Skip to main content

Choose your software

You do not need a collection of AI applications. For the workflow in this guide, three current choices cover most needs.

For a normal Windows user who mainly wants to search PDFs and reports, start with AnythingLLM Desktop. Use LM Studio when your main need is to download, test and run local models. Use Open WebUI when you want a more configurable self-hosted environment and are comfortable managing more of the setup yourself.

The short comparison​

ApplicationBest forWindowsMacDocument searchCan work offlineSetup level
AnythingLLM DesktopBeginner document researchYesYesStrong workspace/RAG workflow with citationsYes, with local providersEasy
LM StudioRunning and testing local modelsYesApple SiliconLocal document chat/RAGYes after models/runtimes are downloadedEasy
Open WebUIAdvanced local or shared AI workspaceYes through supported deployment methodsYes through supported deployment methodsStrong knowledge-base/RAG toolsYes when configured for itMore involved

Ollama is also important, but it is primarily a local model engine rather than the beginner document workspace recommended here. AnythingLLM can use Ollama underneath its interface, and Open WebUI commonly connects to it.

AnythingLLM Desktop​

AnythingLLM Desktop currently provides a one-file desktop installer, requires no account, and can keep models, documents and chat history on the machine. Its document workspaces make the relationship between a set of source files and the questions you ask relatively easy to understand.

It is the best fit for this guide because the reader can move from installation to a useful multi-document workspace without Docker, Python or a terminal.

It supports local language models, local embeddings and a local LanceDB vector database. Its Windows storage directory is inspectable, so you can see the parsed document content, vectors, model files and application database rather than treating “local” as an invisible promise.

Advantages include simple installation, document-centred workspaces, source retrieval, local model support and enough configuration to grow beyond the beginner setup.

Limitations include the fact that it can also connect to cloud providers and external tools, so a user still needs to check what is configured. The default local embedding model is lightweight but mainly English-focused, which is why this guide includes a multilingual embedding option for mixed English and te reo Māori collections.

LM Studio​

LM Studio is a polished desktop application for finding, downloading and running local models. It is particularly useful when you want to understand which model is actually loaded, how much memory it uses, and whether it is running on CPU or GPU.

Its current documentation confirms that downloaded models, document chat/RAG and its local server can operate without internet access. Document processing for local document chat is performed on the device.

LM Studio has become more than a purely local application and can expose both local and cloud choices. For this kaupapa, deliberately choose a model marked as local and confirm offline operation after the required model and runtime files have been downloaded.

Advantages include excellent model management, clear hardware information, straightforward local model testing and useful local API support. It also works well as the local model or embedding provider behind another application.

Limitations are that its document workflow is less focused on reusable research workspaces than AnythingLLM. It is also easy for a new user to become distracted by model choices rather than the document task.

Open WebUI​

Open WebUI is a capable self-hosted AI interface with document knowledge bases, RAG, citations, configurable document extraction and support for local and remote model providers.

It is useful for an organisation or advanced user who wants more control over the whole AI environment, potentially including multiple users and a separate Ollama service. Current Open WebUI documentation describes local document RAG, configurable embedding models and an offline mode.

The extra flexibility also creates extra responsibility. A genuinely offline configuration requires the required models and embeddings to be downloaded first, and external APIs or web tools can still be configured. The Open WebUI offline tutorial itself notes that the documented approach is a community contribution rather than an officially supported air-gap guarantee.

Advantages include strong knowledge-base functions, configurable embeddings and extraction, citations, and the ability to build a controlled shared environment.

Limitations include more installation and administration, more ways to connect external services, and more settings that a beginner can accidentally change.

What about Ollama?​

Ollama runs models locally and exposes them to other software. It is useful infrastructure, but it is not the main beginner interface in this guide. On Windows it can be installed directly, and AnythingLLM Desktop currently includes built-in local LLM support powered through Ollama.

Use Ollama when another application needs a local model service, or when you are comfortable managing models more directly. A beginner who only wants to search reports can start in AnythingLLM without learning the engine first.

Recommendation by reader​

Choose AnythingLLM Desktop if your question is, “How do I put a folder of reports into local AI and ask questions across them?”

Choose LM Studio if your question is, “How do I download and run a particular local model and understand what my computer can handle?”

Choose Open WebUI if your question is, “How do I build a more configurable local or organisational AI workspace and I am prepared to administer it?”

For the rest of this beginner path, the screenshots and steps assume AnythingLLM Desktop on Windows 11 unless stated otherwise.

Sources​

Last verified: 22 August 2026

Choose for the research workflow​

For iwi and hapū researchers, Māori land teams and people working with historical records, the important differences are installation effort, document handling, local storage and whether findings stay tied to source pages. Continue with Why local AI?, Install local AI and Search your documents.

Last updated on