feat(i18n): add Portuguese instructions (European ptPT + Brazilian ptBR)

CLAUDE.md

@@ -0,0 +1,58 @@
+# exercises-dataset
+
+## Purpose
+
+Fork of `hasaneyldrm/exercises-dataset` (1,324 fitness exercises: JSON data + GIF/thumbnail media + instructions in 11 languages). This fork's job: add **European Portuguese (`pt`)** instructions via the pipeline in `scripts/`, then contribute the result upstream as a PR — mirroring how the French PR (commit `6f3031b`) was structured. Not an app; a data repo with a Python pipeline.
+
+## Stack
+
+- Python 3 (system `python3`, currently 3.9) — stdlib only for scripts 1/3/4; script 2 calls any OpenAI-compatible HTTP endpoint. No package manager, no venv, no Node.
+- Data: `data/exercises.json` (18 MB) validated by `data/exercises.schema.json`.
+- `index.html` (17 MB) embeds a full inline copy of the dataset in one `const EXERCISES = [...]` line.
+
+```
+data/            exercises.json + JSON Schema — the source of truth
+scripts/         PT-PT translation pipeline (see scripts/README.md — read it first)
+  glossary_pt.json           pinned PT-PT fitness terminology (~150 terms)
+  en_strings.json            unique EN strings (generated by step 1)
+  translations*.json, chunks/  intermediate outputs — never hand-edit
+images/ videos/  1,324 thumbnails + GIFs (© Gym visual — licensed media, never modify)
+index.html       interactive browser with inline dataset copy
+```
+
+## Commands
+
+Full pipeline, in order (details + env vars in `scripts/README.md`):
+
+```bash
+python3 scripts/1_extract_unique.py     # dedup EN steps → en_strings.json
+python3 scripts/2_translate.py          # needs OPENAI_API_KEY; resumable
+python3 scripts/3_apply.py              # writes pt into exercises.json, enforces 6 invariants
+python3 scripts/4_sync_index_html.py    # regenerates inline JSON in index.html
+python3 scripts/3_apply.py --check      # verify — this is the test suite
+```
+
+## Conventions
+
+- `scripts/README.md` is the pipeline's authoritative doc — keep it in sync with any script change.
+- The 6 invariants in `3_apply.py` are hard rules (schema-valid; `pt` step count == `en`; `instructions.pt` joins the steps; pre-existing fields byte-identical; no empty strings; `pt` last key after `fr`). Never weaken them to make a run pass.
+- Translation register is impersonal 3rd person, no subject pronoun (`Deite-se`, `Mantenha`, `Repita`) — PT-PT, never PT-BR. Terms in `glossary_pt.json` are pinned: if the English source contains a glossary term, the mapped translation is mandatory.
+- Scripts are stdlib-only, `from __future__ import annotations`, runnable on Python 3.9 — don't add pip dependencies.
+- README.md follows upstream's style (badges, emoji headers) — match it when editing; changes here should stay upstream-mergeable.
+
+## Definition of Done
+
+- [ ] `python3 scripts/3_apply.py --check` passes (all 6 invariants)
+- [ ] If `data/exercises.json` changed, `python3 scripts/4_sync_index_html.py` was re-run (data and index.html must never diverge)
+- [ ] No hand edits to generated files (`en_strings.json`, `translations*.json`, `chunks/`, the inline JSON in `index.html`)
+- [ ] `git status` reviewed — media files (`images/`, `videos/`) untouched
+
+## Gotchas
+
+- **Never read `index.html` or `data/exercises.json` fully** — 17–18 MB each. Use `head`, `rg`, or `jq` for targeted slices; `jq '.[0]'` shows a record shape.
+- The inline `EXERCISES` line in `index.html` is one 16 MB line — only `4_sync_index_html.py` may touch it.
+- `translations_partial_*.json` and `chunks/` are resumable-run intermediates of `2_translate.py`; deleting them loses paid translation progress.
+- `2_translate.py` costs money (LLM API) — never run it unprompted; it's resumable, so partial runs are safe to continue.
+- **Preferred translation path: the assistant translates directly** (chunk by chunk from `scripts/chunks/`, glossary-pinned) — never pull another model (ollama etc.) or call a paid API for work you can do yourself. This was corrected explicitly.
+- Media is © Gym visual with its own license terms (see LICENSE/NOTICE.md) — data is MIT, media is not.
+- Upstream is `hasaneyldrm/exercises-dataset`; `origin` is the `tutods` fork. The end goal of local work is an upstream PR, so keep the diff minimal and PR-shaped (like the French PR `6f3031b`).

README.md

@@ -11,12 +11,12 @@
   <img src="videos/0334-DsgkuIt.gif" width="120" alt="dumbbell lateral raise" />
 </p>
 
-**A comprehensive, ready-to-use fitness exercise dataset with 1,324 exercises — each with an animation GIF, 180×180 thumbnail image, category, body-part, equipment, target and muscle-group data, and step-by-step instructions in 10 languages (English, Spanish, Italian, Turkish, Russian, Chinese, Hindi, Polish, Korean, French).**
+**A comprehensive, ready-to-use fitness exercise dataset with 1,324 exercises — each with an animation GIF, 180×180 thumbnail image, category, body-part, equipment, target and muscle-group data, and step-by-step instructions in 12 languages (English, Spanish, Italian, Turkish, Russian, Chinese, Hindi, Polish, Korean, French, European Portuguese, Brazilian Portuguese).**
 
 [![Exercises](https://img.shields.io/badge/Exercises-1324-blue?style=flat-square)](data/exercises.json)
 [![Animation GIFs](https://img.shields.io/badge/Animation%20GIFs-1324-brightgreen?style=flat-square)](videos/)
 [![Thumbnails](https://img.shields.io/badge/Thumbnails-1324-orange?style=flat-square)](images/)
-[![Languages](https://img.shields.io/badge/Languages-10-green?style=flat-square)](#-overview)
+[![Languages](https://img.shields.io/badge/Languages-12-green?style=flat-square)](#-overview)
 [![Mobile App](https://img.shields.io/badge/App-LogPress-111111?style=flat-square&logo=react)](https://github.com/hasaneyldrm/logpress-public)
 [![License](https://img.shields.io/badge/License-MIT%20%2B%20media%20terms-blue?style=flat-square)](LICENSE)
 
@@ -32,7 +32,7 @@
 
 - 1,324 exercises with category, body-part, equipment, target and muscle-group data
 - an animation GIF + 180×180 thumbnail for every exercise (media © [Gym visual](https://gymvisual.com/) — see [License](#-license--use))
-- step-by-step instructions in 10 languages (🇬🇧 English, 🇪🇸 Spanish, 🇮🇹 Italian, 🇹🇷 Turkish, 🇷🇺 Russian, 🇨🇳 Chinese, 🇮🇳 Hindi, 🇵🇱 Polish, 🇰🇷 Korean, 🇫🇷 French)
+- step-by-step instructions in 12 languages (🇬🇧 English, 🇪🇸 Spanish, 🇮🇹 Italian, 🇹🇷 Turkish, 🇷🇺 Russian, 🇨🇳 Chinese, 🇮🇳 Hindi, 🇵🇱 Polish, 🇰🇷 Korean, 🇫🇷 French, 🇵🇹 European Portuguese, 🇧🇷 Brazilian Portuguese)
 - the interactive browser (`index.html`) and developer setup guide (`setup.html`)
 
 ---
@@ -71,7 +71,7 @@ Each exercise entry contains:
 | Muscle Group | Supporting / synergist muscles |
 | Equipment | Equipment required (or `body weight` for bodyweight) |
 | Instructions | Step-by-step instructions for each exercise |
-| Available Languages | 🇬🇧 English · 🇪🇸 Spanish · 🇮🇹 Italian · 🇹🇷 Turkish · 🇷🇺 Russian · 🇨🇳 Chinese · 🇮🇳 Hindi · 🇵🇱 Polish · 🇰🇷 Korean · 🇫🇷 French |
+| Available Languages | 🇬🇧 English · 🇪🇸 Spanish · 🇮🇹 Italian · 🇹🇷 Turkish · 🇷🇺 Russian · 🇨🇳 Chinese · 🇮🇳 Hindi · 🇵🇱 Polish · 🇰🇷 Korean · 🇫🇷 French · 🇵🇹 European Portuguese · 🇧🇷 Brazilian Portuguese |
 | Media | 180×180 thumbnail (`image`) + animation GIF (`gif_url`) per exercise — media © Gym visual, see [License](#-license--use) |
 
 ---
@@ -88,7 +88,7 @@ A fully client-side exercise explorer with:
 - Live search across all 1,324 exercises
 - Filter by category, equipment, and target muscle
 - Infinite scroll grid
-- Click any card to see full details and instructions in English, Spanish, Italian, Turkish, Russian, Chinese, Hindi, Polish, Korean, or French
+- Click any card to see full details and instructions in English, Spanish, Italian, Turkish, Russian, Chinese, Hindi, Polish, Korean, French, European Portuguese, or Brazilian Portuguese
 
 ### `setup.html` — Developer Setup Guide
 
@@ -131,7 +131,7 @@ exercises-dataset/
 | Metric | Count |
 |---|---|
 | Total Exercises | **1,324** |
-| Instruction Languages | **10** |
+| Instruction Languages | **12** |
 
 ### Exercises by Body Part
 
@@ -190,7 +190,9 @@ Each record in `data/exercises.json` follows this structure. A machine-readable
 | `instructions.pl` | `string` | Full step-by-step instructions in Polish |
 | `instructions.ko` | `string` | Full step-by-step instructions in Korean |
 | `instructions.fr` | `string` | Full step-by-step instructions in French |
-| `instruction_steps.<lang>` | `array[string]` | Same instructions split into an ordered array of steps, per language (`en`, `es`, `it`, `tr`, `ru`, `zh`, `hi`, `pl`, `ko`, `fr`) |
+| `instructions.ptPT` | `string` | Full step-by-step instructions in European Portuguese |
+| `instructions.ptBR` | `string` | Full step-by-step instructions in Brazilian Portuguese |
+| `instruction_steps.<lang>` | `array[string]` | Same instructions split into an ordered array of steps, per language (`en`, `es`, `it`, `tr`, `ru`, `zh`, `hi`, `pl`, `ko`, `fr`, `ptPT`, `ptBR`) |
 | `muscle_group` | `string` | Primary synergist muscle group |
 | `secondary_muscles` | `array[string]` | Additional muscles involved |
 | `target` | `string` | Primary target muscle (e.g. `"biceps"`, `"pectoralis major"`) |
@@ -219,7 +221,9 @@ Each record in `data/exercises.json` follows this structure. A machine-readable
     "hi": "अपने घुटनों को मोड़कर और पैरों को ज़मीन पर सपाट रखते हुए अपनी पीठ के बल लेट जाएँ।...",
     "pl": "Połóż się płasko na plecach, ugnij kolana i oprzyj stopy płasko na pod ...",
     "ko": "등을 바닥에 누워 무릎을 구부리고 발을 바닥에 붙입니다. ...",
-    "fr": "Allonge-toi sur le dos, les genoux fléchis et les pieds à plat au sol. ..."
+    "fr": "Allonge-toi sur le dos, les genoux fléchis et les pieds à plat au sol. ...",
+    "ptPT": "Deite-se de costas com os joelhos fletidos e os pés apoiados no chão. ...",
+    "ptBR": "Deite-se de costas com os joelhos flexionados e os pés apoiados no chão. ..."
   },
   "muscle_group": "hip flexors",
   "secondary_muscles": ["hip flexors", "lower back"],
@@ -338,6 +342,8 @@ print(ex["instructions"]["hi"])  # Hindi
 print(ex["instructions"]["pl"])  # Polish
 print(ex["instructions"]["ko"])  # Korean
 print(ex["instructions"]["fr"])  # French
+print(ex["instructions"]["ptPT"])  # European Portuguese
+print(ex["instructions"]["ptBR"])  # Brazilian Portuguese
 ```
 
 ### Python — Load with Pandas
@@ -390,6 +396,8 @@ console.log(ex.instructions.hi); // Hindi
 console.log(ex.instructions.pl); // Polish
 console.log(ex.instructions.ko); // Korean
 console.log(ex.instructions.fr); // French
+console.log(ex.instructions.ptPT); // European Portuguese
+console.log(ex.instructions.ptBR); // Brazilian Portuguese
 ```
 
 ### TypeScript — Type-safe Usage
@@ -412,6 +420,8 @@ interface Exercise {
     pl: string;
     ko: string;
     fr: string;
+    ptPT: string;
+    ptBR: string;
   };
   instruction_steps: {
     en: string[];
@@ -424,6 +434,8 @@ interface Exercise {
     pl: string[];
     ko: string[];
     fr: string[];
+    ptPT: string[];
+    ptBR: string[];
   };
   muscle_group: string;
   secondary_muscles: string[];