Noema: Local AI & Offline LLM

• Python code execution: models can now write and run real Python on your device, entirely offline. Ask for a calculation, a statistical summary, or a reshaping of your own data, and Noema runs the code and shows you the result next to it. • A scientific toolkit is bundled inside the app — NumPy, pandas, SciPy, scikit-learn, SymPy, Matplotlib, and more — so data analysis, statistics, symbolic math, and machine learning work with no network and nothing to install. The packages ship with Noema and cannot be added at runtime. • Charts arrive in the conversation. Matplotlib figures are drawn inline, and a separate Quick Charts tool renders native bar, line, scatter, and pie charts from data a model has already prepared, without running any Python. • Tap a chart to open it full size, zoom in, and share or save it. • Execution is deliberately confined: no network access, no subprocesses, no loading of outside libraries, a temporary working directory it cannot leave, and a 30-second limit on every run. • Turn Python on in Settings under Tools, then arm it from the + menu in a chat. A new Python page lists the runtime version, every bundled package with its version and dependencies, and the health of the package set. • Code blocks inside a model's reasoning keep their formatting instead of being flattened, and images in tool results are sized correctly in the viewer. • Short conversations no longer jump while a reply is streaming in, and the first suggested message hands off to the chat more smoothly. • Interface fixes, bug fixes, and polish throughout.