VOOLDocs

Run a local model#

Install a runtime#

VOOL talks to a local model server. Install one, start it, and pull a model.

bash
ollama serve
ollama pull qwen2.5-coder:7b

Point VOOL at it#

Open Settings → Models. Running local runtimes are detected automatically. Select a model and the mode indicator changes to LOCAL.

Sizing#

Model sizeMemory neededSuited to
3B~4 GBShort edits, quick questions
7–8B~8 GBGeneral use, most coding
14B~16 GBLonger reasoning
32B+32 GB or moreHeavier work, slower

Quantised builds cut memory substantially. A 7B model at 4-bit runs comfortably in about 5 GB.

If it is slow#

  • Choose a smaller or more heavily quantised model.
  • Close other memory-heavy applications.
  • Shorten the context — fewer files in the request means less to process.
VOOL — Parad0x Labs · 2026