# Deep-Dive Analysis — Rethink! EAM 2026 Vortragsportal - **Project:** `rethink_dump` - **Analyst:** Mary (BMAD Business Analyst persona), BMM v6.10.0 - **Date:** 2026-07-07 - **Scope:** Full project deep dive + offline-hardening plan and execution - **Method:** Minto Pyramid (answer-first), evidence grounded in the repository files --- ## 1. Answer first (executive summary) `rethink_dump` is a **single-purpose, dependency-free, static offline web portal** that renders sessions of the *Rethink! EAM 2026* conference as filterable cards. Data originates from a manual browser DOM capture of the [hubs101](https://app.hubs101.com) agenda app, stored as a raw text dump, transformed by one Python script into a JSON manifest, and rendered client-side by a single `app.js`. BMAD tooling (`_bmad/`, `.claude/skills/`) is installed but had produced no artifacts yet. The project's stated purpose is **100 % offline availability of all sessions, downloads, and videos**. Before this pass it was *mostly* offline but had three concrete gaps that broke that promise. All three are now closed (see §6). --- ## 2. System overview ``` Browser (hubs101 agenda) ──manual capture──► output/hubs101_agenda_dump.txt (source of truth: raw) │ scripts/parse_sessions.py (transform) ▼ data/sessions.json + data/sessions.js (manifest + file:// fallback) downloads/attachments/.md (offline note per session) downloads/videos/.html (offline video page per session) │ index.html + app.js + styles.css (render layer) ▼ Filterable session-card UI ``` ### Component inventory | Component | File | Role | |---|---|---| | Raw agenda dump | `output/hubs101_agenda_dump.txt` | Human-captured source; 6 sessions, block format `=== Session N ===` | | Parser | `scripts/parse_sessions.py` | Single reproducible transform → all generated artifacts | | Session manifest | `data/sessions.json` | Machine-read data consumed by the UI | | file:// fallback | `data/sessions.js` | Same data as `window.__SESSIONS__` so the app runs without a server | | Offline notes | `downloads/attachments/.md` | Self-contained per-session archive (full text) | | Offline video pages | `downloads/videos/.html` | Self-contained per-session player + drop-in slot for a real `.mp4` | | App shell | `index.html` | Layout, hero, toolbar, script wiring | | Renderer | `app.js` | Fetch/fallback data, search + type + date filters, card render | | Styling | `styles.css` | Dark theme, responsive grid, no external fonts/CSS | --- ## 3. Data model A session object (manifest schema): | Field | Type | Notes | |---|---|---| | `id` | string | `session-`, positional | | `type` | enum | `NETWORKING` \| `OPENING` \| `PRESENTATION` (fallback `PRESENTATION`) | | `date` / `time` / `timezone` | string | As captured, German locale | | `location` | string | Room / area | | `title` / `description` | string | `description` is bullet-flattened to one line for card display | | `speaker` / `role` / `company` | string | May be empty (networking/track blocks) | | `source` | string | Provenance pointer to the raw dump | | `attachments[]` | array | `{label, file, kind, type}` — raw export, offline note, offline video | **Captured content (updated 2026-07-07):** the **full agenda of both conference days — 62 sessions** (35 on `Mo, 29.06.2026`, 27 on `Di, 30.06.2026`), pulled from the live we.CONECT/hubs101 API, not a hand capture. Types: 27 PRESENTATION, 18 WORKSHOP, 10 NETWORKING, 2 OPENING, 2 CLOSING, 1 KEYNOTE, 1 DISCUSSION, 1 DEMO. Full descriptions and speaker biographies included. Source of truth is `output/hubs101_agenda_full.json` (raw API response); the parser regenerates everything else from it. > **Provenance:** the hubs101 app (`app.hubs101.com`) is a login-gated SPA and cannot be scraped > headless. The same agenda is served publicly by the marketing backend at > `https://api.we-conect.com/events/63507a6a-605d-411e-ba21-38c97fa6a768/agenda` (event UUID resolved > from `api.we-conect.com/events`). That JSON is the authoritative offline capture. --- ## 4. Architecture assessment **Strengths** - **Zero runtime dependencies** — pure HTML/CSS/JS + one stdlib-only Python script. Nothing to install, nothing to break, trivially archivable. - **Clean separation** raw → transform → manifest → view. Provenance is explicit (`source`). - **XSS-safe render** — `escapeHtml()` applied to every interpolated field in `app.js`. - **Deterministic parser** — regex block split, tolerant key/value parsing, safe type fallback. **Risks / limitations (pre-pass)** 1. **`file://` breakage** — `fetch('./data/sessions.json')` fails under `file://` (CORS), so "just open index.html" did *not* work; a server was mandatory. → **fixed** via `sessions.js`. 2. **Parser drift** — `sessions.json` was hand-enriched with offline-note attachments the parser did not generate. Re-running the parser would have *deleted* that enrichment. → **fixed**: the parser is now the single source of truth for every attachment and offline file. 3. **Thin offline archive** — notes were metadata stubs without the session description, so the "offline note" did not actually preserve the session content. → **fixed**: notes now embed the full description and bullet list. 4. **No video layer** — the goal names videos, but no video structure existed. → **fixed**: a self-contained offline video page per session with a drop-in `.mp4` slot. 5. **Coverage** — *resolved.* The prior pass held only 6 hand-captured sessions. The full agenda for both days (62 sessions) is now captured from the public we.CONECT API and integrated. See §7 for the one remaining refresh caveat. --- ## 5. BMAD context - Install: BMAD **v6.10.0**, module **bmm** (`software-development` team), languages English, user `Deltatree`, skill level intermediate. - Agents available: Analyst (Mary), Tech Writer (Paige), PM (John), UX (Sally), Architect (Winston), Dev (Amelia). - `_bmad-output/` (planning + implementation artifacts) and `docs/` were **empty** — this document is the first project-knowledge artifact. --- ## 6. Offline-hardening — what changed this pass 1. **Parser is now the single source of truth.** `scripts/parse_sessions.py` regenerates, from the one raw dump: `data/sessions.json`, `data/sessions.js`, every `downloads/attachments/.md`, and every `downloads/videos/.html`. One command reproduces the entire offline archive. 2. **Enriched offline notes** — each note now contains the full metadata *and* the complete description with bullet points preserved: a genuinely self-contained offline record. 3. **Offline video pages** — one self-contained HTML page per session, dark-themed, no external assets, with a `