# Alfred.report (Full System Reference) > One wake word. Closed work comes back. You stay out of the middle. Alfred is a Cloudflare-native operational harness and voice-first AI Chief of Staff operated by OpenRoyleAl. ## 1. System Architecture Alfred replaces disjointed agent chat sessions with a unified execution spine: ``` iPhone / Siri / Shortcuts → voice.alfred.report (Cloudflare Worker) → MiMo TTS (WAV) ↓ Alfred.report D1 (agent-alfred-knowledge) ↓ Mission Kernel & Evidence Gates ↓ alfred.report Dashboard (Cloudflare Worker + Static Assets) ``` ### Components 1. **ORAL Worker (`voice.alfred.report`)**: Voice endpoints, MiMo neural text-to-speech engine, session management, and Durable Object orchestration. 2. **Board Worker (`alfred.report`)**: Mission management dashboard, REST API proxy, iOS Siri onboarding flow, and developer resources. 3. **Storage & State**: Cloudflare D1 SQL database (`agent-alfred-knowledge`), Cloudflare Agent Memory (`hermes` namespace), and Cloudflare R2 object store. 4. **AI Gateway**: `oracle960` gateway routing all LLM and neural TTS token streams with strict cost monitoring. ## 2. Mission Model & Lifecycle Every task in Alfred.report is modeled as a formal `Mission`: - **States**: `drafted` → `active` → `gated` → `complete` | `failed` | `aborted` - **Budget**: Bounded token and USD limits (`max_tokens`, `max_usd`, `spend_class`). - **Acceptance Criteria**: Explicit array of requirements that must be validated prior to closure. - **Evidence Trail**: Cryptographic or structured attestations (test passes, artifact links, screenshots) proving requirements were met. - **Closure Gate**: Missions attempting to complete without verified evidence or with zero recorded usage are rejected. ## 3. Detailed API Reference ### Voice Endpoints (`https://voice.alfred.report`) #### GET /voice/briefing - **Parameters**: `token` (string, required query param or Bearer header). - **Behavior**: Analyzes active, gated, and closed missions in D1 for the tenant; generates a concise speech summary; converts to audio using the tenant's selected voice persona; returns a standard `audio/wav` stream. - **Response**: `200 OK` with `Content-Type: audio/wav`. #### GET / POST /voice - **Parameters**: `token` (string), `text` (string), `style` (optional string). - **Behavior**: Synthesizes provided text using the MiMo neural TTS model into natural, resonant speech. - **Response**: `200 OK` with `Content-Type: audio/wav`. #### GET /auth - **Parameters**: None. - **Behavior**: Generates and persists a secure session token in D1, then redirects the user to `https://iphone.alfred.report/?token={TOKEN}` for 1-click iOS Siri setup. ### Command API (`https://alfred.report`) #### GET /api/command/missions - **Description**: Lists recent missions filtered by status (`active`, `gated`, `closed`, `all`). - **Response**: JSON array of Mission objects. #### POST /api/command/missions - **Description**: Creates a new mission with objective, budget, and acceptance criteria. - **Body**: `{ "objective": string, "acceptance": string[], "budget": { "max_usd": number, "max_tokens": number } }` #### GET /api/command/stats - **Description**: Returns 24-hour mission closure rate, token burn, and active counts. ## 4. When to Use Alfred - Autonomous multi-step operations requiring auditability and evidence verification. - Hands-free voice debriefs on iOS, watchOS, and macOS via "Hey Siri, Alfred report". - SEO portfolio monitoring (140+ domains) with automated site audits and rank verification. ## 5. Trust, Governance & Contact - **Operator**: OpenRoyleAl - **Headquarters**: 71-75 Shelton Street, Covent Garden, London, WC2H 9JQ, United Kingdom - **Email**: hans@icebergmedia.co.uk - **Documentation**: https://alfred.report/api/docs - **OpenAPI 3.1 Spec**: https://alfred.report/api/openapi.json - **Sitemap**: https://alfred.report/sitemap.xml