ollama pull llama3.2:3b
zypheron config set ai.provider ollama
zypheron config set ai.model llama3.2:3b
zypheron ai test --provider ollama
This stays aligned with the repo: choose a local model, set provider and model explicitly, then verify the path before expecting higher-level workflows to behave. That is more reliable than hand-waving about local AI support without showing the actual CLI flow.
Why Local AI Matters
- Lower exposure for prompts, target context, and artifacts.
- Useful for air-gapped or low-connectivity environments.
- Predictable model choices that can be persisted for reuse.
Hosted Providers Still Fit
Local does not replace hosted providers. The repo supports both. The point is choosing the right backend for the workflow instead of forcing one provider model across everything.
If you need MCP after this, the current follow-up path is zypheron mcp config, then zypheron mcp start, then zypheron mcp status.