Feature Guide

An AI co-pilot that knows your stream.

Ask a question mid-stream and get a straight answer from your own live stats and channel notes. Ask it to switch scenes and it does. The AI Sidekick runs on your own PC — no cloud model, no per-message cost, and it keeps working with the internet off.

From switched on to hands-free in four steps.

Turn it on → tell it about your stream → ask → let it act. All of it on your own hardware.

Step 1 · Switch it on

One toggle starts the local AI.

The Sidekick runs on Ollama, on your own PC. Flip the AI Engine toggle and streamerOS starts the engine for you — no terminal, nothing to babysit — and stops it again when you want the resources back. Pick and download models from Settings; llama3.2 for chat and nomic-embed-text for memory are the defaults.

Settings · AI

Local AI
AI Engine
Chat modelllama3.2
Memory modelnomic-embed-text
Address127.0.0.1
Running locally — nothing leaves this PC.

Step 2 · Tell it about your stream

Write your Streamer Bible once.

Your Streamer Bible is a JSON file in your workspace: your persona, current game, moderation rules and OBS scene names. Save an edit and the Sidekick reloads it automatically. Those facts sit above everything else it knows, so “which scene is gameplay?” is answered from your file, not a guess.

Workspace

streamer_bible.json
PersonaChill ranked grinder
GameValorant
ModerationNo spoilers, no slurs
scene.gameplayScene 3
scene.brbScene 5
Saved — the Sidekick reloaded your Bible.

Step 3 · Ask anything, mid-stream

Plain words in, a short answer out.

Replies stream in token by token, designed to show the first word in under ~800 ms on RTX 3060-class hardware. The Sidekick sees your live stats — chat speed, the room’s vibe and sentiment — alongside your Bible, and keeps answers to a line or two because you are live.

Sidekick · streaming reply

YouHow’s chat feeling right now?
SidekickChat’s moving fast and the mood is positive — that last round landed.
YouWhich scene is gameplay again?
SidekickScene 3 — straight from your Bible.

Step 4 · Let it act

Ask, and it does the thing.

Say “switch to gameplay” and it switches your OBS scene. It can also read recent chat, search your Chat Archive, look up this stream’s Super Chat revenue and design an Aura Scene overlay — chaining up to four steps for one request. If OBS isn’t connected, it tells you instead of pretending.

Sidekick · action

Youswitch to gameplay
OBS scene → Scene 3 (gameplay)
SidekickDone — you’re on gameplay.

Actions

Six things you can just ask for.

No menus mid-match. Type the request the way you would say it, and the Sidekick picks the right tool.

Switch your OBS scene

“switch to gameplay”

Matches the scene names in your Streamer Bible and switches over your local OBS connection. If OBS is not connected, it says so.

How OBS Bridge works

Read recent chat

“what is chat asking right now?”

Pulls the latest live chat lines, so you can catch the question you scrolled past mid-fight.

Search your Chat Archive

“what did chat say about my crosshair last stream?”

Searches the chat you saved from earlier streams — straight from your disk.

See the Chat Archive

Check Super Chat revenue

“how much have I made in Super Chats this stream?”

Looks up this stream’s Super Chat total and per-currency breakdown — the same numbers as your revenue panel.

See the revenue panel

Design an overlay

“design a starting soon screen”

Builds an Aura Scene overlay from your description, saves it, and pushes it live to your overlay.

See Aura Scene

Recall what you told it

“who is my co-host?”

Brings in facts you asked it to remember, automatically, whenever they are relevant to the question.

How Creator Memory works

Creator Memory

Tell it once. It remembers.

Start a message with “remember that…”, “note that…” or “keep in mind…” and the Sidekick stores the fact in a private vector database (LanceDB) on your disk. Your co-host’s name, a sponsor’s do-not-say list, the running joke with your mods.

Later questions pull in the relevant memories automatically — you never have to repeat yourself or dig for a setting. Nothing leaves the PC. Why streamerOS stays zero-cloud.

Sidekick · memory

YouRemember that my co-host is Max.
Saved to creator memory on this PC

Next stream

YouWho’s joining me for the duo queue?
SidekickMax, your co-host.

AI Insights

Advice from your own analytics.

Import the analytics CSV exports from YouTube or Twitch and the Sidekick writes two to four insight cards from your numbers, each tagged growth, brand or content — so you know whether a tip is about reach, sponsors or what to stream next.

It needs at least 7 days of data. With less than that, it asks you to import more rather than padding the list with guesses.

AI Insights · example

growth

Weekend streams hold viewers longer

Your peak viewers land on Saturdays. Test moving one weekday slot to Sunday.

content

Chat spikes during ranked clutches

Your fastest chat comes in close rounds. Lead your next VOD title with the clutch.

brand

Steady schedule reads well to sponsors

Consistent weekly streams make an easier pitch. Put the schedule in your media kit.

Frequently asked questions.

Does the AI Sidekick send my prompts or stream data to the cloud?

No. The Sidekick runs on Ollama, a local AI engine on your own PC, and streamerOS only talks to it at 127.0.0.1. Your questions, live stats, Streamer Bible and memories stay on the machine. There is no per-message cost, and once your models are downloaded it keeps working with the internet off.

What do I need to run it?

Ollama installed on the PC, and an RTX 3060-class GPU is recommended. The defaults are llama3.2 for chat and nomic-embed-text for memory; you choose and download models from Settings. The AI Engine toggle starts or stops the local engine for you, so there is nothing to manage in a terminal.

Is it as smart as ChatGPT or other cloud AI?

No, and it is not trying to be. Local 3B-class models are fast and private, but they are less capable than large cloud models. The Sidekick is built for short, grounded answers about your own stream and for simple actions like switching a scene — not long essays or deep research. If your hardware can handle a bigger model, you can pick one in Settings.

Will running a local AI model slow down my stream?

Generating a reply uses your GPU while it runs, so the Sidekick keeps answers to a sentence or two and only generates when you send it a message. The AI Engine toggle stops the engine completely when you want every bit of headroom. Note that the 1.8% CPU figure quoted for streamerOS is the core app under a live 1080p60 game — it does not include AI inference, which depends on your model and GPU.

Can it switch OBS scenes on its own?

Only when you ask. Say "switch to gameplay" and it maps that to the scene names in your Streamer Bible and switches over your local OBS connection, chaining up to four steps for one request. If OBS is not connected, it tells you so instead of pretending the switch happened.

Does it make things up?

It is grounded, not infallible. Facts from your Streamer Bible are put ahead of everything else it knows, revenue and chat answers come from your own stored data, and a chat-archive search that finds nothing is reported as nothing rather than invented. Like any language model it can still get something wrong, so check a number before you read it out on stream.

What happens when the free trial ends?

streamerOS is free for 7 days, then $29 once. Without a licence, plain chat with the Sidekick stays free for up to 10 messages a day. Actions (scene switching, chat, archive and revenue lookups, overlay design), creator memory, the Streamer Bible and AI insights are part of the $29 licence.

More questions answered on the full FAQ.

An assistant that never leaves your PC.

The AI Sidekick ships at launch in streamerOS, running on the local AI engine on your own hardware. Free for 7 days, then $29 once.

Launching February 2027
Back to all features