Phonton CLI v0.24.0

phonton-cli v0.24.0 — A first local run that reads well

Polish from recording the launch demo on a fresh runner with
qwen2.5-coder:3b. Behavior is unchanged; what you read along the way is
shorter and plainer.

Setup you can read

In a terminal, phonton models setup MODEL now ends with a summary instead
of a JSON record: the runtime, the model, each calibration probe
(search/replace edit, unified diff edit, structured create, tool call
format) as pass or FAIL with its time, and the one edit format Phonton will
ask that model for. The full record is still printed with --json or when
the output is piped.

Download progress rewrites a single line on a terminal instead of printing a
new line every 250 ms, shows a human size beside the exact byte count, and
no longer repeats a layer's digest after Ollama's own pulling label.

A shorter plan review

The TUI's local plan review is six labeled lines (Goal, Repo, Edits, Checks,
Model, Budget) and a few bullet notes. Checks are shown as a command line;
arguments with spaces, quotes or shell characters are quoted, so the
boundaries you approve are the ones that run. The notes were rewritten in
plain language, for example: "These checks were inferred from the repository
and have not run yet. Phonton asks before running them on this machine."

A clearer memory error

When calibration finds no context size that fits in memory, the error now
says which memory is short (or that the model's limit is missing) and what
to do, instead of "no safe automatic context".

Install

npm install -g phonton-cli
phonton version    # phonton 0.24.0
phonton models setup qwen2.5-coder:3b