Skip to content

RFC: --quick mode with static renderer + fallback [DO NOT MERGE]#12

Draft
mikeyobrien wants to merge 1 commit intonicobailon:mainfrom
mikeyobrien:feat/quick-mode-static-renderer
Draft

RFC: --quick mode with static renderer + fallback [DO NOT MERGE]#12
mikeyobrien wants to merge 1 commit intonicobailon:mainfrom
mikeyobrien:feat/quick-mode-static-renderer

Conversation

@mikeyobrien
Copy link

@mikeyobrien mikeyobrien commented Feb 26, 2026

Summary

image

Adds an opt-in --quick mode to visual-explainer while preserving the original mechanism as default.

  • Default behavior unchanged: no flag = full one-shot HTML generation.
  • Quick behavior: --quick routes through JSON spec + static renderer.
  • Safety: if quick mode fails or request doesn’t fit quick schema, prompts explicitly fall back to full workflow.

What changed

  • Added quick renderer assets:
    • quick/render-architecture.mjs
    • quick/base.css
    • quick/README.md
  • Documented quick mode + fallback contract in:
    • SKILL.md
    • README.md
  • Wired --quick instructions into prompts:
    • prompts/generate-web-diagram.md
    • prompts/diff-review.md
    • prompts/plan-review.md
    • prompts/project-recap.md
  • Added overflow/cutoff hardening in quick CSS (pipeline grid + wrap protections).

Benchmark data

Measured on 5-cycle runs:

  • Baseline (full HTML by model): 165.17s avg
  • Quick v1: 41.09s avg (-75.1%)
  • Quick v2 (rich quick output): 31.42s avg (-81.0%)

Token/output reduction vs baseline:

  • Quick v2 estimated output tokens: 1295 vs 7809 baseline (-83.4%)

Parity results (v2):

  • Functional quality rubric: 11.0 -> 12.0
  • Semantic coverage: 7.0 -> 7.0
  • Richness rubric: 2.6 -> 11.0
  • Spec→HTML fidelity coverage: 99.546%
  • Overall verdict: PASS++++

E2E validation

Verified end-to-end quick mode generation for:

  • /generate-web-diagram --quick ...
  • /diff-review --quick ...
  • /plan-review --quick ...
  • /project-recap --quick ...

Also verified full mode remains active when --quick is omitted.

Notes

  • Quick mode currently uses architecture-schema rendering (intentional condensed format).
  • Full mode remains available for maximal, fully custom layouts.

- add quick static renderer pipeline (base.css + render-architecture.mjs + schema docs)
- document quick mode in SKILL and README while preserving default full-html behavior
- wire --quick routing + fallback instructions into generate-web-diagram, diff-review, plan-review, and project-recap prompts
- improve quick CSS overflow handling to reduce clipped content on dense outputs
Copy link

@chatgpt-codex-connector chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 407373e35b

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

2. Read `./quick/README.md` for the JSON schema and optional rich fields.
3. Generate a **JSON spec** (not HTML) for the architecture renderer and save it to `/tmp/ve-quick-<timestamp>.json`.
4. Render HTML using:
- `node $HOME/.pi/agent/skills/visual-explainer/quick/render-architecture.mjs /tmp/ve-quick-<timestamp>.json ~/.agent/diagrams/<descriptive-name>-quick.html`

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Resolve quick renderer path without Pi-only hard-coding

This quick-mode command is pinned to $HOME/.pi/agent/skills/..., so --quick rendering breaks for non-Pi installs (for example the README’s Claude install path ~/.claude/skills/visual-explainer) and immediately falls back to full generation, effectively disabling the new feature in those environments. Please derive the renderer path from the loaded skill location (or handle agent-specific install roots) instead of using a Pi-specific absolute path; the same pattern is repeated in prompts/diff-review.md, prompts/plan-review.md, and prompts/project-recap.md.

Useful? React with 👍 / 👎.

@mikeyobrien mikeyobrien marked this pull request as draft February 26, 2026 04:53
@mikeyobrien mikeyobrien changed the title feat: add --quick mode with static renderer + fallback RFC: --quick mode with static renderer + fallback [DO NOT MERGE] Feb 26, 2026
@mikeyobrien
Copy link
Author

Marking this as RFC / feedback-only for now — not intended to merge yet.

I’m looking for focused feedback on:

  1. Interface/UX

    • Is --quick the right flag name/shape?
    • Should quick mode be architecture-only for v1?
  2. Safety / fallback

    • Is the quick->full fallback behavior explicit enough?
    • Any edge cases where we should force full mode instead?
  3. Quality bar

    • Is the condensed quick output acceptable for diff-review, plan-review, and project-recap?
    • Any must-have sections missing in quick mode?
  4. Perf evidence

    • Are benchmark + parity metrics sufficient as decision criteria?
    • Any additional measurements you want before we consider merge?

If you can, please comment with Approve direction / Blocker / Nice-to-have so I can triage follow-ups cleanly.

@mikeyobrien
Copy link
Author

Love the visual-explainer skill - only painpoint is how long it takes to generate. The intent here is trade-off the visual aesthetic for faster generation time. Curious what your thoughts are.

@mikeyobrien
Copy link
Author

Created paired tracking issue for structured feedback + acceptance checklist: #13

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant