First Launch

The first-run wizard appears until setup is complete. Its job is to help you connect one working model path so Row-Bot can answer a first chat message. Everything else can be configured later.

The first-run wizard helps you choose a local, hosted, subscription, or custom model path before optional setup begins.

Choose A Model Path

A model is the AI system that writes responses and reasons through tasks. Row-Bot can work with several kinds of model providers:

Path	Good For	Tradeoffs
Local Ollama	Privacy-focused local chats and tool use when your machine has enough resources.	Requires installing models locally. Larger models need more memory and disk space.
Hosted API provider	Strong models without local downloads.	Requires an API key, internet access, and provider billing. Prompt content can be sent to that provider.
Subscription account	Using a supported account-backed provider from inside Row-Bot.	Requires sign-in or token import and follows that provider's account terms.
Custom endpoint	Advanced local or self-hosted runtimes such as LM Studio, vLLM, llama.cpp, LocalAI, LiteLLM, or SGLang.	Requires endpoint details and compatible model behavior. Tool use needs enough context and function-calling support.

For beginners, start with the provider path you already trust. If you are unsure, Ollama is the simplest local-first path, while an API provider is usually the quickest path to strong hosted models.

Install Private Knowledge Search

Every normal desktop, source, and official Docker install uses the same setup step. The wizard offers Mixedbread Embed Large v1, a separate local model for semantic memory and document search, as a checked-by-default 675 MB download. It is not the model that writes chat responses.

The download starts when you finish setup and requires internet access to Hugging Face. Files go into Row-Bot's private cache; documents and memories are not uploaded, and normal recall stays offline after the initial download. Docker keeps this cache in its named /data volume.

You can uncheck the option and finish without it. Row-Bot continues with bounded lexical and graph fallback, so chat and memory do not stop working. Later, open Settings → Documents and use Download model. Retry local load retries an already cached model; Repair local model deliberately replaces a damaged download. Rebuild the document and memory vectors after changing the embedding provider or model.

Settings → Documents shows whether the local embedding model is ready and provides explicit download, retry, repair, and index rebuild actions.

Setup Center

Setup Center lets you finish optional setup areas later without blocking the first chat.

Setup Center is the place to finish or revisit optional setup after the first launch. It can guide you through models, documents, workflows, Designer, Developer, channels, accounts, MCP, plugins, Buddy, and voice. The wizard gets you started; Setup Center helps you build out the rest of your app over time.

Privacy, Cost, And Credentials

Local model runs can stay on your machine. Hosted, subscription, realtime voice, web search, account, channel, MCP, and plugin features can contact outside services when configured and used. Provider keys and account tokens should be entered only through the relevant Settings or account flow. Cost depends on the provider or service you choose.

Troubleshooting

• If the wizard cannot find a local model, start Ollama and install a model first.
• If a hosted provider connects but no models appear, refresh Providers and Models.
• If the private knowledge model download fails, check internet access to Hugging Face and try finishing setup again, or uncheck it and install it later from Settings → Documents.
• If you skip optional setup, open Setup Center or Settings later.