Tracks playback position via frame counting in stream_audio().
On stop/skip, saves URL + elapsed time to bot.state (SQLite).
!resume reloads the track and seeks to the saved position via
ffmpeg -ss. State persists across bot restarts.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Non-URL input (e.g. !play classical music) searches YouTube for 10
results and picks one randomly. Also fixes --flat-playlist returning
"NA" as the URL for single videos by falling back to the original
input URL.
_resolve_title replaced with _resolve_tracks using --flat-playlist to
enumerate playlist entries. cmd_play enqueues each track individually,
with truncation when the queue is nearly full. Single-video behavior
unchanged.
stream_audio now accepts a callable for volume, re-read on each PCM
frame instead of capturing a static float at track start.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
asyncio's SSL memory-BIO transport silently drops voice packets even
though text works fine. pymumble uses blocking ssl.SSLSocket.send()
which reliably delivers voice data.
- Rewrite MumbleBot to use pymumble for connection, SSL, ping, and
voice encoding/sending
- Bridge pymumble thread callbacks to asyncio via
run_coroutine_threadsafe for text dispatch
- Voice via sound_output.add_sound(pcm) -- pymumble handles Opus
encoding, packetization, and timing
- Remove custom protobuf codec, voice varint, and opus ctypes wrapper
- Add container patches for pymumble ssl.wrap_socket (Python 3.13) and
opuslib find_library (musl/Alpine)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Connect to multiple IRC servers concurrently from a single config file.
Plugins are loaded once and shared; per-server state is isolated via
separate SQLite databases and per-bot runtime state (bot._pstate).
- Add build_server_configs() for [servers.*] config layout
- Bot.__init__ gains name parameter, _pstate dict for plugin isolation
- cli.py runs multiple bots via asyncio.gather
- 9 stateful plugins migrated from module-level dicts to _ps(bot) pattern
- Backward compatible: legacy [server] config works unchanged
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
HTTP POST endpoint for external services (CI, monitoring, GitHub).
HMAC-SHA256 auth, JSON body, single POST endpoint at /.
- asyncio.start_server with raw HTTP parsing (zero deps)
- Body validation: channel prefix, non-empty text, 64KB cap
- !webhook admin command shows address, request count, uptime
- Module-level server guard prevents duplicates on reconnect
- 22 test cases in test_webhook.py
4-tier permission model: user < trusted < oper < admin.
Commands specify a required tier via tier= parameter.
Backward compatible: admin=True maps to tier="admin".
- TIERS constant and Handler.tier field in plugin.py
- _get_tier() method in bot.py with pattern matching
- _is_admin() preserved as thin wrapper
- operators/trusted config lists in config.py
- whoami shows tier, admins shows all configured tiers
- 32 test cases in test_acl.py
Admin-only plugin for interval-based command execution.
Supports add/del/list, 1m-7d intervals, 20 jobs/channel.
Persists via bot.state, restores on reconnect.
Includes test_cron.py (~38 cases).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Bot.shorten_url() method delegates to flaskpaste plugin when loaded.
RSS, YouTube, and pastemoni announcements auto-shorten links.
Includes test_flaskpaste.py (9 cases) and FakeBot updates in 3 test files.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add cmd_paste to flaskpaste plugin (create paste, return URL).
Add test suites for encode, hash, defang, cidr, and dns plugins
(83 new test cases, 1093 total).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
canary: generate realistic fake credentials (token/aws/basic) for
planting as canary tripwires. Per-channel state persistence.
tcping: TCP connect latency probe through SOCKS5 proxy with
min/avg/max reporting. Proxy-compatible alternative to traceroute.
archive: save URLs to Wayback Machine via Save Page Now API,
routed through SOCKS5 proxy.
resolve: bulk DNS resolution (up to 10 hosts) via TCP DNS through
SOCKS5 proxy with concurrent asyncio.gather.
83 new tests (1010 total), docs updated.
YouTube's InnerTube /player endpoint now returns LOGIN_REQUIRED for the
WEB client. Switch to ANDROID client context which still returns full
videoDetails. Fixes missing duration in announcements and broken channel
resolution from video URLs.
Extract shared _innertube_player() helper to deduplicate payload
construction between _resolve_via_innertube and _fetch_duration.
Free, keyless API returning open ports, hostnames, CPEs, tags, and
known CVEs for any public IP. All requests routed through SOCKS5.
21 test cases (927 total).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Drop GeoLite2-ASN.mmdb dependency (required license key) in favor of
iptoasn.com ip2asn-v4.tsv (no auth, public domain). Bisect-based
lookup in pure stdlib, downloaded via SOCKS5 in update-data.sh.
Adds 30 test cases covering load, lookup, and command handler.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Fetches duration via InnerTube player API for new videos at
announcement time. Displayed as compact h:mm:ss before views/likes.
Gracefully omitted for Shorts and unavailable content.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Alert backends now populate structured `extra` field with engagement
metrics (views, stars, votes, etc.) instead of embedding them in titles.
Subscription plugins show richer announcements: Twitch viewer counts,
YouTube views/likes/dates, RSS published dates.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Poll Pastebin archive and GitHub Gists for keyword matches,
announce hits to subscribed IRC channels. Follows rss.py
polling/subscription pattern with state persistence.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add Bot.long_reply() that sends lines directly when under threshold,
or creates a FlaskPaste paste with preview + link when over. Refactor
abuseipdb, alert history, crtsh, dork, exploitdb, and subdomain
plugins to use long_reply(). Configurable paste_threshold (default: 4).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Event-driven plugin that auto-fetches page titles for URLs posted in
channel messages. HEAD-then-GET via SOCKS5 pool, og:title priority,
cooldown dedup, !-suppression, binary/host filtering. 52 tests.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace per-request SOCKS5+TLS handshakes with urllib3 SOCKSProxyManager
connection pool (20 pools, 4 conns/host). Batch _fetch_og calls via
ThreadPoolExecutor to parallelize OG tag enrichment in alert polling.
Cache flaskpaste SSL context at module level.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Mastodon, Bluesky, GitHub, GitLab, npm, PyPI, and arXiv backends
no longer truncate content/descriptions in titles. Full text is
shown on the PRIVMSG line; only !alert history keeps truncation.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
ACTION carries the tag/date/URL, PRIVMSG carries the uncropped title.
Removes _truncate on alert output for better readability.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add _fetch_many() helper using ThreadPoolExecutor to query instances
in parallel. Refactors PeerTube, Mastodon, Lemmy, and SearXNG from
sequential to concurrent fetches. Also adds retries parameter to
derp.http.urlopen; multi-instance backends use retries=1 since
instance redundancy already provides resilience.
Worst-case wall time per backend drops from N*timeout to 1*timeout.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Per-backend error counts with exponential backoff: after 5 consecutive
failures a backend is skipped every 2^(n-5) cycles (capped at 32).
Working backends are no longer penalized by one flaky backend doubling
the entire poll interval.
Migrates last_error (string) to last_errors (dict per backend).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Reply immediately with empty seen, run a silent _poll_once in a
background task to populate seen IDs, then start the poller.
Eliminates the 30-120s blocking wait from 27 sequential backend queries.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Store shortened URLs in the results DB at poll time alongside the
original URL. History output uses the stored short URL directly,
only regenerating (and persisting) when no short URL exists yet.
Original URL always preserved for re-shortening if needed.
When secrets/flaskpaste/derp.crt and derp.key are present, load them
into the SSL context for mutual TLS auth and skip the PoW challenge
entirely. Fall back to PoW only when no client cert is available.
- PoW-authenticated paste creation and URL shortening via FlaskPaste
- !paste <text> creates a paste, !shorten <url> shortens a URL
- Module-level shorten_url/create_paste helpers for cross-plugin use
- Alert plugin auto-shortens URLs in announcements and history output
- Custom TLS CA cert support via secrets/flaskpaste/derp.crt
- No SOCKS proxy -- direct urllib.request to FlaskPaste instance
Tor exit nodes poison plain DNS on port 53 and MITM some HTTPS
connections. Replace raw TCP DNS with DoH (Google, Cloudflare, Quad9)
and retry up to 5 times across providers to find a clean exit node.
MX results are now sorted by priority.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The 512-byte IRC limit includes the :nick!user@host prefix the server
prepends when relaying. Reserve 64 bytes for it and prefer splitting at
space boundaries instead of mid-word. Also strip the command prefix and
"Commands:" label from help output to keep the listing compact.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
v2.0.0 sprint 1 -- five standalone plugins requiring no core changes:
- jwt: decode JWT header/payload, flag alg=none/expired/nbf issues
- mac: IEEE OUI vendor lookup, random MAC generation, OUI download
- abuseipdb: IP reputation check + abuse reporting (admin) via API
- virustotal: hash/IP/domain/URL lookup via VT APIv3, 4/min rate limit
- emailcheck: SMTP RCPT TO verification via MX + SOCKS proxy (admin)
Also adds update_oui() to update-data.sh and documents all five
plugins in USAGE.md and CHEATSHEET.md.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add Wikipedia, Stack Exchange, GitLab, npm, PyPI, Docker Hub,
arXiv, Lobsters, DEV.to, Medium, and Hugging Face backends to
the alert plugin (16 -> 27 total). Fix PyPI backend to use RSS
updates feed (web search now requires JS challenge). Fix DEV.to
to use public articles API (feed_content endpoint returns empty).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Video URLs (watch, shorts, embed, youtu.be) now resolve the channel
ID through the InnerTube player API -- a small JSON POST instead of
fetching the full 1MB watch page. Much more resilient to transient
proxy failures. Page scraping remains as fallback for handle URLs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Each alert result gets a deterministic 8-char base36 ID derived from
backend:item_id. IDs appear in announcements and history, and can be
looked up with !alert info <id> for full details. Existing rows are
backfilled on startup.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Hacker News (hn) uses Algolia search_by_date API for stories,
appends point count to title, falls back to HN discussion URL
when no external link. GitHub (gh) searches repositories sorted
by recently updated, shows star count and truncated description.
Both routed through SOCKS5 proxy via _urlopen.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Bluesky (bs) searches public post API, constructs bsky.app URLs
from at:// URIs. Lemmy (ly) queries 4 instances (lemmy.ml,
lemmy.world, programming.dev, infosec.pub) with cross-instance
dedup. Odysee (od) uses LBRY JSON-RPC claim_search for video,
audio, and documents with lbry:// to odysee.com URL conversion.
Archive.org (ia) searches via advanced search API sorted by date.
All routed through SOCKS5 proxy via _urlopen.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Kick (kk) searches channels and livestreams via public search API.
Dailymotion (dm) queries video API sorted by recent. PeerTube (pt)
searches across 4 federated instances with per-instance timeout.
All routed through SOCKS5 proxy via _urlopen.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
DuckDuckGo (dg) searches via HTML lite endpoint with HTMLParser,
resolves DDG redirect URLs to actual targets. Google News (gn)
queries public RSS feed, parses RFC 822 dates. Both routed through
SOCKS5 proxy via _urlopen.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Query general, news, videos, and social media categories
separately with time_range=day. Dedup results by URL across
categories to avoid announcing the same item twice.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Search Reddit posts (rd) via JSON API and Mastodon hashtag
timelines (ft) across 4 fediverse instances. Both public,
no auth required.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When the bot accepts an admin INVITE, the channel is stored in
bot.state under chanmgmt/autojoin:<channel>. On reconnect, persisted
channels are rejoined alongside configured ones. If the bot is kicked,
the channel is removed from the auto-rejoin list.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Matched results were announced then discarded. Add a dedicated SQLite
database (data/alert_history.db) to store every announced result with
channel, alert name, backend, title, URL, date, and timestamp. Add
!alert history <name> [n] subcommand to query recent results.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
YouTube InnerTube is a public API -- no need to route through SOCKS5
proxy, which was causing SSL EOF errors on every poll. Switch to
direct urllib.request.urlopen.
Remove _MAX_ANNOUNCE cap; all matched results are now announced
individually instead of truncating with "... and N more".
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
SearXNG can return "publishedDate": null, which bypasses the default
value in dict.get() and passes None to _parse_date / re.search.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Extract dates from multiple sources:
- SearXNG: publishedDate field from search results
- YouTube: publishedTimeText from InnerTube response
- OG fallback: article:published_time, og:updated_time, date,
dc.date, dcterms.date, sailthru.date meta tags
Date is shown as (YYYY-MM-DD) or relative time after the tag prefix.
OG tags are fetched for date even when title/URL already matched.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When a search result's title/URL doesn't contain the keyword, fetch
the page's first 64 KB and parse og:title and og:description meta
tags. If the keyword appears there, the result is announced. Prefers
og:title as display title when it's richer than the search result
title.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
