@gonrocca/zero-pi 0.1.26 → 0.1.28

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -13,294 +13,89 @@
13
13
 
14
14
  **The zero spec-driven development workflow, packaged for [pi](https://pi.dev).**
15
15
 
16
- [![npm](https://img.shields.io/npm/v/@gonrocca/zero-pi?color=af8aff&label=npm)](https://www.npmjs.com/package/@gonrocca/zero-pi)
17
- [![repo](https://img.shields.io/badge/repo-github-7497ff?logo=github&logoColor=white)](https://github.com/gonzalonicolasr/zero-pi)
18
- [![license](https://img.shields.io/npm/l/@gonrocca/zero-pi?color=eebe5c)](./LICENSE)
19
- [![node](https://img.shields.io/node/v/@gonrocca/zero-pi?color=4fddab&label=node)](https://nodejs.org)
20
-
21
- **Repositorio:** [github.com/gonzalonicolasr/zero-pi](https://github.com/gonzalonicolasr/zero-pi)
16
+ [![npm](https://img.shields.io/npm/v/@gonrocca/zero-pi?color=af8aff&label=npm)](https://www.npmjs.com/package/@gonrocca/zero-pi) [![repo](https://img.shields.io/badge/repo-github-7497ff?logo=github&logoColor=white)](https://github.com/gonzalonicolasr/zero-pi) [![license](https://img.shields.io/npm/l/@gonrocca/zero-pi?color=eebe5c)](./LICENSE) [![node](https://img.shields.io/node/v/@gonrocca/zero-pi?color=4fddab&label=node)](https://nodejs.org)
22
17
 
23
18
  </div>
24
19
 
25
20
  ---
26
21
 
27
- zero-pi adds the **`/forge`** SDD pipeline, skill auto-learning, adaptive
28
- per-phase models, a metered-billing guard, and an animated `ZERO` banner —
29
- **without modifying pi itself**. Same idea as `gentle-pi`: pi stays untouched;
30
- zero-pi is a package pi loads. Every prompt, skill, extension and theme below
31
- ships inside this one package.
32
-
33
- ## Contents
34
-
35
- - [Install](#-install)
36
- - [The SDD workflow](#-the-sdd-workflow)
37
- - [Quality-of-life extensions](#-quality-of-life-extensions)
38
- - [Commands](#-commands)
39
- - [Environment variables](#-environment-variables)
40
- - [Files it reads & writes](#-files-it-reads--writes)
41
- - [Relationship to `zero`](#-relationship-to-zero)
42
- - [Development](#-development)
43
-
44
- ---
22
+ zero-pi is an installable **[pi](https://pi.dev)** package that adds a
23
+ disciplined spec-driven development pipeline plus a handful of quality-of-life
24
+ extensions — **without modifying pi**. Same idea as `gentle-pi`: pi stays
25
+ untouched; zero-pi is a package it loads.
45
26
 
46
27
  ## 📦 Install
47
28
 
48
29
  ```
49
30
  pi install npm:@gonrocca/zero-pi
31
+ pi install npm:pi-subagents # required — the pipeline delegates to sub-agents
50
32
  ```
51
33
 
52
- That registers zero-pi in `~/.pi/agent/settings.json` and makes its prompts,
53
- skills, theme, and extensions available in every pi session.
54
-
55
- | Requirement | Detail |
56
- | ----------- | ------ |
57
- | `pi-subagents` | **Required** — the SDD pipeline delegates each phase to a sub-agent. Install with `pi install npm:pi-subagents`. |
58
- | Node | ≥ 20.6 — pi loads the TypeScript extensions directly. |
59
- | After upgrade | Restart pi for the new version to take effect. |
60
-
61
- Remove it with `pi remove npm:@gonrocca/zero-pi`.
62
-
63
- ---
34
+ Needs Node 20.6. Restart pi after an upgrade.
64
35
 
65
- ## 🛠 The SDD workflow
36
+ ## 🛠 `/forge` — the SDD pipeline
66
37
 
67
- ### `/forge` the four-phase pipeline
68
-
69
- A spec-driven development run, driven through four phases in order:
38
+ The core of zero-pi. Run **`/forge <feature>`** and the orchestrator drives the
39
+ work through four phases, each delegated to its own sub-agent:
70
40
 
71
41
  | Phase | Does |
72
42
  | ----- | ---- |
73
43
  | **explore** | Investigate the codebase read-only; produce findings. |
74
44
  | **plan** | Write requirements, design, and an ordered task list. |
75
45
  | **build** | Implement the plan. |
76
- | **veredicto** | Review the build adversarially and record a verdict. |
77
-
78
- Start it with `/forge <feature>`. The orchestrator drives phase order and
79
- enforces a hard build/veredicto iteration cap — a `corregir` verdict re-runs
80
- `build`, a `replantear` re-runs `plan`, and when the cap is reached without a
81
- `pasa` the run stops and reports the result as **not verified**.
82
- `/forge --continue [slug]` resumes an interrupted run from its `.sdd/<slug>/`
83
- artifacts.
84
-
85
- A run can also start from natural language: describe the work and signal SDD
86
- intent — "hacelo con sdd", "usá el pipeline" — and the `sdd-routing` skill
87
- routes it into `/forge`. It triggers only on a clear signal phrase; `/forge`
88
- stays the primary, explicit entry point.
89
-
90
- ### Language & output
91
-
92
- A `/forge` run reads as a short, calm progress stream, in Spanish:
93
-
94
- - **Language Boundary** — every user-facing message is in Spanish (Rioplatense
95
- voseo); sub-agent briefs stay in English for token efficiency; identifiers
96
- (verdicts, slugs, paths, model ids, commands) are kept verbatim.
97
- - **Output Contract** — each phase reports a bounded summary
98
- (`Estado` / `Resumen` / `Artefactos` / `Siguiente`), never free-form prose.
99
- No raw tool output, file dumps, sub-agent listings, or triple-backtick code
100
- fences reach the chat. The phase-start line names the model and provider the
101
- phase runs on plus a brief gloss of what it does — so a slow phase reads as
102
- working, not frozen. The run ends stating `verificado` or `no verificado`.
103
-
104
- ### Review Workload Forecast
105
-
106
- The `plan` phase keeps tasks reviewable. Every planned task carries a
107
- `review: ~N changed lines` estimate, and `tasks.md` gains a `## Review Workload`
108
- section with a bold run total. Tasks are sized against a fixed budget of
109
- **400 changed lines per task** (borrowed from gentle-ai). A task over budget is
110
- split; one that genuinely cannot be split stays whole and is recorded as an
111
- over-budget exception with a reason.
112
-
113
- ### SDD sub-agents — `sdd-agents.ts`
114
-
115
- `pi-subagents` discovers agents from `~/.pi/agent/agents/**/*.md`, but a
116
- `pi install` ships only the phase *prompts*. This extension closes the gap: on
117
- every load it generates the four `zero-<phase>` agent files under
118
- `~/.pi/agent/agents/zero/` from the package's phase prompts and the per-phase
119
- models in `~/.pi/zero.json`, so they stay in sync with the prompts and with
120
- `/zero-models`.
121
-
122
- ### Per-phase models — `/zero-models`
123
-
124
- A real pi command — a code handler, not an LLM prompt — for the SDD models. Run
125
- it with no argument for the **provider-aware** interactive picker
126
- (**phase → provider → model**, sourced from pi's model registry), or set one
127
- directly:
128
-
129
- ```
130
- /zero-models interactive picker
131
- /zero-models build=claude-opus-4-7 set one phase
132
- /zero-models build=codex/gpt-5-codex set one phase with an explicit provider
133
- ```
134
-
135
- It reads and writes `~/.pi/zero.json` — a `models` map and a parallel
136
- `providers` map. The orchestrator picks the change up on the next `/forge` run.
137
-
138
- ### Run memory — Cortex
139
-
140
- Every run reads from and writes to Cortex, the persistent-memory MCP server.
141
- Before exploring, the orchestrator recalls prior `zero-run/*` traces; when the
142
- run ends it saves a run-trace — verdict, correction rounds, gotchas — under
143
- `topic_key: zero-run/<slug>`. The next run starts from what the last one
144
- learned. With `--no-mcp`, or when Cortex is unreachable, the loop degrades
145
- silently and never blocks a run.
146
-
147
- ### Canonical specs & `/zero-sync`
148
-
149
- zero keeps a **canonical, project-wide spec store** that accumulates accepted
150
- requirements across runs, so each `/forge` run builds on the last.
151
-
152
- - **The store** — `.sdd/specs/requirements.md`: a `# ` title followed by
153
- `### REQ: <name>` blocks. The `plan` phase reads it as the baseline.
154
- - **The plan artifacts** — every run's `plan` phase writes `proposal.md`,
155
- `spec.md` (the **delta**: `## ADDED` / `## MODIFIED` / `## REMOVED`),
156
- `design.md`, and `tasks.md` into `.sdd/<slug>/`.
157
- - **`/zero-sync <slug>`** — a deterministic, unit-tested merge that folds the
158
- delta into the store atomically. Guardrails reject a bad delta before
159
- anything is written. The orchestrator invokes it automatically after a `pasa`
160
- verdict — never on `corregir`, `replantear`, or a cap-reached run.
161
- - **The archive** — `.sdd/archive/<YYYY-MM-DD>-<slug>/`: an append-only audit
162
- trail of every sync.
163
-
164
- ### Adaptive model profiles — autotune
165
-
166
- zero learns which model fits each SDD phase from your own run history.
167
-
168
- - **The metrics log** — `~/.pi/zero-runs.jsonl`: every completed run appends one
169
- JSON line (slug, per-phase models, verdict, round count, per-round verdict
170
- sequence). Append-only.
171
- - **Phase attribution** — a `corregir` round blames `build`, a `replantear`
172
- blames `plan`. Autotune upgrades **only the phase at fault**, one tier at a
173
- time. `explore` and `veredicto` are never tuned.
174
- - **Cross-machine sync** — each run pushes its metrics to Cortex; `/forge`
175
- pulls the shared log back, so autotune sees runs from other machines too.
46
+ | **veredicto** | Review it adversarially and record a verdict. |
176
47
 
177
- The `autotune` mode in `~/.pi/zero.json` controls what happens:
48
+ The verdict is `pasa` (done), `corregir` (re-run build), or `replantear`
49
+ (re-run plan). A hard iteration cap bounds the build↔veredicto loop — reached
50
+ without a `pasa`, the run stops and is reported as **not verified**.
51
+ `/forge --continue [slug]` resumes an interrupted run.
178
52
 
179
- | Mode | Behaviour |
180
- | ---- | --------- |
181
- | `auto` _(default)_ | Applies the adjustment to `~/.pi/zero.json` and notifies you what changed. |
182
- | `ask` | Records the recommendation; you apply it from `/zero-models`. |
183
- | `off` | Records run metrics, but never changes or recommends anything. |
53
+ The run speaks **Spanish**, in a bounded, low-noise format — one short summary
54
+ per phase, naming the model that phase runs on, no raw tool output. Or just
55
+ describe the work and say "hacelo con sdd": the `sdd-routing` skill routes it
56
+ into `/forge` for you.
184
57
 
185
- Set it with `/zero-models autotune=<auto\|ask\|off>`. In `ask` mode a waiting
186
- recommendation shows as a leading `★ aplicar sugerencia` entry in `/zero-models`.
58
+ ## What else it adds
187
59
 
188
- ### Skill auto-learning `skill-loop`
189
-
190
- Gives the agent a closed learning loop — distill a reusable skill from a
191
- substantial task, store it, surface relevant skills on a new task, refine an
192
- existing skill instead of duplicating it — so solutions are reused, not
193
- re-derived.
194
-
195
- ---
196
-
197
- ## ✨ Quality-of-life extensions
198
-
199
- ### Provider guard — `provider-guard.ts`
200
-
201
- The `anthropic` provider runs two ways: a Claude Pro/Max **subscription** login
202
- (OAuth, via `/login` — `pi-claude-oauth-adapter` smooths this path) or an **API
203
- key**, which bills per token from your metered extra-usage pool. Same provider
204
- id, different billing. The guard watches model switches and reads pi's auth mode
205
- (`modelRegistry.isUsingOAuth`): when a model runs on `anthropic` with an API key
206
- it emits a single non-blocking warning suggesting `/login`. On OAuth, or on any
207
- other provider, it stays silent.
208
-
209
- ### Startup banner — `startup-banner.ts`
210
-
211
- Renders the `ZERO` wordmark as ASCII-safe Tetris cells in pure ANSI 24-bit
212
- colour, no runtime dependencies. The cells assemble from the bottom up, settle,
213
- then run a short **sparkle pass** — a few cells glint bright for about a second.
214
- The render runs synchronously before pi draws its UI.
215
-
216
- | `ZERO_BANNER` | Effect |
217
- | ------------- | ------ |
218
- | _(unset)_ / `shimmer` | Animated assembly + sparkle, then settle (default) |
219
- | `static` | Completed banner only, no animation |
220
- | `off` | Render nothing |
221
-
222
- Colour is skipped automatically when `NO_COLOR` is set or output is not a TTY.
223
-
224
- ### Working-phrase ticker — `working-phrases.ts`
225
-
226
- Replaces pi's static `Working...` line with a context-aware, rotating Spanish
227
- phrase — tool-specific while a tool runs (`Leyendo archivos…`), the SDD phase
228
- while a sub-agent runs (`Planeando la solución…`), playful verbs while the model
229
- thinks (`Maquinando…`) — plus a theme-tinted braille spinner.
230
-
231
- ### Conversation resume — `conversation-resume.ts`
232
-
233
- When pi exits normally, writes a local handoff note at `.pi/zero-resume.md` —
234
- the exact restore command for the persisted session plus a concise conversation
235
- tail. Generate it any time with `/zero-resume`. zero-pi creates `.pi/.gitignore`
236
- so conversation context is never committed. Set `ZERO_RESUME=off` to disable the
237
- automatic write.
238
-
239
- ### Windows process-tree kill — `win-tree-kill.ts`
240
-
241
- On Windows, `kill()` terminates only the target process — so aborting a turn can
242
- leave an orphaned `claude` process streaming (Esc appears to do nothing). This
243
- extension patches `child_process.spawn` so every later subprocess tree-kills via
244
- `taskkill /T /F`. No-op on non-Windows. **Keep it enabled on Windows.**
245
-
246
- ### ZERO terminal theme — `themes/zero-sdd.json`
247
-
248
- A dark, high-contrast pi theme with cyan, amber, mint, rose, and violet accents
249
- tuned for SDD work. Select it from `/settings`, or set `"theme": "zero-sdd"`.
250
-
251
- ---
252
-
253
- ## ⌨️ Commands
254
-
255
- | Command | What it does |
60
+ | Feature | What it does |
256
61
  | ------- | ------------ |
257
- | `/forge <feature>` | Run the four-phase SDD pipeline for a feature. |
258
- | `/forge --continue [slug]` | Resume an interrupted SDD run. |
259
- | `/zero-models [<phase>=[<provider>/]<model>]` | Show or set the per-phase SDD models. |
260
- | `/zero-models autotune=<auto\|ask\|off>` | Set the autotune mode. |
261
- | `/zero-sync <slug>` | Fold a run's delta `spec.md` into the canonical store. |
262
- | `/zero-resume` | Write `.pi/zero-resume.md` now. |
62
+ | **`/zero-models`** | Pick the model + provider for each SDD phase — interactive or direct. |
63
+ | **Autotune** | Learns which model fits each phase from your run history and re-tunes itself. |
64
+ | **`/zero-sync`** | Folds each run's spec delta into a canonical, project-wide spec store. |
65
+ | **Run memory** | Every run recalls and saves traces to Cortex, so runs learn from each other. |
66
+ | **Provider guard** | Warns when the `anthropic` provider runs on a metered API key instead of your subscription. |
67
+ | **Working-phrase ticker** | Swaps pi's `Working...` for a context-aware Spanish phrase + spinner. |
68
+ | **Conversation resume** | Writes `.pi/zero-resume.md` on exit — the restore command + a conversation tail. |
69
+ | **Windows tree-kill** | Aborting a turn kills the whole process tree — no orphaned `claude`. |
70
+ | **Skill auto-learning** | Distills reusable skills from substantial tasks and surfaces them later. |
71
+ | **`zero-sdd` theme** | A dark, high-contrast pi theme tuned for SDD work. |
263
72
 
264
- ## 🔧 Environment variables
73
+ ## ⌨️ Commands
265
74
 
266
- | Variable | Effect |
267
- | -------- | ------ |
268
- | `ZERO_BANNER` | `shimmer` (default) · `static` · `off` — startup-banner mode. |
269
- | `ZERO_RESUME` | `off` / `0` disables the automatic conversation-resume write. |
270
- | `NO_COLOR` | Standard disables banner colour. |
75
+ | Command | Does |
76
+ | ------- | ---- |
77
+ | `/forge <feature>` | Run the SDD pipeline `--continue [slug]` resumes. |
78
+ | `/zero-models [<phase>=[<provider>/]<model>]` | Show or set per-phase models `autotune=auto\|ask\|off`. |
79
+ | `/zero-sync <slug>` | Fold a run's delta into the canonical spec store. |
80
+ | `/zero-resume` | Write the session handoff note now. |
271
81
 
272
- ## 📂 Files it reads & writes
82
+ ## 🔧 Configuration
273
83
 
274
- | Path | Role |
275
- | ---- | ---- |
276
- | `~/.pi/zero.json` | Per-phase `models` / `providers` and the `autotune` mode. |
277
- | `~/.pi/zero-runs.jsonl` | Append-only run-metrics log autotune learns from. |
278
- | `~/.pi/agent/agents/zero/` | Generated `zero-<phase>` sub-agent files. |
279
- | `.sdd/<slug>/` | Per-run plan artifacts. |
280
- | `.sdd/specs/requirements.md` | The canonical, project-wide spec store. |
281
- | `.sdd/archive/` | Append-only audit trail of every `/zero-sync`. |
282
- | `.pi/zero-resume.md` | Local session handoff note. |
84
+ zero-pi keeps its state in `~/.pi/zero.json` (per-phase models + autotune mode)
85
+ and `~/.pi/zero-runs.jsonl` (the run-metrics log); per-project artifacts live
86
+ under `.sdd/`. Set `ZERO_RESUME=off` to disable the conversation-resume note.
283
87
 
284
88
  ## 🔗 Relationship to `zero`
285
89
 
286
- `zero-pi` is the pi-specific layer of the **zero** integrator. The `zero` CLI
287
- installs this layer onto pi (bootstrapping pi.dev itself when missing) and
288
- writes the per-phase model configuration. You can also install `zero-pi`
289
- directly with `pi install npm:@gonrocca/zero-pi` if you only want the pi layer.
90
+ zero-pi is the pi layer of the **zero** integrator. The `zero` CLI installs it
91
+ onto pi and writes the per-phase model config; you can also install zero-pi on
92
+ its own with the command above.
290
93
 
291
- ## 🧪 Development
94
+ ## Development
292
95
 
293
96
  Dependency-free, no build step — pi loads the TypeScript extensions directly.
294
- Run the test suite with:
97
+ Run the test suite with `npm test`.
295
98
 
296
- ```
297
- npm test
298
- ```
299
-
300
- ---
99
+ ## License
301
100
 
302
- <div align="center">
303
-
304
- MIT © Gonzalo Rocca
305
-
306
- </div>
101
+ MIT © Gonzalo Rocca · [github.com/gonzalonicolasr/zero-pi](https://github.com/gonzalonicolasr/zero-pi)
@@ -15,8 +15,8 @@
15
15
  //
16
16
  // All decisions live in `autotune.ts`; this file only reads files, calls those
17
17
  // functions, and applies/notifies. The whole handler is wrapped in a swallowing
18
- // `try/catch` — exactly like `startup-banner.ts`, a failure must never break a
19
- // pi session. The package stays dependency-free: `node:fs`/`node:os`/`node:path`
18
+ // `try/catch` — a failure must never break a pi session. The package stays
19
+ // dependency-free: `node:fs`/`node:os`/`node:path`
20
20
  // only, plus minimal local interfaces for the pi API.
21
21
 
22
22
  import { readFileSync, writeFileSync } from "node:fs";
package/package.json CHANGED
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "name": "@gonrocca/zero-pi",
3
- "version": "0.1.26",
4
- "description": "zero-pi — an installable layer for pi (pi.dev): the zero spec-driven development workflow, skill auto-learning, and an animated ZERO startup banner. Adds capability to pi without modifying pi.",
3
+ "version": "0.1.28",
4
+ "description": "zero-pi — an installable layer for pi (pi.dev): the zero spec-driven development workflow, per-phase model autotune, and skill auto-learning. Adds capability to pi without modifying pi.",
5
5
  "type": "module",
6
6
  "keywords": [
7
7
  "pi",
@@ -23,7 +23,6 @@
23
23
  "./themes"
24
24
  ],
25
25
  "extensions": [
26
- "./extensions/startup-banner.ts",
27
26
  "./extensions/working-phrases.ts",
28
27
  "./extensions/win-tree-kill.ts",
29
28
  "./extensions/sdd-agents.ts",
@@ -38,7 +37,6 @@
38
37
  "prompts",
39
38
  "skills",
40
39
  "themes",
41
- "extensions/startup-banner.ts",
42
40
  "extensions/working-phrases.ts",
43
41
  "extensions/win-tree-kill.ts",
44
42
  "extensions/sdd-agents.ts",
@@ -1,280 +0,0 @@
1
- // zero-pi - animated startup banner.
2
- //
3
- // A pi extension that renders the ZERO wordmark as ASCII-safe Tetris cells.
4
- // The animated mode progressively assembles the wordmark from the bottom up,
5
- // then settles into the complete banner. The extension stays dependency-free
6
- // and keeps the historical ZERO_BANNER controls.
7
-
8
- type RGB = [number, number, number];
9
-
10
- const COLORS: Record<string, RGB> = {
11
- cyan: [80, 210, 255],
12
- blue: [116, 151, 255],
13
- mint: [79, 221, 171],
14
- amber: [238, 190, 92],
15
- rose: [255, 106, 122],
16
- violet: [175, 138, 255],
17
- peak: [255, 245, 218],
18
- };
19
-
20
- /** Number of rows in every glyph. */
21
- const ROWS = 6;
22
-
23
- // Five-cell-wide glyphs. Non-space characters become visible Tetris cells.
24
- const FONT: Record<string, string[]> = {
25
- Z: ["ZZZZZ", " ZZ", " ZZ ", " ZZ ", "ZZ ", "ZZZZZ"],
26
- E: ["EEEEE", "EE ", "EEEE ", "EE ", "EE ", "EEEEE"],
27
- R: ["RRRR ", "RR RR", "RRRR ", "RR RR", "RR R", "RR R"],
28
- O: [" OOO ", "OO OO", "OO OO", "OO OO", "OO OO", " OOO "],
29
- P: ["PPPP ", "PP PP", "PP PP", "PPPP ", "PP ", "PP "],
30
- I: ["IIIII", " II ", " II ", " II ", " II ", "IIIII"],
31
- "-": [" ", " ", "-----", " ", " ", " "],
32
- " ": [" ", " ", " ", " ", " ", " "],
33
- };
34
-
35
- const PIECE_COLORS: Record<string, RGB> = {
36
- Z: COLORS.cyan,
37
- E: COLORS.amber,
38
- R: COLORS.mint,
39
- O: COLORS.rose,
40
- P: COLORS.violet,
41
- I: COLORS.blue,
42
- "-": COLORS.amber,
43
- };
44
-
45
- interface Matrix {
46
- rows: string[];
47
- width: number;
48
- totalCells: number;
49
- ranks: Map<string, number>;
50
- }
51
-
52
- function glyphFor(ch: string): string[] {
53
- return FONT[ch.toUpperCase()] ?? FONT[" "];
54
- }
55
-
56
- function matrixForText(text: string): Matrix {
57
- const rows = Array.from({ length: ROWS }, () => "");
58
- for (const [index, raw] of Array.from(text).entries()) {
59
- const glyph = glyphFor(raw);
60
- for (let r = 0; r < ROWS; r++) {
61
- rows[r] += (index === 0 ? "" : " ") + glyph[r];
62
- }
63
- }
64
-
65
- const cells: Array<{ row: number; col: number }> = [];
66
- for (let row = 0; row < ROWS; row++) {
67
- for (let col = 0; col < rows[row].length; col++) {
68
- if (rows[row][col] !== " ") cells.push({ row, col });
69
- }
70
- }
71
-
72
- // Tetris reads as a pile building upward: bottom rows settle first.
73
- cells.sort((a, b) => b.row - a.row || a.col - b.col);
74
-
75
- const ranks = new Map<string, number>();
76
- for (const [rank, cell] of cells.entries()) {
77
- ranks.set(`${cell.row}:${cell.col}`, rank);
78
- }
79
-
80
- return {
81
- rows,
82
- width: rows[0]?.length ?? 0,
83
- totalCells: cells.length,
84
- ranks,
85
- };
86
- }
87
-
88
- /** Lay a word out as six rows of ASCII-safe Tetris cells. */
89
- export function bannerLines(text = "ZERO"): string[] {
90
- return matrixForText(text).rows.map((row) =>
91
- Array.from(row)
92
- .map((cell) => (cell === " " ? " " : "[]"))
93
- .join(""),
94
- );
95
- }
96
-
97
- function lerp(a: number, b: number, t: number): number {
98
- return a + (b - a) * t;
99
- }
100
-
101
- function lerpColor(c1: RGB, c2: RGB, t: number): RGB {
102
- const k = Math.max(0, Math.min(1, t));
103
- return [
104
- Math.round(lerp(c1[0], c2[0], k)),
105
- Math.round(lerp(c1[1], c2[1], k)),
106
- Math.round(lerp(c1[2], c2[2], k)),
107
- ];
108
- }
109
-
110
- function ansiFg([r, g, b]: RGB, text: string): string {
111
- return `\x1b[38;2;${r};${g};${b}m${text}\x1b[0m`;
112
- }
113
-
114
- function colorForPiece(piece: string, row: number, rank: number, active: boolean): RGB {
115
- if (active) return COLORS.peak;
116
- const base = PIECE_COLORS[piece] ?? COLORS.blue;
117
- const heightWarmth = (ROWS - 1 - row) / Math.max(1, ROWS - 1);
118
- const rankPulse = (rank % 7) / 18;
119
- return lerpColor(base, COLORS.peak, heightWarmth * 0.18 + rankPulse);
120
- }
121
-
122
- function paintMatrixLine(
123
- matrix: Matrix,
124
- row: number,
125
- visibleCells: number,
126
- sparkles?: ReadonlySet<string>,
127
- ): string {
128
- let out = "";
129
- const line = matrix.rows[row] ?? "";
130
- for (let col = 0; col < matrix.width; col++) {
131
- const piece = line[col] ?? " ";
132
- if (piece === " ") {
133
- out += " ";
134
- continue;
135
- }
136
- const key = `${row}:${col}`;
137
- const rank = matrix.ranks.get(key) ?? Number.POSITIVE_INFINITY;
138
- if (rank >= visibleCells) {
139
- out += " ";
140
- continue;
141
- }
142
- // A cell glints when it is the assembly's leading cell or when this
143
- // sparkle frame picked it.
144
- const glint = rank === visibleCells - 1 || (sparkles?.has(key) ?? false);
145
- out += ansiFg(colorForPiece(piece, row, rank, glint), "[]");
146
- }
147
- return out;
148
- }
149
-
150
- /** Pick a handful of distinct settled cells to glint for one sparkle frame. */
151
- function pickSparkles(matrix: Matrix, count: number, rand: () => number): Set<string> {
152
- const keys = Array.from(matrix.ranks.keys());
153
- const chosen = new Set<string>();
154
- const target = Math.min(Math.max(0, count), keys.length);
155
- for (let guard = 0; chosen.size < target && guard < target * 8; guard++) {
156
- chosen.add(keys[Math.floor(rand() * keys.length)]);
157
- }
158
- return chosen;
159
- }
160
-
161
- interface OutStream {
162
- write(chunk: string): unknown;
163
- isTTY?: boolean;
164
- }
165
-
166
- /** Whether colour output is appropriate for the given stream. */
167
- function shouldColor(stream: OutStream): boolean {
168
- if (process.env.NO_COLOR) return false;
169
- if (process.env.FORCE_COLOR) return true;
170
- return Boolean(stream && stream.isTTY);
171
- }
172
-
173
- /** The three banner modes, resolved from the ZERO_BANNER environment variable. */
174
- export type BannerMode = "off" | "static" | "shimmer";
175
-
176
- /** Resolve the banner mode from an environment-variable value. */
177
- export function resolveBannerMode(raw: string | undefined): BannerMode {
178
- const value = (raw ?? "").trim().toLowerCase();
179
- if (value === "off" || value === "0" || value === "false" || value === "none") return "off";
180
- if (value === "static" || value === "plain" || value === "no-animate") return "static";
181
- return "shimmer";
182
- }
183
-
184
- // Atomics.wait gives us a synchronous sleep without adding a dependency. The
185
- // banner renders before Pi's TUI takes over, so blocking briefly is intentional.
186
- const SLEEP_SLOT = new Int32Array(new SharedArrayBuffer(4));
187
- function sleepSync(ms: number): void {
188
- if (ms > 0) Atomics.wait(SLEEP_SLOT, 0, 0, ms);
189
- }
190
-
191
- function easeOutCubic(t: number): number {
192
- return 1 - Math.pow(1 - t, 3);
193
- }
194
-
195
- /** Options for {@link renderBanner}. */
196
- export interface RenderOptions {
197
- mode?: BannerMode;
198
- text?: string;
199
- frames?: number;
200
- frameMs?: number;
201
- /** Number of post-settle sparkle frames (shimmer mode). `0` disables it. */
202
- sparkleFrames?: number;
203
- /** Delay between sparkle frames, in milliseconds. */
204
- sparkleMs?: number;
205
- stream?: OutStream;
206
- }
207
-
208
- /**
209
- * Render the ZERO banner synchronously.
210
- *
211
- * On a non-colour stream the banner is printed plain. Otherwise the animated
212
- * mode assembles cells frame by frame, then settles on the completed wordmark.
213
- */
214
- export function renderBanner(options: RenderOptions = {}): void {
215
- const mode = options.mode ?? resolveBannerMode(process.env.ZERO_BANNER);
216
- if (mode === "off") return;
217
-
218
- const stream: OutStream = options.stream ?? process.stdout;
219
- const matrix = matrixForText(options.text ?? "ZERO");
220
-
221
- if (!shouldColor(stream)) {
222
- stream.write("\n" + bannerLines(options.text ?? "ZERO").join("\n") + "\n\n");
223
- return;
224
- }
225
-
226
- stream.write("\n" + Array.from({ length: ROWS }, () => "").join("\n") + "\n");
227
-
228
- // Repaint the six rows in place; optionally with this frame's sparkle set.
229
- const paintRows = (sparkles?: ReadonlySet<string>): void => {
230
- stream.write(`\x1b[${ROWS}A\r`);
231
- for (let r = 0; r < ROWS; r++) {
232
- stream.write("\x1b[2K" + paintMatrixLine(matrix, r, matrix.totalCells, sparkles) + "\n");
233
- }
234
- };
235
-
236
- if (mode === "shimmer") {
237
- const frames = Math.max(2, options.frames ?? 42);
238
- const frameMs = Math.max(1, options.frameMs ?? 18);
239
- for (let f = 0; f < frames; f++) {
240
- const t = easeOutCubic(f / (frames - 1));
241
- const visibleCells = Math.max(1, Math.ceil(matrix.totalCells * t));
242
- stream.write(`\x1b[${ROWS}A\r`);
243
- for (let r = 0; r < ROWS; r++) {
244
- stream.write("\x1b[2K" + paintMatrixLine(matrix, r, visibleCells) + "\n");
245
- }
246
- sleepSync(frameMs);
247
- }
248
- }
249
-
250
- // Settle on the completed wordmark.
251
- paintRows();
252
-
253
- // Sparkle pass — a few settled cells glint bright each frame, then a clean
254
- // final settle. Shimmer only; `sparkleFrames: 0` disables it.
255
- if (mode === "shimmer") {
256
- const sparkleFrames = Math.max(0, options.sparkleFrames ?? 22);
257
- const sparkleMs = Math.max(1, options.sparkleMs ?? 36);
258
- for (let f = 0; f < sparkleFrames; f++) {
259
- paintRows(pickSparkles(matrix, 3 + (f % 4), Math.random));
260
- sleepSync(sparkleMs);
261
- }
262
- if (sparkleFrames > 0) paintRows();
263
- }
264
-
265
- stream.write("\n");
266
- }
267
-
268
- /**
269
- * The pi extension entry point.
270
- *
271
- * The pi argument is accepted for API compatibility. The banner is rendered
272
- * defensively: visual sugar should never break an interactive session.
273
- */
274
- export default function register(_pi?: unknown): void {
275
- try {
276
- renderBanner();
277
- } catch {
278
- // A banner failure must never break a pi session.
279
- }
280
- }