Local Models

Local Models is the cross-provider model management screen inside RealTimeX.

Open it from Settings > AI Providers > Local Model Management.

What this page is for

Use this page when you want one place to inspect and manage models across local backends.

The current product supports:

• Ollama
• LocalAI
• LM Studio
• llama.cpp
• RealTimeX Local

This page is broader than RealTimeX Local, which focuses only on the managed built-in GGUF runtime.

Provider status

Each provider tab can appear as:

• online
• offline
• unconfigured
• loading

If a provider is unconfigured, RealTimeX expects you to finish its base URL or provider setup first in Settings > AI Providers > LLM, then return here.

What changes by provider

Not every provider supports the same actions.

In the current product:

• Ollama and LocalAI are pull-oriented model registries
• LM Studio mainly reflects models you already loaded inside LM Studio
• llama.cpp reflects the running server and local GGUF files
• RealTimeX Local uses the managed local models directory and built-in runtime

That means this page is best understood as one shell over several different local model ecosystems.

Common actions

Depending on the provider, the current UI can let you:

• add or pull a new model
• review installed model counts
• inspect size, modified time, digests, and IDs
• copy model identifiers
• load a model
• open model details
• edit per-model settings
• refresh or update metadata
• repair incomplete models
• delete models from local storage

Some actions are provider-specific, so a tab may expose fewer controls than another.

RealTimeX Local inside Local Models

The RealTimeX Local tab is where you manage the built-in GGUF inventory across the managed local models directory.

The current flow supports:

• recommended model downloads
• Hugging Face search
• manual repository or file entry
• default-model selection
• load and warmup behavior
• repair for incomplete artifacts

Use this tab when you want deeper inventory management than the simpler RealTimeX Local settings page.

Ollama and LocalAI

For Ollama and LocalAI, the current UI is oriented around pulling models by name or gallery ID.

Use these tabs when:

• you already run one of those providers
• you want RealTimeX to attach to that local model inventory
• you prefer provider-native model management instead of the managed built-in runtime

LM Studio

LM Studio behaves differently from pull-based providers.

In the current app, the usual expectation is:

• load a model inside LM Studio first
• then return to Local Models to inspect it from RealTimeX

If the tab looks empty, it usually means LM Studio is not currently serving a loaded model to the instance.

llama.cpp

The llama.cpp tab is tied to the running llama.cpp server and its model directory.

The current UI can help you:

• inspect the active model
• review GGUF files on disk
• download GGUF files from Hugging Face
• understand when the model directory or server configuration is incomplete

If the server connection fails or no model directory is available, fix the llama.cpp runtime setup first, then return to the tab.

Model settings

Some providers expose a model settings modal with runtime tuning controls such as:

• keep-alive time
• temperature
• context window size
• max tokens
• advanced sampling controls

These settings are useful when you want a model to stay warm longer or behave differently without changing the whole provider configuration.

Incomplete and repairable models

The current UI can flag unhealthy local models, including cases like:

• missing multimodal projector files
• missing split GGUF parts
• partially downloaded artifacts

When a model is marked incomplete, repair it or remove and re-download it before using it in production workflows.

Typical workflow

1. Configure the provider in Settings > AI Providers > LLM if needed.
2. Open Settings > AI Providers > Local Model Management.
3. Choose the provider tab you want to work with.
4. Pull or download a model if none exist yet.
5. Inspect details or adjust per-model settings if needed.
6. Load the model where that provider supports explicit loading.
7. Set the related provider as your active LLM in Settings > AI Providers > LLM.

When to use Local Models

• Use it when you manage more than one local provider.
• Use it when you need visibility into model files, digests, health, and provider-specific operations.
• Use it when you want to troubleshoot why a local provider looks empty, offline, or incomplete.

For the managed built-in local path, see RealTimeX Local. For provider selection at the instance level, see Large Language Models.