Chat and create
Ask questions, draft content, plan work, and get streaming responses in one browser workspace.
Private by design · open source · self-hosted
AI work that stays in your hands.
LiteMindUI brings chat, document questions, web search, and voice into one browser workspace. Run the model on your machine or connect a hosted provider when the job calls for it.
Start with the prebuilt Docker images
curl -fsSL https://raw.githubusercontent.com/debabratamishra/litemind-ui/main/install.sh | bash- Local or cloudChoose the model path
- Per-user isolationBuilt into the backend
- Browser nativeOpen at localhost:3000
LiteMindUI / workspace
What changed in the operations brief?
Source: operations-brief.pdf · page 4
Ask anything...
A focused workspace for the work already on your machine.
One workspace for the work you already do
The interface stays out of the way. The useful parts are close at hand.
Ask your documents
Upload PDFs, notes, reports, and spreadsheets. Hybrid search combines semantic and keyword retrieval before answering.
Search the web
Add live web results when an answer needs current information, with sources kept beside the response.
Work by voice
Use microphone input or the realtime WebRTC voice pipeline for hands-free conversations.
Why teams look twice
Private is a deployment decision, not a slogan.
LiteMindUI gives you the controls to decide where inference, identity, documents, and conversation history live. The project is open for inspection, and the hosted model path is optional.
Read the data and auth modelLocal first, cloud optional
Run Ollama on your machine. OpenRouter and Nvidia NIM remain available when you choose a hosted model.
Clear data boundaries
Self-hosted authentication and per-user isolation keep conversations, sessions, and RAG contexts separated.
A familiar interface
Once the stack is running, use LiteMindUI from any modern browser at http://localhost:3000.
Open for inspection
Read the source, run it yourself, and adapt the FastAPI backend or Next.js frontend to your needs.
See the shape of the product
A browser window, not another app to manage.
Bring the conversation, the documents, and the model settings into one place. The same interface works for a first local experiment and a larger self-hosted rollout.
- Frontend
- Next.js / TypeScript
- Backend
- FastAPI
- Retrieval
- ChromaDB + BM25
localhost:3000.From clone to conversation
Start with a small model. Grow from there.
Use the installer for the shortest path, or run the source with Docker when you want to inspect the stack. Local inference keeps the first run private; hosted providers are a configuration change.
Open the quick startRun the installer
Pull the prebuilt images and start the Docker stack.
curl -fsSL https://raw.githubusercontent.com/debabratamishra/litemind-ui/main/install.sh | bashOpen your workspace
Visit http://localhost:3000, create an account, and sign in.
Choose a model
Connect Ollama for local inference or configure a hosted provider.
For the people extending it
A backend you can trace. A frontend you can change.
Chat and RAG route through a pluggable skill layer. Retrieval combines vector and keyword search, and voice runs as a separate realtime pipeline.
Read the developer guideA practical base for a private rollout
Start with the open-source project, then adapt the parts your organization needs to own.
Private deployment
Run the application inside your network and keep model, document, and identity services under your control.
Identity and access
Adapt the self-hosted authentication layer and per-user data isolation to your access model.
Custom integrations
Connect internal tools, private models, document stores, and business-specific workflows.
Enterprise support
Tell us what has to stay private.
For deployment planning, identity integration, or a custom model and data-source connection, start with the use case and the constraints.
Discuss your deploymentBefore you ask
The details that affect a real deployment.
Short answers first. The repository and developer guide carry the implementation detail.
Which operating systems are supported?
LiteMindUI runs through Docker on systems that support the provided Compose setup. The browser interface works in current Chrome, Edge, Firefox, and Safari releases.
Can it run without the internet?
Yes, after dependencies and a local model are available. Offline use requires Ollama or another local inference service; web search and hosted models need network access.
Where is my data stored?
Your deployment controls storage. The canonical conversation and user-memory stores use PostgreSQL, while document and ingestion paths are configured on the host.
Do I need an OpenRouter or Nvidia account?
No. Ollama provides the local model path. Hosted providers are optional and require their own API credentials.
Is LiteMindUI ready for a company rollout?
The open-source project provides the base application, self-hosted authentication, and per-user isolation. Production deployments should be reviewed against your security, support, and compliance requirements.
Your workspace, your rules