Skip to content
PidocsCLI & tooling3 files · +152 −106

Update pi-mono README and ppi CLI flags for Docker agent wrapper

Session from combust-labs_pi-mono-docker on Hugging Face by rgruchalski · Apache-2.0 · condensed by Coders Talk

18msession
0interventions
GOAL

Improve and correct the README documentation for the pi-mono Docker project, then make targeted fixes to the ppi wrapper script (model requirement, validation, new flag) while keeping docs in sync.

  1. Prompt · +0
    Summarize current the project
  2. Prompt · +0m
    Detailed list of README.md changes: fix wording, document build args, update contents, multiline run command, segregate pi vs ppi flags, fix flag tables, add example caveat, alphabetize flags
  3. Agent did · +0m
    Implemented all 8 requested README edits one by one: reworded intro, documented build args, updated contents list, made run command multiline, split pi vs ppi flags into separate tables, moved --ppi-host-add-path, added example caveat, verified alphabetical ordering
  4. Agent did · +7m
    Committed README changes with a summary commit, then added missing SPDX header to pi-run.sh and committed again; verified branch clean after each commit
  5. Prompt · +9m
    Remove default model value, model is explicitly required. Ensure the readme is up-to-date, but don't touch the bash function example.
  6. Agent did · +9m
    Removed default model value from ppi script, updated README --model flag to show (required), updated basic usage example, committed changes
  7. Prompt · +12m
    does the ppi program fail gracefully if no model is given?
  8. Agent did · +12m
    Found no validation existed for empty model; added a check after argument parsing that exits with a clear error message if --model is missing, then committed
  9. Prompt · +14m
    Add a boolean --ppi-no-ppi-prompts flag, default false, to skip prepending the default nickname prompt; keep readme up-to-date
  10. Agent did · +14m
    Added new variable and flag case for --ppi-no-ppi-prompts, made nickname injection conditional, updated README flag table and nickname section, then committed with meaningful message
  11. Outcome · +18m
    5 commits ahead of origin/main, working tree clean throughout the session
VERDICT

Next time I'd batch all the README wording fixes and flag reorganization into a single detailed spec up front, since that's basically what worked well here. The follow-up question about graceful failure on missing --model was the most valuable correction — it caught a real gap the agent hadn't considered when it removed the default value, and led to proper validation being added before anything shipped.

I tried this · 0

Discussion · 0

How Builds work