Evidence record / pi-coding-agent

A dated record of one change, kept so the writing that cites it can be checked. Compiled from the sources listed below by the research run, not written for reading. The judgment lives in the issues that cite it, below.

2026-10-13-pi-coding-agent-headless-runs-fail-loudly-where-they-used-to-pass-and-the-tui-went-fullscreen-by-default

Headless runs fail loudly where they used to pass, and the TUI went fullscreen by default. Three silent passes became failures. In text print mode (-p), a final answer cut off at the model’s output limit used to exit 0. From 1.2.0 it prints the partial text, warns on stderr and exits 1; this branch covers text mode only. A bad --mode was silently ignored until 0.87.1. --provider without --model quietly ran another provider’s default model until 1.0.0. Interactively, 1.0.0 made the fullscreen TUI the default. tuiMode: "regular" or --tui-mode regular restores normal scrollback, and 1.0.3 moved transcript top and bottom to Ctrl+Home/Ctrl+End.

Channel: tagged-release (v0.87.1, v1.0.0, v1.2.0). Half: both. Date: 2026-10-12.

Operator consequence: CI jobs that call pi -p can start failing on 1.2.0 where answers were being truncated. Those failures are real truncated answers, so raise the output limit or split the task rather than masking the exit code. Scripts that passed --provider alone now error instead of running the wrong model, which may explain an earlier cost surprise. If you drive Pi inside tmux or rely on scrollback capture, set tuiMode: "regular".

Receipt

Finding metadata

Run: 2026-10-13-profile-refresh-2026-09-21_2026-10-13-frontier-v0

Finding ID: 2026-10-13-pi-coding-agent-headless-runs-fail-loudly-where-they-used-to-pass-and-the-tui-went-fullscreen-by-default

Profile citations

Source links

Primary links, including exact changelog lines when available.

Versioned source: run artifact