Frequently Asked Questions

A good FAQ should answer the questions users ask after the first excitement wears off and real usage begins.

This page focuses on operational reality: where data lives, what sync does and does not do, why models forget, what Premium actually buys, and what tradeoffs come with a privacy-first system.

Where is my data stored?

Your working data is stored locally on your device first.

If you enable Premium cloud sync, the synced payload is meant to be encrypted before it is stored remotely. The important operational point is that Abolitus is not supposed to treat your story as normal plaintext server content.

In practical terms, think of the local device as the main workspace and the cloud as an optional continuity layer.

Is Abolitus a hosted chatbot service or a client?

It is best understood as a client and workspace rather than a traditional hosted chatbot destination.

You bring your own provider routes, your own keys, your own model preferences, and your own local workflow. Abolitus is the environment that helps you organize, route, persist, and continue those sessions.

Does Abolitus train on my chats?

The product is designed specifically to avoid turning your chats into normal readable server-side content. That is one of the central reasons the architecture is local-first and zero-knowledge oriented.

However, remember the separate provider boundary: if you send prompts to an external provider, that provider's own policies still matter.

What happens if I clear browser site data?

If you wipe the local site data before exporting or syncing what you care about, you should assume the local workspace is gone from that device.

This is one of the biggest real risks in any local-first product.

If you care about durability, do not rely on memory or good intentions. Use backup, export, or encrypted sync deliberately.

What is the safest backup mindset?

Treat important creative work the way you would treat important writing in any serious tool:

1. Keep local continuity in mind.
2. Export or sync before risky device changes.
3. Do not assume browser data survives every cleanup, reinstall, or storage policy change.
4. Keep recovery material organized if you use encrypted cloud continuity.

Why did the AI forget older parts of the story?

Because the model only sees the part of the workspace that still fits into the active context window.

Older messages can remain saved locally while no longer fitting into the next prompt.

If continuity matters, move durable facts into:

• Personas.
• Lorebooks.
• Summaries.

Forgetting is usually a context-management problem, not proof that the whole app lost your data.

Many users confuse these two situations:

• the chat still exists in storage,
• but the model cannot see all of it in the next request.

Those are very different problems.

Does Premium fix forgetting?

No.

Premium helps with encrypted continuity across devices. It does not create infinite context inside the model.

Premium is about persistence and device flow, not about changing the fundamental token limits of the underlying model route.

If Premium does not fix forgetting, what does?

Usually some combination of:

• cleaner token budget management,
• stronger summaries,
• better lore organization,
• more intentional persona setup,
• and choosing a model route with enough context for the job.

The model refused or became generic. Is Abolitus censoring me?

Usually no. The most common cause is the model route, not Abolitus itself.

Try:

1. A stronger or better-suited wrapper preset.
2. A different model family.
3. A local route if the current cloud route is heavily aligned.
4. A cleaner prompt and lower context clutter.

Newer mainstream cloud families can differ a lot in how strongly they resist aggressive scenes.

If you want a cleaner diagnosis, change one variable at a time. If you swap provider, wrapper, sampler, and persona all at once, you lose the ability to tell what actually helped.

What kinds of models can I use?

That depends on the provider route you configure.

Abolitus is designed to work with supported provider surfaces rather than shipping a single bundled model catalog. In practice, model availability changes over time as routes expose different families.

The important point is that the app is not limited to one model personality. You can choose routes based on tone, price, speed, context, and alignment behavior.

Do API keys sync across devices?

No.

Provider credentials and TTS credentials stay on the device where you entered them.

This is intentional. Credentials are a different risk category than ordinary synced settings.

Why do API keys stay local?

Because syncing secrets casually is one of the fastest ways to make a privacy-first app less private.

Keeping provider credentials local reduces the blast radius if one device, one sync surface, or one account workflow is compromised.

It also means you can choose different providers on different devices if that suits your workflow.

Do I need Premium to use local chat?

No.

Local chat, OpenRouter routing, NanoGPT routing, and most core creative features remain usable without Premium.

Premium is not meant to gate the basic existence of the product. It mainly pays for continuity features that require operating remote infrastructure for you.

What exactly does Premium pay for?

Premium pays for encrypted continuity features such as background cloud sync and Desktop Tunnel. It does not replace your provider bill and does not unlock the basic existence of local chat.

Does Premium include model usage?

No. Premium is separate from the cost of your model provider.

If you use a paid provider route, that billing relationship remains between you and the provider.

What Premium options exist?

Current pricing and availability can change, but the product currently distinguishes between time-based extensions and the limited Founder tier. Always check the live billing page for the latest details before making assumptions.

Can Abolitus recover my data if I lose my recovery material?

No.

That is a direct consequence of the zero-knowledge trust model. If the service does not hold your recovery secret, it also cannot reconstruct it for you later.

This is one of the most important tradeoffs to understand before you depend on encrypted cross-device continuity.

Is that a bug or a feature?

It is a feature with real cost.

People often want both of these at the same time:

• "do not let the service know my secret,"
• and "let the service recover my secret whenever I ask."

Those two goals conflict. A serious privacy model forces you to choose where recovery power lives.

Can I use current flagship models in Abolitus?

Often yes, if your chosen provider route exposes them.

For example, current families available through supported routes may include Claude Opus 4.7, GPT-5.5, Gemini 3.1 Pro, Gemma 4, and others. Availability changes with the provider route, not with a bundled Abolitus model catalog.

The safest way to think about this is: Abolitus is model-flexible, but the exact menu comes from the route you connect.

What is the safest way to move between devices?

Use the method that matches the job:

• Premium cloud sync for durable cross-device continuity.
• Device Handoff for moving an active scene quickly.
• Desktop Tunnel for using a desktop-hosted local model remotely.

Those features solve different problems, and choosing the right one prevents a lot of confusion.

When should I use Device Handoff instead of sync?

Use Device Handoff when you want to continue the current working session on another device quickly.

Use sync when you want durable encrypted continuity across time, not only a short movement of the active experience.

When should I use Desktop Tunnel?

Use Desktop Tunnel when your preferred model lives on your desktop machine but you want to access it from somewhere else. It is about reachability to your own desktop-hosted route, not about replacing all other sync or handoff features.

Why does the app feel different on mobile?

Because mobile is a different environment: less screen space, different typing ergonomics, different battery constraints, and often shorter sessions.

That is why some visual and workflow choices are intentionally local or adaptive rather than mirrored perfectly from desktop.

Why is my wallpaper or theme setup different on another device?

Some appearance choices are intentionally device-local so one machine's visual preference does not overwrite another machine's practical setup.

Can support read my chats to debug a problem?

The product is intentionally structured to reduce reliance on that kind of support model.

Expect support to work better with:

• reproducible steps,
• route names,
• screenshots that you intentionally sanitize,
• high-level symptoms,
• and diagnostics that do not require pasting private scenes.

That limitation is not accidental. It is part of what a zero-knowledge posture looks like in practice.

How should I ask for support in a zero-knowledge product?

Expect support to rely more on high-level diagnostics, reproducible steps, and sanitized technical signals than on handing over your plaintext stories. That limitation is part of the privacy model, not an accidental weakness.

The best support request usually includes:

1. what you were trying to do,
2. what route or provider was involved,
3. what changed after the failure,
4. whether the issue is local-only or cross-device,
5. and the smallest reproducible example you can provide without exposing sensitive content.

That gives support something actionable while preserving the reasons you chose a privacy-first tool in the first place.