StreamLinkr Docs

Choose the active provider: Ollama, Gemini, OpenAI, or Grok. Only one provider powers replies at a time. Use Compare providers to evaluate local privacy, hosted convenience, cost, speed, and model quality.

Enable "Ollama is running locally" - Host and Port fill in automatically (localhost:11434). Shows a Program path field and Auto-launch Ollama toggle. If Ollama is on another machine, disable that toggle and enter the LAN IP and port manually. "Allow relaxed fallback" lets the model accept slightly lower quality if the strict quality check would have dropped the reply.

Enable "Use Ollama Cloud" and enter a Cloud API key. Model defaults to gemma3:27b.

Enter the API key, click Refresh models, choose a model. API timeout and Bot reply timeout are set per provider.

Temperature (0.1–2), Top P (0.1–1), Top K (1–100), Repeat penalty (1–1.5), Max context tokens, Max output tokens.

Temperature (0–2), Top P (0–1), Max tokens (100–500), Frequency penalty (−2 to 2), Presence penalty (−2 to 2).

Temperature (0–2), Top P (0–1), Top K (1–100), Max output tokens (number field).

Temperature (0–2), Top P (0–1), Max tokens (100–500).

Manage persona presets, identity, personality, guidelines, tone guide, stay-in-character notes, and speech style.

Control forbidden opening and closing patterns so the bot never starts or ends replies with unwanted phrasing.

Positive and negative sentiment vocabulary words (comma-separated) used to gauge chat tone. Reply timeout/delay now live in Connections → Timeouts, and offended terms/responses moved to Responses → Events.

Add paired viewer inputs and preferred bot responses to teach style. AI failure fallback messages are configured here too - one is chosen at random when the provider fails to return a reply.

Apply built-ins, clone them, save custom presets, update custom presets, delete them, export to JSON, or load a preset file. Cloning copies the current form values and lets you rename before saving.

Base chance (%), minimum chance, maximum chance, activity threshold (messages per window), familiarity bonus (extra % for known viewers). Higher activity raises the effective chance toward the maximum.

Conversation buffer limit (messages kept in active context), reply cache size (anti-repetition history depth), buffer TTL hours (how long the buffer stays active before expiring between streams).

Enable proactive chat, idle minutes (quiet time before the bot speaks unprompted), send startup message, send shutdown message, idle message examples, startup prompt examples, shutdown prompt examples.

Auto greeting toggle, Use AI greetings toggle, greeting message examples (supports {user}).

Enable follow messages, Use AI responses toggle, Require stream online toggle, example messages.

Enable shoutouts, auto on raid, auto on subscription, show tags, message length limit. Per-platform fetch timeouts and gift sub shoutout cooldown are set in Connections → Timeouts.

Enable offended responses toggle, offended terms list, and offended response examples - lets the bot react in character when chat turns hostile.

Post a "stream is live" message to a webhook (Discord-compatible) when the stream goes online. Mention @everyone or a custom role ID (mutually exclusive), per-platform opt-out (Twitch/Kick/YouTube), and AI-generated or static example messages with {NAME}/{URL} placeholders.

Global cooldown seconds between announcements. Per announcement: name, examples, interval minutes, minimum chat lines before eligible, priority (1–10), command trigger, Use AI toggle, Command-only mode (hides from auto-rotation), Moderator-only command toggle.

help, shoutout (so), comment, tts, roast, compliment, lurk, unlurk, stfu, purge - depending on which features are enabled.

points, leaderboard, givepoints, gamble, duel, accept, decline, winrate, raffle, join, mytickets, givetickets, addpoints, removepoints - managed from Engagement and Permissions.

sr, queue, wrongsong, skip, clearqueue. Moderators and the broadcaster bypass most per-user limits.

clip - creates a Twitch clip via Streamer.bot. Twitch-only, configured and rate-limited from Clips.

If a command works on Twitch but not Kick or YouTube, check the platform-specific role access before changing the command itself.

Set in Community → Moderation → Redeems. Applies to all chat commands globally.

Master engagement on/off switch plus per-platform toggles (Twitch, Kick, YouTube). Engagement commands and point accrual are inactive until the global toggle is on.

Points per message, lurk point interval minutes (minimum 5), lurk points per interval. Inspect balances per platform, filter by username, edit individual balances, reset individual viewers, or reset all balances for a platform.

Enable gamble (!gamble <amount>, !winrate), win chance (%), payout multiplier (stake × multiplier on win). Enable AI-generated gamble messages with separate win and loss example sets.

Enable duels (!duel <user> <amount>, !accept, !decline, !winrate), accept window 10–60 seconds. Enable AI-generated duel messages with result, timeout, and decline example sets.

Enable raffle (!raffle <amount>, !join), raffle duration 10–60 seconds (default 30). Enable AI-generated raffle winner message and provide winner examples.

A recurring, watch-time-driven prize draw, separate from Raffle - viewers earn tickets over time instead of entering a single round. Enable lotto, exclude streamer/exclude moderators toggles, optional overlay wheel with winner reveal delay (0–20s). Viewer and subscriber ticket weights (1–10x). Watch-time accrual: interval in minutes and tickets per interval (needs Streamer.bot's Present Viewers → Live Update → Live). Channel point redeem support with tickets-per-redeem. Manual controls: start, pause/resume accrual, pick winner, re-roll, clear all tickets. Viewers check tickets with !mytickets and can gift them with !givetickets.

Enable giveaways. Viewer ticket weight, subscriber ticket weight, exclude moderators toggle. Per-platform giveaway support. Live entry panel. Using overlay toggle holds the chat announcement until the overlay wheel finishes spinning, with a configurable winner reveal delay (default 9.5s).

Recent point transactions showing user, delta, and reason - audit trail for balance changes, edits, transfers, and resets.

Name, command trigger, initial value, increment amount, increment permission, enabled state, persistence toggle, show in overlay toggle, and optional AI example messages.

Edit existing counters, trigger an increment from the UI, set an exact value, reset, delete, or save changes per counter.

Each counter has its own increment permission. Admin set/reset commands are controlled from Moderation → Permissions.

AI counter messages support {USER}, {COUNTER}, {COUNT}, and {DELTA}.

Enable ad-run messages, use AI toggle, and example messages (supports {ad_length}). Ad-run messages fire automatically when a Twitch ad starts.

Usernames in this list are silently ignored - no AI response, no command processing. Add bots like nightbot, streamelements, etc.

Ban, Timeout, Untimeout, and Unban each have a dedicated page with a global enable toggle, per-platform toggles, and example messages.

Role-based access per command per platform. Commands: comment, tts, roast, compliment, lurk, unlurk, so, purge, points, leaderboard, givepoints, gamble, duel, raffle, sr, skip.

Command prefix (applies globally), global command cooldown seconds.

Map Twitch action ID and Kick action ID to: comment, tts, roast, compliment, stfu. Each redeemable command has a "Disable chat command" toggle.

Enable song requests, tune cooldowns, per-user limits, queue size, and command behavior. Now-playing overlay toggle runs independently of the !sr command - it can stay on to show whatever is playing even if song requests themselves are disabled.

Configure YouTube with pear-desktop or Spotify with developer credentials, connection state, and playback requirements. YouTube search does not need an API key.

Choose the confirmation style: default, roast, compliment, or AI-generated response examples.

Inspect the live queue, skip entries, clear the queue, and understand current pending requests.

Requires Premium, an active playback device, a Developer app, and the redirect URI http://127.0.0.1:51542/api/spotify/callback.

Requests are stored locally in DB_Files/song_queue.db. Clear the queue between streams if you do not want skipped entries to accumulate.

Host, Port, Endpoint, Password for the WebSocket connection. Program path to auto-launch the executable. Auto-launch Streamer.bot toggle. Auto-stop delay slider (0–120 minutes) for after the stream goes offline.

Auto-start chatbot toggle (starts the bot when StreamLinkr launches), start when stream goes online toggle, require stream online toggle (chat behavior only runs while a stream is live).

Reply timeout (s) and reply delay (s) for AI responses, plus AI API timeout (s) for non-Ollama providers. Per-platform shoutout fetch timeouts (Twitch / Kick / YouTube) and gift sub shoutout cooldown.

Expand "Streamer.bot action group import" in the Streamer.bot section and click Copy import data, then paste into Streamer.bot → Import.

Enable TTS globally master toggle. Without this on, all TTS is inactive regardless of other settings.

Enable random bot-reply TTS. Trigger chance (%), random cooldown seconds, minimum text length, maximum text length. Action name and Voice alias must match Streamer.bot and Speaker.bot configuration.

Ban message TTS toggle + example messages. Subscription TTS toggle + example messages.

Enable !comment command toggle, enable !tts command toggle, speak cooldown seconds, TTS command cooldown seconds, custom blocked terms list.

Enable Discord integration, webhook URL. Alert toggles: ban, timeout, unban, untimeout. Quippy message toggles.

Backup on launch toggle, retention days, maximum backups, launch backup delay seconds. Manual backup creation, refresh list, restore from backup, delete backup.

Context relevance, human memory, contradiction awareness toggles. User-specific memory, fact history window, fact staleness hours, memory time window.

Conversation history and learned facts help the bot personalise replies without relearning the same viewer context every stream.

Fact staleness and time-window controls decide how much old context should influence new replies.

Recent response history helps the bot avoid repeating the same phrasing or structure too often.

Use manual or scheduled backups before major updates, persona changes, or large engagement resets. Stop the bot before restoring a backup.

Enter the license key used when license enforcement is enabled. The field is locked while the chatbot is running - stop the runtime before changing it.

Validate the key through the local backend and read the returned status message. Validation failure shows the exact reason.

Once validated, this screen shows whether the key is a Trial or Standard license. Trial keys carry an expiry date and show a warning banner once they are within 7 days of running out.

While a trial key is active, the bot occasionally posts a short StreamLinkr promo message in chat (roughly every 45 minutes, only while live). This stops automatically once you upgrade to a Standard key.

Number of concurrent AI task workers. Raise this if the message queue builds up faster than workers can drain it.

Maximum number of tasks that can wait in the queue before new ones are dropped.

Seconds a worker waits before abandoning a stalled task and freeing the slot.

Total workers, active workers, idle workers, and queued message work.

Engagement worker capacity and queue pressure.

CPU, heap, memory, uptime, and live metrics connection status.

Filter logs, clear output, open logs folder, toggle auto-scroll.

Health and metrics output reveal memory pressure, worker queue buildup, Streamer.bot disconnects, provider failures, and database startup issues.

The runtime panel shows bot status and Start, Stop, Restart buttons. When using Ollama locally, a Restart AI model button also appears.

If program paths are configured, Start buttons appear for Streamer.bot, Ollama, Speaker.bot, and pear-desktop so you can launch dependencies from within StreamLinkr.

The topbar tracks unsaved changes. Save changes writes config to the backend; Reload config discards local edits and reloads from the backend.

Topbar search finds settings by label or keyword and jumps directly to the relevant screen, focusing the panel.

Closing the desktop app asks whether to keep running in tray or quit. Quit requests a graceful backend shutdown before closing.

When minimised to tray, the app can restore the window and expose start, stop, restart, show app, and quit actions from the tray menu.

The desktop shell blocks duplicate app instances and can offer administrator relaunch when install or data paths need elevated write access.

Enable Twitch, Kick, and/or YouTube independently. Set up at least one platform so the bot knows where to listen.

Bot name (defaults to StreamLinkrBot) and Streamer name per platform - used to recognize the bot's own messages and the streamer's messages in chat.

Fuzzy-match sensitivity from 0 to 1 (default 0.75) used when matching usernames against the configured bot/streamer names.

Twitch-only feature - silently ignored on Kick and YouTube. Requires Twitch enabled in Accounts.

Streamer.bot action name (e.g. "Create Twitch Clip") that StreamLinkr fires when a viewer runs !clip.

Per-user cooldown in seconds (default 60) and a global cooldown in seconds (default 30) that applies after any !clip use.

Use personality toggle posts an AI-styled chat announcement with the clip URL once Streamer.bot confirms the clip was created.

Post clip to Discord toggle plus a dedicated clip webhook URL, separate from the general Discord alert webhook.