I think I have it!
| ID | Requirement | Details | |----|-------------|---------| | | Capture Entry Point | • A persistent floating “Lo Re Paku” FAB (Floating Action Button) on iOS/Android. • Long‑press opens a modal with three capture modes: Text , Voice , Image . | | FR‑2 | Automatic Metadata | • Device‑level data: timestamp, GPS (optional), current foreground app bundle ID. • User can edit/override metadata after capture. | | FR‑3 | Session Detection | • Sliding‑window clustering (default 15 min idle) based on time, location, and app context. • Users can manually merge or split sessions. | | FR‑4 | AI Summarization | • Backend service using LLM (e.g., OpenAI GPT‑4o / Claude 3.5) with a custom prompt: “Summarize the following fragments into a concise 2‑sentence overview, preserving key entities and actions.” | | FR‑5 | Depth Options | • Brief (≤2 sentences) • Standard (≈5 bullet points) • Detailed (full paragraph). | | FR‑6 | Task Conversion | • Each bullet or sentence gets a “Create Task” button → opens task‑creation UI pre‑filled with text, optional due date (extracted via date‑recognition). | | FR‑7 | Export/Share | • PDF (styled), Markdown, plain text, and a sharable public link (read‑only). | | FR‑8 | Security | • AES‑256 at‑rest encryption. • TLS 1.3 for all API calls. • Optional Zero‑Knowledge mode where the server stores only ciphertext (key managed on device). | | FR‑9 | Integration API | • POST /sessions webhook with JSON payload (session_id, summary, fragments, metadata). • OAuth 2.0 client‑credentials flow for third‑party apps. | | FR‑10 | Settings | • Enable/disable GPS, choose default summarization depth, toggle “End‑to‑End Encryption,” manage API keys. | lo re paku
+-------------------+ +-------------------+ | Home Screen | | Any App Screen | | | | | | [ + ] | <-- | [ Lo Re Paku ] | | (FAB) | | (floating) | +-------------------+ +-------------------+ I think I have it
The term "Lo Re Paku" can sometimes lead to confusion due to the linguistic components of its title: | | FR‑2 | Automatic Metadata | •
| Category | Requirement | |----------|-------------| | | Summarization latency ≤ 3 seconds for ≤ 20 fragments (average). | | Scalability | Backend must handle 10 k concurrent users with burst peaks of 2×. | | Reliability | 99.9 % uptime SLA for the summarization service. | | Usability | Capture flow ≤ 2 taps + 1 second voice recording. | | Accessibility | All UI elements accessible via VoiceOver/TalkBack; provide ARIA labels. | | Localization | UI strings ready for i18n; summarization works for English, Spanish, French, Mandarin (via LLM language support). | | Privacy | GDPR & CCPA compliant; ability to delete a session permanently (right‑to‑be‑forgotten). |
| Criterion | Test Description | |-----------|------------------| | | When the FAB is tapped, the capture modal appears within 300 ms . | | AC‑2 | A voice memo of up to 30 seconds is saved and transcribed with ≥ 95 % accuracy (based on a benchmark set). | | AC‑3 | After three captures within a 10‑minute window, the system automatically creates a single session. | | AC‑4 | The AI‑generated brief summary contains ≤ 2 sentences and includes at least 2 of the top‑5 extracted entities. | | AC‑5 | Tapping “Create Task” opens the native task‑creation UI with the summary text pre‑filled. | | AC‑6 | Exported PDF matches the on‑screen layout and is ≤ 200 KB for a session of ≤ 15 fragments. | | AC‑7 | Enabling “End‑to‑End Encryption” results in the server storing only ciphertext; a security audit can’t retrieve raw content. | | AC‑8 | The webhook fires within 1 second of session completion, delivering correct JSON payload. | | AC‑9 | Deleting a session removes all associated data from the device and the backend within 5 seconds . |