Questions before you play.

SoloRoleplayer// is a private solo roleplay shell built by WordBlock Labs, a solo dev project. You write your side. The companion writes its side. It is made for people who want scene play, not a generic chatbot.

You need three things: SoloRoleplayer// Free, Cute LM, and the tested model mlx-community/Qwen3.5-9B-MLX-4bit. If you want the stronger lane, add the optional SoloRoleplayer// LoRA after that.

No. The current Free lane is built around Cute LM instead. Cute LM runs locally on your machine and is preset for SoloRoleplayer//.

Download mlx-community/Qwen3.5-9B-MLX-4bit from Hugging Face. That is the tested model lane the Free version is shaped around.

Cute LM is the free local runtime from WordBlock Labs. It starts the tested model on your own machine at 127.0.0.1:1234 so SoloRoleplayer// can find it without a bunch of setup pain.

No. The local Free lane does not need a SoloRoleplayer// account. Cute LM runs locally. Your local play flow can stay on your own machine. However, we'd love your membership to our Ko-fi, and our newsletter, so we can keep cool projects like this one free.

You need internet one time to download the app, Cute LM, the model, and any optional LoRA. After that, if you run the local lane only, your play can stay offline. Shards, memory, and saves stay with you.

A Shard is your portable save. It is the thing you back up and carry with you. SoloRoleplayer// can also rebuild its local memory index from that save when you load it again.

STASIS means Short-Term AI Smart Intelligence System. It is the scene-memory lane. It helps the shell remember where you are, what just happened, what people are wearing, what the current goal is, and other in-character continuity details.

It is not supposed to. SoloRoleplayer// is built for a turn-based roleplay style, not a collaborative writing style. The player owns the player character, and we have guardrails in place to keep the companion in its own lane. The base LLM can still head-hop now and again. If you want to remove that problem almost entirely, we suggest our LoRA.

Because local model people complain about the same two problems all the time: head-hopping and persona loss. Our SoloRoleplayer// LoRA was built to almost completely stop those problems in the tested Qwen 9B lane.

Yes. The Free lane works with the tested model and Cute LM by themselves. The LoRA is the stronger add-on lane.

/stop leaves Holo, takes you back to Construct, and offers a shard backup.

/quit offers a shard backup flow, then logs you out to setup or world select.

/reset now keeps things simple. It lets you either factory reset the world back to its start state or delete the world and exit that slot.

/purge is the privacy-cleanup command. It wipes SoloRoleplayer// local world traces from the machine, including slot data, recovery leftovers, imported-world leftovers, preference leftovers, local media blobs, and local Tauri slot shards. Best practice: save a Shard first, then run /purge.

Yes. Use @build here in supported rooms. You get up to two room-asset slots per base room, and each one can have a name, alias, outside view, inside view, pictures, and music.

Temperature controls how wild or restrained the AI's word choices are. Lower temperature usually means more stable, direct, predictable replies. Higher temperature usually means more variety, more risk, and sometimes more weird mistakes. For roleplay, less is often more if you want stronger continuity and cleaner behavior.

Top P controls how wide the AI's word-choice pool is. Lower Top P narrows the pool and tends to make output cleaner and tighter. Higher Top P allows a broader pool and can make replies feel more surprising, but also less disciplined. If the model starts getting messy, repetitive, or overly random, try lowering Temperature first, then Top P.

WordBlock Labs. It is a solo dev tinkerer project. If you want to help support more work on SoloRoleplayer// and Cute LM, please visit Ko-fi.