Common setup questions (FAQ)

[scottconverse]

Collecting common setup questions here. Post your own below and I'll add answers.

---

Q: Ollama screen in the installer stays stuck — Continue never enables.

Ollama isn't running or the model isn't loaded. Open a terminal:

ollama serve
ollama pull gemma4:e4b

Then refresh the installer page. The screen polls /api/ollama/status every few seconds — it will unlock once the model is reported as loaded.

---

Q: Run immediately shows an error.

AgentSuite (the underlying engine) isn't installed or the Python path is wrong. Make sure you ran pip install -e . from the repo root with the same virtualenv that's running uvicorn. Check pip show agentsuite — it should resolve.

---

Q: Token speed is really slow (< 5 tok/s).

Your machine is likely swapping. Switch to the Light tier model (gemma4:e2b) which uses ~3 GB less RAM. Go to Settings → Model to switch without reinstalling.

---

Q: Artifacts are very short or cut off mid-sentence.

The model's context window was exhausted during Spec or Execute. Try the Balanced or Pro tier model — larger context window.

---

Q: Where are the artifacts on disk?

~/AgentSuite/.agentsuite/runs/{run-id}/ — the run ID is shown in the Runs view. Promoted (approved) artifacts are at ~/AgentSuite/.agentsuite/_kernel/{project}/{agent}/.

---

Q: Can I use a different model than the ones listed?

Not through the UI in v0.1, but the backend honours whatever Ollama reports. Pull any model with ollama pull <name>, update AGENTSUITE_MODEL env var (if AgentSuite respects it), and restart the backend. The model picker in the installer is cosmetic — it sets a preference that AgentSuite reads at run time.

---

Post your question below.