WORLD AI GUIDE
EN
Tool families

Local and open AI: verify the control you really get

Running 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.

Local computing workstation with an automation diagram
AI-generated illustration.

When to explore this family

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.

How to compare tools

  1. Verify what genuinely works offline.

  2. Measure memory use, speed and quality on your own hardware.

  3. Inspect licences, updates and outbound connections for each component.

A repeatable trial

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.

  1. Isolation

    Install model and interface on a test machine. Disconnect it, then check which functions genuinely remain available.

  2. Measurement

    Record memory use, response time, answer quality and passage retrieval on the same questions.

  3. Maintenance

    Document versions, licences, updates, backups and how the setup can be exported or replaced.

Evidence for a decision

Output
An acceptable result on your hardware with controlled network flows.
Checks
Memory, latency, model licence and remote dependencies.
Evidence
Versions, configuration, memory measurements and disconnected trial.
Failure signal
A local claim when another component transmits the data.
Compare actual cost and keep a decision record →
20 active services in the directory

Services to examine

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 & openlocal

Docker Model Runner

local model runtime

Docker · US

Official site
View every tool in this familyUnderstand the selection method →

Related guides

RAG guide · 12 min

Build a reliable, citable RAG knowledge base

Start with a limited corpus and a named owner. Remove stale versions, retain source metadata and permissions, test retrieval separately from generation, display cited passages and teach the system to say when the corpus is insufficient.

Read the guide →

Questions before choosing

Does local mean nothing leaves the machine?

Not automatically. An interface may enable telemetry, updates or remote APIs; inspect actual connections.

Is the largest model always best?

No. Compare speed, memory, task quality and maintenance on your own hardware.

How do model and interface differ?

The model generates answers; the interface manages files, users and interactions. Their hosting and licences may differ.

Explore other tool families