v1.1.3¶
Release date: 2026-03-18
Overview¶
This patch release hardens the final mile of the local Mistral/Ollama workflow. The SDK now validates a real tiny generation during local-check, retries once on short disconnects, and gives much clearer guidance when a local model is too heavy for the current machine state.
Highlights¶
iints ai local-checknow performs a tiny generation smoke-test by default- Local Ollama generation retries once before failing on transient disconnects
- Generation disconnects now surface explicit recovery guidance:
- confirm direct inference with
ollama run ... - rerun
iints ai local-check - switch to
ministral-3:3bon tighter hardware - Documentation now explains the practical fallback path when
ministral-3:8bis installed but unstable during inference
Recommended Verification Flow¶
python -m pip install -U "iints-sdk-python35[mdmp]==1.1.3"
ollama pull ministral-3:3b
iints ai local-check --model ministral-3:3b
iints ai report results/<run_id> --model ministral-3:3b
Stability Notes¶
- Research use only
- Not a medical device
- No clinical dosing advice
- AI output remains blocked unless MDMP verification succeeds and the minimum grade requirement is met