LocalAI
Read the detailed profile →local AI engine
LocalAI
Official siteRunning a tool locally can give more control over data, but brings hardware, configuration and maintenance needs. Separate the model, interface and any external services that may be enabled.

Try a model without sending a document to a remote service.
Search internal documents with access controls you understand.
Set up a reversible environment for testing several models.
Verify what genuinely works offline.
Measure memory use, speed and quality on your own hardware.
Inspect licences, updates and outbound connections for each component.
Starting situation. Ask questions about ten internal documents on an isolated workstation. The test must work without a network connection and cite the passages used.
Install model and interface on a test machine. Disconnect it, then check which functions genuinely remain available.
Record memory use, response time, answer quality and passage retrieval on the same questions.
Document versions, licences, updates, backups and how the setup can be exported or replaced.
These active services are starting points from the directory. Editorial highlighting and score order this selection; they do not replace a trial on your own task.
local AI engine
LocalAI
Official sitelocal models
Ollama · US
Official sitemodel serving
vLLM Project
Official sitelocal models
LM Studio · US
Official sitelocal model runtime
Docker · US
Official siteself-hosted language models
Open WebUI · US
Official sitemodel serving
SGLang
Official sitelocal models
Jan AI · US
Official sitelocal model runtime
Apple · US
Official siteNot automatically. An interface may enable telemetry, updates or remote APIs; inspect actual connections.
No. Compare speed, memory, task quality and maintenance on your own hardware.
The model generates answers; the interface manages files, users and interactions. Their hosting and licences may differ.