh, sharing one time budget and preserving failed progress in private checkpoints; hold recurring and cloud publication until separately qualified.
- Identify private checkpoint API failures without exposing private output, and inspect a retained failed migration through a bounded read-only diagnostic.
- Preserve complete native Discord crawl state in verified private checkpoints, with atomic restore, bounded retention, and an explicit migration from the retired workflow cache transport.
- Export current history-permission denials alongside retained window receipts, fail the sync if that status cannot be saved, and clear the unavailable marker after a successful full or targeted retry.
- Include voice and stage text chats in message sync, history backfills, and coverage reports; preserve stage channel types when resuming and require CONNECT when verifying their history access.
- Record permission-verified coverage for explicit bounded full-history scans, including empty windows, without completing older history or extending coverage from routine refreshes.

0.15.8 - 2026-10-16
Highlights: Recover cloud logins automatically after confirmed GitHub rate limits.
- Retry confirmed GitHub cloud-login rate limits after their cooldown, with bounded waits and distinct credential diagnostics.
- Update Crawlkit to v0.17.0 and refresh provider, network, telemetry, and Go analysis dependencies while retaining Go 1.27.0 and the matching SQLite/libc runtime; expose an explicit release-pipeline dry run for build verification without publishing.

0.15.7 - 2026-10-11
Highlights: Fix nine reachable Go standard-library vulnerabilities and the gRPC xDS denial-of-service advisory.
- Build with Go 1.27.2 to fix nine reachable standard-library vulnerabilities, including in Docker and optional language helpers, while retaining the existing minimum Go versions; refresh Go system, text, cryptography, terminal ANSI, and character-width dependencies, and run analyzers with the compatible module-pinned export reader.
- Update transitive gRPC to v1.83.2 to fix the xDS server denial-of-service advisory (GHSA-2v4p-qf9q-27wj).
- Refresh setup-node to v7.1.0 and TruffleHog to v3.99.0 across CI and existing publication workflows. Thanks @dependabot.
- Update CodeQL initialization and analysis together to v4.38.2 and group future updates to avoid incompatible action versions; refresh TruffleHog to v3.97.9. Thanks @dependabot.
- Use one shared autoreview installation from openclaw/agent-skills; repository entrypoints now receive upstream fixes without copied helpers or test suites.

0.15.6 - 2026-09-30
Highlights: Search Discord Components V2 cards and run exact semantic searches up to 3.7× faster.
- Keep Discord Components V2 messages, such as app-posted GitHub cards, searchable: bot sync keeps their component tree in raw_json and indexes its text, link buttons and media descriptions for search and snapshots, and Desktop imports index the same text. Messages archived before this fix keep empty text until they are re-read; the sync command docs describe backup-first recovery for bot-synced and Desktop-imported rows.
- Speed up exact semantic search 3.7× (10,000 × 1,536-dimensional embeddings: 506 ms → 137 ms, allocations 15.5M → 140k) by updating Crawlkit to v0.16.6 and skipping a redundant norm pass per candidate.
- Select the actual latest metrics observation at nanosecond precision, including imported timestamps with different offsets, without rewriting retained history.
- Stop cloud adoption when inventory pagination repeats a cursor, and release unread HTTP error responses before reporting publication failures.
- Distinguish cloud login and publisher authorization failures with bounded diagnostics that preserve private response details.
- Refresh SQLite to v1.60.1 with its required libc v1.77.1, go-strftime to v1.1.0, and golangci-lint to v2.14.0, retaining the Go 1.27.0 minimum.
- Update the TruffleHog secret-scanning action to v3.97.8. Thanks @dependabot.

0.15.5 - 2026-09-20
- Add explicit public Discord invite-count collection, scoped history import, and read-only status in a separate metrics database, preserving partial observations and existing archives. Thanks @hannesrudolph.
- Keep empty DM date-window hints consistent with the suggested follow-up command, excluding uncatalogued conversations only when removing the final date filter. Thanks @rnavarro.
- Refresh SQLite to v1.59.0 with its matching libc v1.75.7, go-humanize to v1.1.0, and terminfo to v1.2.0; retain the Go 1.27.0 minimum and current published Crawlkit v0.16.4.
- Update the optional Korean helper's native Kiwi library and model to 0.24.0, preserving the separate helper process and documented lexical-index rebuild procedure.
- Refresh CodeQL, GoReleaser, and pinned Go container images; align the Docker source smoke with Go 1.27.1.

0.15.4 - 2026-09-20
- Update Crawlkit to v0.16.4 so normal snapshot updates accept equivalent attachment schemas with reordered columns, preserving locally retained history without requiring forced replacement.

0.15.3 - 2026-09-19
- Hydrate Desktop-imported channels and their ancestors during targeted bot syncs so publication can use authoritative types, parents, and permissions without resetting stored history or weakening privacy filters.

0.15.2 - 2026-09-19
- Exclude private-thread crawl cursors from public-only snapshots while preserving public-thread cursors, full backups, and local sync state.
- Fix large message listings and TUI reply/mention hydration failing with SQLite's parameter limit, preserving all requested rows and display names.
- Add opt-in D1/R2 publication to the existing backup workflow, using one filtered export with resumable uploads and guarded archive adoption. Thanks @obviyus.
- Recover missing source members and messages through manual Discord reconciliation before adopting an existing hosted archive, preserving permission checks and existing source records. Thanks @obviyus.
- Update Crawlkit to v0.16.3, retaining the existing Go minimum and SQLite runtime pairing.

0.15.1 - 2026-09-13
- Explain empty search, messages, and dms results with stderr diagnostics for restrictive filters and missing local data, while preserving stdout, JSON output, and exit codes. Thanks @rnavarro.
- Avoid repeated requests to unavailable Discord channels for seven days during routine sync, while full and targeted syncs retry immediately, unavailable-only backlogs keep discovering channels, and doctor reports marker health using the same timestamp policy. Thanks @rnavarro.
- Preserve cancellation for interrupted SQLite work and closed transactions, while retaining the underlying diagnostic.
- Keep the generated documentation identity stable when building from a renamed checkout.
- Update CrawlKit to v0.16.2 and the optional Chinese tokenizer to gse v1.1.0 with cedar v0.50.0, retaining the Go minimum and SQLite runtime pairing.

0.15.0 - 2026-09-11
Highlights: Keep embeddings current during continuous capture, recover restart gaps on demand, and inspect live archive freshness.
- Add opt-in tail --embed-live to keep embeddings current during Discord capture, with coordinated queue ownership, provider batch and timeout settings, and background-worker health in status output. Thanks @hannesrudolph.
- Add opt-in tail --repair-on-start to recover restart gaps immediately after Gateway connection, using the existing serialized repair and writer ownership. Thanks @hannesrudolph.
- Exclude nested channels and threads beneath excluded categories during catalog discovery and live capture. Thanks @hannesrudolph.
- Include the last captured Gateway event time as optional last_tail_event_at in local status --json, separately from the last completed sync. Thanks @hannesrudolph.
- Speed per-channel changed-message queries with an additive update-cursor index; document synchronous first-writable-open costs and backup-first evaluation before broader rollout. Thanks @hannesrudolph and @vincentkoc.
- Allow explicit OS keyring credentials for embeddings, with safe initialization retries during capture and unchanged environment-based defaults. Thanks @hannesrudolph.
- Generate deterministic, aggregate-only Discord field notes alongside daily published activity reports, with paired Markdown/JSON artifacts, separate markers preserving legacy narrative notes, explicit timestamp and coverage limits, and filtered-publication cleanup. Thanks @vincentkoc.
- Use CrawlKit v0.16.1 to bound live-embedding shutdown when cancellation interrupts retry or completion persistence and cleanup storage is unavailable. Thanks @hannesrudolph.
- Refresh terminal-width, Go system, text and cryptography dependencies, Go analysis tools, and CodeQL while retaining the existing Go minimum and SQLite runtime pairing.

0.14.1 - 2026-09-09
- Accept the application's legacy migrated attachment column order in the backup-first repair helper while retaining exact schema and repair safeguards.
- Add a manually gated, backup-first maintenance workflow for narrowly scoped attachment text repair, with strict snapshot validation before installing a new Actions database cache.
- Keep extracted attachment text within its byte limit without splitting valid UTF-8 characters, and identify invalid admitted attachment columns in contents-free export errors without changing stored data or weakening snapshot validation.
- Limit disposable Discord backup clones to current main and required descendants, without fetching tagged history.
- Provide checked-out source and workflow run identity to Discord archive publications.
- Update CrawlKit to v0.15.0 while retaining the existing Go minimum and preferred toolchain.
- Preserve stable published shard paths for released subscribers, roll back handled pre-manifest installation failures, and report post-manifest cleanup failures without committing or pushing a partial result.
- Add optional manifest-bound producer.json receipts for explicitly identified publishers, preserving unrelated files and checking the binding before publication push retries. Receipts are unsigned publisher assertions, not authenticated provenance.
- Keep publication Git arguments bounded with literal NUL-delimited paths and one path-only commit, preserving unrelated staged work.
- Report incomplete selected channel or category permissions in publication preflight, including inherited thread permissions, without changing public-only export selection.
- Accept exact manifest-owned table generation paths alongside legacy snapshot shards without broadening publication cleanup or Git staging.
- Remove previously published embedding and media files when omitted, and commit only exact archive outputs and explicitly generated reports while preserving unrelated staged work.
- Exclude deleted messages from cloud row publication and SQLite bundles, using the existing replacement protocol and matching live-row counts.
- Preserve cached channel permissions during targeted sync and exclude unobserved permissions from public-only snapshots.
- Exclude nested unselected guild metadata from filtered snapshots while retaining local raw observations.
- Keep local Wiretap DMs out of embedding enqueue, rebuild and provider requests, including jobs queued by older versions.
- Keep inferred Git fingerprints separate from snapshot integrity metadata when importing legacy snapshots, including historical restores, while preserving incremental merge checkpoints.
- Exclude local-only direct messages from published README activity reports and add report --published for standalone shared reports, while preserving full local reports and filtered-publication safeguards.
- Remove prior manifest-declared embedding files when export is disabled or its model changes, preserving undeclared embedding/media siblings, unrelated staged work, and local vectors.
- Validate imported embedding rows against their manifest identity and vector dimensions; reject malformed or non-finite vectors transactionally and skip missing, deleted, and local-only message targets while preserving existing DM vectors.

0.14.0 - 2026-09-07
Highlights: Search Korean, Japanese, Chinese, and Arabic word forms with optional local language analyzers.
- Add opt-in multilingual lexical indexes with independent relevance ranking, safe analyzer-change rebuilds, and discrawl lexical rebuild; keep the default binary free of helper dictionaries and preserve existing search behavior when languages are disabled. Thanks @vkehfdl1.
- Reject unsafe snapshot shard paths with CrawlKit v0.14.9 while preserving the SQLite-required libc runtime.

0.13.4 - 2026-09-05
Highlights: Recover missing channel history and cached messages, and keep routine sync discovering new channels and archived threads.
- Verify supposedly complete channels with no stored messages, recover their full history, and preserve retryability across bounded cancellation cleanup or process interruption. Thanks @rnavarro.
- Recover unresolved Desktop cache entries across upgrades and input modes without overwriting newer edits, duplicating events, or rescanning completed full-cache files. Thanks @Nicolas0315.
- Discover newly archived threads during incremental sync with per-endpoint cursors and resume incomplete thread histories without rescanning the full archive. Thanks @obviyus.
- Bootstrap newly discovered channels with one newest message page during routine latest-only sync, preserving a backfill cursor for later full-history recovery. Thanks @obviyus.
- Keep Wiretap dry runs from creating archives or runtime directories; optional coverage reads an existing archive without migrations.
- Stop guild and archived-thread pagination with a cursor error when Discord repeats a page instead of hanging sync. Thanks @SebTardif.
- Reject repeated or missing message-page cursors without losing the last usable backfill checkpoint. Thanks @SebTardif.
- Bound release-check HTTP requests to 30 seconds with Crawlkit v0.14.8 so an unresponsive server cannot hang update checks indefinitely.
- Require Go 1.27.0 and prefer Go 1.27.1 for builds; refresh SQLite to v1.58.0, crypto and terminal dependencies, and container, analyzer, security-scan, and docs-build tooling.

v0.13.3 - 2026-08-17

Fixes
- Add persistent exact snapshot updates for dedicated Git-only readers so rows omitted by a newer filtered snapshot are removed without changing the merge-by-default behavior of richer local archives.
- Skip Discord members with a missing user object during pagination and record conversion so public sync no longer panics on a nil User. Thanks @SebTardif.

v0.13.2 - 2026-08-14

Maintenance
- Require Go 1.26.6 and refresh the pinned Go container images for the latest standard-library security fixes.

0.13.1 - 2026-08-09
Highlight: tail --verbose now tells you why a message was not archived.

Fixes
- Trace live Discord message receipt, handling, scope filtering, and archive completion in verbose tail logs. Each event reports gateway_received, handler_started, ignored with a reason, or archived, so an out-of-scope message is no longer indistinguishable from one the Gateway never delivered. archived is emitted only after the canonical write, event append and cursor update all succeed. Traces carry event type, stage and IDs only — never message content or author metadata.

Maintenance
- Update Crawlkit to v0.14.6, Kong to v1.16.1, the modernc SQLite stack to v1.56.0, and refresh Go analysis tooling.

0.13.0 - 2026-08-03

Changes
- Surface stored message-to-channel catalog integrity in diagnostics, annotate output whose channel metadata is missing, and warn when zero-row SQL results may hide orphaned messages. Thanks @htydev.

Maintenance
- Update Crawlkit to v0.14.5 and go-colorful to v1.4.1.

Documentation
- Rewrite the README as a concise front door to installation, archive sources, search, and the existing documentation site.

0.12.0 - 2026-08-02

Changes
- Queue live tail, failure replay, and repair messages for embedding behind explicit tail --with-embeddings opt-in, and support OpenAI projection dimensions (#159, based on #154 — thanks @hannesrudolph).

Maintenance
- Update Crawlkit to v0.14.4, Alpine to 3.24, and refresh the SQLite, libc, terminal width, and terminal detection dependency stack.
- Update CodeQL Action to v4.37.4 and actions/stale to v11 while preserving the existing security-scan and stale-item policies.
- Let the unified release workflow own version tags and organization-managed signing, notarization, publication, and Homebrew delivery.

0.11.10 - 2026-07-27

Maintenance
- Move official releases to the shared signed, notarized, independently verified GitHub Actions pipeline while preserving archive names, contents, checksums, and Homebrew delivery.

0.11.8 - 2026-07-20

Highlights
- Preserve member and guild history across syncs and Git-share merges with source-attributed tombstones, explicit restore semantics, and revision-aware reconciliation.
- Keep large archives responsive and attachment retrieval secure with bounded coverage aggregation and strict Discord CDN redirect validation.

Fixes
- Prevent global coverage reports from timing out on large archives by aggregating messages in one sequential scan and allowing a longer bounded deadline.
- Preserve guild and member history with source-attributed tombstones, explicit restore semantics, omission-safe member refreshes, and revision-aware Git-share merges.
- Keep attachment fetches compatible with three allowed Discord CDN redirects and injected HTTP transports while preserving final-response host validation.
- Reject malformed tombstone timestamps before routine incremental snapshot imports mutate the archive.

Maintenance
- Update golang.org/x/crypto to v0.54.0 and modernc.org/libc to v1.74.3, while confirming Crawlkit v0.14.3 remains current.

Highlights
- Ship official macOS binaries notarized by Apple, with hardened runtime, trusted timestamps, stable Foundation identity requirements, and online ticket verification.
- Keep incomplete macOS artifacts out of releases by preserving GoReleaser outputs until signing, notarization, and every post-notarization check succeeds.

Maintenance
- Require the private runtime notary profile for official releases and verify notarization tickets again from published macOS archives.
- Update Crawlkit to v0.14.3 and its compatible SQLite and terminal-detection stack.

0.11.6 - 2026-07-17

Highlights
- Keep live Discord tails recoverable across handler timeouts and process restarts with a private durable spool and exact create, update, and delete replay. Thanks @hannesrudolph.
- Make CLI discovery and setup smoother with descriptive root and nested help, side-effect-free -h, and actionable missing-configuration guidance. Thanks @0xdevalias.
- Harden archive boundaries by blocking attachment redirects off Discord CDN hosts and rejecting malformed tombstone timestamps before import mutations. Thanks @GrantTheAssistant.
- Report accurate versions for source-installed binaries.

Changes
- Add bounded tail failure replay through discrawl tail --replay-failures-only and --replay-limit, with privacy-safe diagnostics. Thanks @hannesrudolph.
- Improve root, command, and nested command help with descriptions and side-effect-free -h handling. Thanks @0xdevalias.

Fixes
- Keep Discord Gateway tails fail-closed while durably spooling and replaying timed-out message creates, updates, and deletes. Thanks @hannesrudolph.
- Reject attachment redirects whose final URL leaves Discord's allowlisted CDN hosts. Thanks @GrantTheAssistant.
- Reject malformed message tombstone timestamps before either SQLite snapshot import path mutates the archive. Thanks @GrantTheAssistant.
- Show setup guidance instead of a raw missing-file error when configuration-dependent commands cannot find config.toml. Thanks @0xdevalias.
- Report module build metadata for source-installed binaries instead of a stale hard-coded release version.

Maintenance
- Update Crawlkit to v0.14.2 and Kong to v1.16.0, alongside current x/sys, x/text, and TruffleHog releases.
- Harden GoReleaser macOS archive verification before signed publication.
- Sync canonical AutoReview tooling and strengthen repository secret scanning. Thanks @vincentkoc.

0.11.5 - 2026-07-09
- Add Developer ID signing and release verification for official macOS binaries under OpenClaw Foundation Team ID FWJYW4S8P8, while keeping local and cross-platform snapshot builds credential-free.
- Update CrawlKit to v0.13.4 and Go to 1.26.5, including the standard-library crypto/tls security fixes.
Full changelog

Discrawl v0.11.0

Changes
- Add optional turbovec semantic-search scoring via [search.embeddings].vector_backend, while keeping exact cosine as the default backend. Thanks @vincentkoc.
- Added the Homebrew install command to the discrawl.sh landing hero and agent docs index, with a one-row desktop layout and copy button.
- Update crawlkit through v0.12.0.
- Add read-only Cloudflare remote archive scaffolding with [remote] config,
subscribe-cloud, GitHub-backed remote login with OAuth or token-env
bootstrap, remote status, remote archives, and cloud-mode status --json
output that does not open or create a local SQLite database.
- Route cloud-mode search and filtered messages reads to Worker named
queries so subscribers can inspect live D1 data without local SQLite.
- Add discrawl cloud publish to export non-DM local SQLite rows into the
Cloudflare remote archive ingest API without changing Git snapshot
publishing.
- Mirror the non-DM local SQLite archive into the Worker-backed R2 object store
during discrawl cloud publish, alongside the D1 row ingest used for live
queries.
- Compress the sanitized SQLite mirror as a gzip chunk bundle with an explicit
privacy/count manifest before uploading to R2.

Fixes
- Kept resumed sync --full backfills from moving channel latest-message checkpoints backward, avoiding duplicate head recrawls on large interrupted channels. Thanks @hannesrudolph.
- Made messages --sync fail fast with an omit---sync hint when a live tail process owns the sync lock, while plain messages reads continue without waiting. Thanks @jeanmonet.

0.9.1 - 2026-05-18

Changes
- Add cached release checks with discrawl check-update and passive terminal notices when a newer Discrawl release is available.

0.9.0 - 2026-05-17

Changes
- Media-enabled discrawl publish now migrates shared attachment media to gzip-compressed files while still importing older raw-media snapshots.
- Semantic search now scores lightweight embedding rows first and hydrates full message details only for the winning results.

Fixes
- Bounded gzip media restore and hash verification so malformed shared snapshots cannot decompress unbounded data.
- Cancelled concurrent message-sync workers when a peer hits a fatal channel error.
- Rejected inaccessible explicit guild targets during init --guild and sync --guild/--guilds instead of silently treating them as successful empty syncs.
- Hardened embedding snapshot imports against unsafe manifest paths, symlink escapes, and unbounded gzip input.
- Limited FTS search fallback to missing or unsupported FTS infrastructure errors so unrelated query failures are reported.
- Re-sniffed previously skipped wiretap cache files instead of treating unchanged skipped fingerprints as permanently unchanged.

Changes
- Added attachment media caching with discrawl attachments, attachments fetch, sync --with-media, and Git snapshot backup/restore for cached non-DM media files.
- Documented media backup flow, including CDN fetch failures, local cache behavior, and Git snapshot publishing.
- Docker: add a local image with /data persistence and CI smoke coverage.
- Moved stable store SQL for sync state, messages, attachments, embedding jobs, members, and status reads/writes to sqlc-generated typed wrappers while leaving dynamic FTS, semantic search, report, share, and user SQL handwritten.

Fixes
- Kept large Git snapshot imports and FTS rebuilds from exhausting memory on small hosts by using file-backed SQLite temp storage and a bounded import cache. (#65) Thanks @hxy91819.

Fixes
- Kept Git snapshot imports incremental when metadata, attachment, mention, or event tables changed alongside the message tail, avoiding full archive replays for routine updates.

Fixes
- Repeated sync --source wiretap runs now skip unchanged Discord Desktop cache files and report unchanged file counts, making steady-state local-cache refreshes much faster.
- sync --full --skip-members now also skips member crawls when resuming incomplete stored channels, so backfills do not unexpectedly refresh the full guild member list.

Maintenance
- Refactored sync-mode handling so routine latest syncs, --all-channels, --full, and member-refresh decisions share clearer internal paths with regression coverage.
- Refreshed Go module dependencies and CI tool/action pins, including staticcheck, gofumpt, gosec, govulncheck, gitleaks, setup-node, and GoReleaser.
- Hardened report README writes and Discord Desktop cache reads with root-scoped filesystem access to satisfy the latest gosec checks.

Changes
- dms now lists local wiretap DM conversations and can read or search one DM thread with --with, --last, and --search, so common DM queries no longer require raw SQL.
- search --dm and messages --dm now target the local-only @me archive directly and skip Git snapshot auto-update, since DMs are never imported from the shared mirror.
- Go module dependencies and lint rules were refreshed for the current Go toolchain, including stricter JSON marshal checks and modern simplification rules.

Fixes
- Wiretap now infers fallback DM channel names from cached Discord user/profile data, so channels discovered only from route/message cache entries resolve to names like Vincent K instead of channel-*.
- Wiretap message output now preserves sanitized author labels in stored metadata, improving dms and messages output without storing raw desktop cache payloads.

Tests
- Added regression coverage for DM channel-name inference from cached profile data when Discord Desktop cache lacks explicit channel recipient metadata.
- Added coverage for local DM conversation listing/filtering, DM cleanup paths, share import/export helpers, CLI DM windows, and Discord Desktop import helper edge cases.
- CI now runs uncached test and race suites, checks go mod tidy, and performs a snapshot GoReleaser build before release tags.

- 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

- 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
- OpenClaw config reuse plus 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 80%+ 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