How To Use SheevChat A.I. Features

SheevChat A.I. can write messages in a consistent Bot Voice, respond to stream events, create reusable announcements, and help draft content. You decide which modules this installation may use and which features may run automatically.

Connect and Enable A.I.

  1. Open Settings → Connect.
  2. Find SheevChat A.I. and select Connect SheevChat A.I.
  3. Open Bot Commands → A.I..
  4. Expand Modules & Usage.
  5. Enable each granted module you want this installation to use.
  6. Expand a module and configure its individual features.

Access and local permission are separate. An account grant makes a module available; the module switch allows this SheevChat installation to use it. Newly granted modules start off. Refresh Access checks for updated grants.

Where shown, the module card also displays its current allowance and usage. Turning off a module stops its registered features locally but does not remove the account grant.

Choose the Bot Voice

Open Bot Voice and choose a built-in voice, or select Create Custom Voice to describe how the bot should sound. The active voice is shared by A.I. features used through Commands, Automations, Timers, and SheevPad.

Under Response Rules, keep the content protections you want and add any channel-specific rules. Response Rules take priority over the selected voice. Save voice edits and Response Rules separately.

Action Builder also provides A.I. → Set Bot Voice. It can activate a saved voice or create one from custom instructions and variables. Remember that the selected voice remains active for later A.I. output; it is not limited to that one sequence.

Understand Automatic and Action-Based Features

An automatic feature watches its supported event or cadence once you enable its Run automatically switch. Use its platform switches to choose where it may send.

An Action Builder feature runs only when a Command, Automation, Timer, or SheevPad button reaches that action. This gives you exact control over the trigger, filters, destination, and action order.

For event-scoped automatic features:

  • Cross-post off: respond only on the platform where the event happened.
  • Cross-post on: also send to every enabled destination under Send to.

For cadence features, About every is intentionally approximate. They wait while no selected chat is live and avoid crowding other SheevChat messages. A configured Timer or Automation using the equivalent action takes priority over that feature's automatic cadence.

Conversation Starters

Conversation Starters supplies prepared, sourced questions that can give chat something easy to discuss.

  1. Enable Conversation Starters → Prepared chat questions.
  2. Choose the categories you want included.
  3. Set its automatic cadence and destination platforms, or leave automatic use off.
  4. Expand Conversation Starters to inspect the current pool.

Use Refresh Questions to check for a newer prepared package. Next Topic selects an unused question. Mark Used advances without sending it, while Release returns it to the unused pool.

Action Builder includes:

  • Refresh Conversation Starters
  • Select Next Conversation Starter
  • Send Current Conversation Starter
  • Mark Conversation Starter Used
  • Release Conversation Starter

Category choices apply to the A.I. panel and to Commands, Automations, Timers, and SheevPad actions.

Raids & Creator Intros

This module has two different features:

  • Raid greeting can automatically greet incoming Twitch or Kick raids. It uses the raider identity and available creator context.
  • Creator introduction is action-based and creates a short introduction for a specified Twitch, YouTube, or Kick creator.

For a controlled raid greeting, create an Incoming raid Automation and add A.I. → Generate raid greeting. That action is valid only for incoming Twitch and Kick raid events.

Use Generate creator introduction in a Command, Automation, Timer, or SheevPad button. Supply a username or variable such as {user}, choose the creator platform, and choose where the finished introduction should be sent. TikTok can receive generated output, but TikTok creator-profile lookup is not available.

Viewer Greetings

Viewer Greetings covers:

  • First-time chatters
  • Returning viewers
  • New followers
  • !lurk and !unlurk responses

Enable Viewer greetings for automatic event responses and select its destinations. In Action Builder, Generate viewer greeting must be used with a First Chat Ever, Welcome Back, or New Follow Automation trigger.

If a generated viewer greeting is rejected by the safety checks, SheevChat uses a short neutral greeting instead of exposing diagnostic or unsafe text to chat.

Lurk and return replies prepares the matching !lurk and !unlurk responses together. The platforms enabled on the actual Lurk Commands remain authoritative; use Open Lurk Commands from the A.I. panel to review them.

Community Reactions

Community Reactions creates voice-aware responses to supported events such as subscriptions, gifts, Hype Trains, birthdays, and other community-support activity.

For exact control, add A.I. → Generate community reaction to an Automation. Choose:

  • Chat only to send the response as a message.
  • Overlay alert only to display the generated text in an Overlay Builder alert.
  • Chat and overlay alert to do both.

Overlay output requires a saved Overlay Builder source and alert preset. Set the alert duration and keep that Browser Source active in OBS or Streamlabs Desktop.

Channel Point celebrations do not generate implicitly. Create a Channel Point Redemption Automation and add the appropriate A.I. community-reaction action when you want a redemption to receive an A.I. response. This keeps paid or limited redemptions under your explicit control.

Stream Moments

Stream Moments contains three features:

  • Stream opening message writes an opening message when the stream begins or when you invoke its action.
  • After-ad message waits for the triggering Twitch ad to finish, applies the configured extra delay, and welcomes viewers back.
  • Recent chat summary summarizes meaningful recent viewer conversation. Idle and emote-only windows are skipped.

When a recap exceeds a platform's message limit, SheevChat sends it as ordered chat messages so the complete recap is preserved.

Action Builder provides Generate stream opening message, Generate after-ad message, and Generate recent chat summary. The after-ad action must be inside an Ad Break Automation. The opening and recap actions can also be used in Commands, Timers, and SheevPad.

Optional guidance can focus the result, such as Tonight is community game night or Focus on viewer suggestions.

Scheduled Chat Writer

Scheduled Chat Writer turns one announcement into less repetitive recurring chat messages.

Automatic mode uses its configured cadence while a selected chat is live. For a dedicated announcement:

  1. Create a Timer.
  2. Under Do, add A.I. → Generate scheduled message variations.
  3. Enter the base announcement and optional goal.
  4. Configure the Timer's platforms and schedule.

The first run creates and stores 10 Bot Voice variations. Later runs send the next unused variation. After all 10 are used, the following run prepares a new set. A configured Timer's base announcement takes priority over the module's general automatic schedule.

Writing Tools

Writing Tools can rewrite, add hype, calm the tone, shorten, translate, or draft a poll, stream title, social post, or clip caption.

For a safe preview:

  1. Expand Writing Tools in the A.I. tab.
  2. Select a tool or starter example.
  3. Enter the source text and optional guidance.
  4. Select Generate Draft.
  5. Review and copy the result.

The Writing Tools workspace never posts automatically.

To generate and send text from a workflow, add A.I. → Generate with Writing Assistant to a Command, Automation, Timer, or SheevPad button. Choose the tool, source text, optional guidance, translation language when applicable, and Send to destination.

Bot Conversations

Bot Conversations lets the bot answer viewers who directly mention or reply to it. Enable Direct bot replies, then choose the platforms on which the bot may respond.

Replies use the active Bot Voice, Response Rules, and relevant available chat context. It does not answer every ordinary message; the viewer must directly address the bot through a mention or reply.

Start with one platform and observe the results before enabling it everywhere. Tighten the Bot Voice and Response Rules if replies are too long, too energetic, or outside your channel style.

Local Chat Context and Privacy

Local Chat Context is optional and off by default. When enabled, messages are stored in SheevChat's local SQLite database for context-aware features.

You control:

  • Retention from 1–365 days
  • Maximum stored messages
  • Immediate removal with Clear History

Enabling storage does not send the entire history anywhere by itself. A registered feature sends only the bounded context required for its request. Accounts marked AI Context under Settings → Filters are never stored. Bot and self echoes may remain locally for continuity but are excluded from generated context.

Private contact, payment, and account identifiers remain protected regardless of custom Response Rules.

A.I. Is Separate from Local A.I. Voices

Piper and Kokoro are optional local text-to-speech engines under Settings → Dependencies and Settings → Audio. They convert text into speech; they do not write the text or enable SheevChat A.I. modules.

You may use a SheevChat A.I. feature to generate a message and a Piper or Kokoro voice to speak it, but they are configured independently.

Common Problems

  • The A.I. tab is locked or missing: Connect SheevChat A.I. and confirm the account has a current module grant.
  • A module says Granted · Off: Enable that module locally under Modules & Usage.
  • A feature does not run automatically: Enable both the module and the feature's automatic switch, then select at least one live destination platform.
  • An action is unavailable: The required module or individual feature is not granted and enabled.
  • An action reports the wrong trigger: Some actions require a specific Automation event, such as Incoming raid, Ad Break, First Chat Ever, Welcome Back, or New Follow.
  • Nothing posts from Writing Tools: The workspace is preview-only. Copy the draft or use the Writing Assistant action when sending is intentional.
  • A cadence feature is quiet: It pauses without a selected live chat, varies its timing, and may defer to a configured Timer or Automation.
  • Replies use the wrong style: Check the active Bot Voice, then review Response Rules because rules override voice instructions.
  • Output goes to the wrong chat: Review the feature's Send to platforms, Cross-post setting, or the action's destination.
  • Usage is exhausted or unavailable: Review the module allowance, refresh access, and confirm the connected account still grants that feature.
  • An event response appears on only one platform: This is the safe default. Enable cross-posting explicitly if the response should also reach other selected platforms.

For workflow fundamentals, see Action Builder and Action Sequences, Creating Automations, and Creating Timers.