Model freedom

Local Models in Disp8ch: run the core workspace without a model-provider key

Disp8ch can use hosted providers, OpenRouter, or local OpenAI-compatible runtimes. The local path supports Ollama, LM Studio, llama.cpp, vLLM, SGLang, and similar endpoints.

Install Disp8chRead quickstart
Disp8ch local model setup showing model recommendations and hardware fit
What it does

Turns the feature into usable workspace objects.

Disp8ch treats Local Models as part of the operating loop. The result is visible, editable, and connected to the other surfaces.

  • Runs core chat, memory, workflows, agents, boards, Council, local document work, and design work through local endpoints.
  • Advises on model fit using RAM, CPU, GPU, VRAM, installed runtimes, and installed models.
  • Lets each agent or workflow use the model configuration that fits its job.
How it works

Context flows through the app instead of being trapped in chat.

The feature reads from the same workspace objects used by the rest of Disp8ch, then writes back through confirmation-gated actions when work changes state.

  • Onboarding can inspect the machine and recommend balanced, speed, or quality local options.
  • The advisor shows commands for runtimes such as Ollama or llama-server, but it does not download or activate models without your action.
  • OpenAI-compatible base URLs make local and hosted model routes consistent.
What makes it different

Built for operators who need auditability.

The feature is designed around explicit state, reviewable outputs, and practical handoffs to the next place work should go.

  • Local setup is treated as a first-class path, not a hidden developer mode.
  • Recommendations are explainable and conservative about memory risk.
  • Cloud keys remain optional for core local workflows.
Works with

Local Models is strongest when it connects to the rest of the workspace.

A typical job starts with a user request, pulls in sources or app state, creates a plan, asks for confirmation when something changes, and leaves a trace in the right surface.

WebChatAgentsWorkflowsMemoryData Sources
Prompt to tryLocal Models
Compare three local model runtimes for an 8 GB VRAM laptop and recommend a practical setup.
Related features

Keep moving through the operating loop.

All features
Disp8ch WebChat showing an assistant conversation with app actions and workspace context
Plain English control

Agentic WebChat

WebChat is the command surface for the whole workspace. Ask questions, build workflows, inspect state, create tasks, run research, and hand work to agents without losing the audit trail.

diagram of Disp8ch agent operations across tools, approvals, budgets, and work outputs
Agent control plane

Agent Operations

Agents in Disp8ch have profiles, tools, model overrides, skills, extensions, budgets, channel routing, workspace files, and observable activity.

Disp8ch Memory screen showing stored context, candidates, and review controls
Durable context

Memory

Memory in Disp8ch is local, scoped, searchable, and reviewable. The agent can propose what it learned, but you stay in charge of what becomes durable.

Install

Try Local Models on your own machine.

Start local, connect only the providers and channels you want, and keep risky actions approval-gated.

Linux, macOS, or WSL
curl -fsSL https://raw.githubusercontent.com/aaronnat23/disp8ch/main/scripts/install.sh | bash -s -- --repo https://github.com/aaronnat23/disp8ch.git