Cross-Pollination Architecture Specification

**Cross-pollination** is PULSE's proactive intelligence layer. It predicts what you need *before you ask* by synthesizing signals from your calendar, tasks, location, time patterns, and conversation history — then delivers contextual predictions to the right device at the right moment. Think of it as a personal chief-of-staff who knows your schedule, remembers your habits, and quietly prepares what you'll need next. | Principle | Description | |---|---| | **Anticipatory, not intrusive** | Predictions surface only when confidence is high and timing is right | | **Privacy-first** | All context stays local or in user-controlled infrastructure; no third-party telemetry | | **Self-improving** | Every accept/reject trains the system to be more accurate over time | | **Cost-aware** | Hard caps on inference calls, token usage, and prediction frequency | | **Multi-modal delivery** | Text cards, voice, push notifications — matched to device and context | The Context Collector continuously gathers and normalizes signals from multiple sources into a unified context frame. | Source | Method | Refresh Rate | Priority | |---|---|---|---| | **Calendar** | Google Calendar API / Apple EventKit | Every 5 min | HIGH | | **Task History** | Clawdbot memory files + conversation logs | On change | HIGH | | **Location** | Node `location_get` (when available) | Every 15 min | MEDIUM | | **Time Patterns** | Derived from historical interaction timestamps | Daily rebuild | MEDIUM | | **Conversation History** | Last N messages from active channels | On message | HIGH | | **Weather** | External API (OpenWeatherMap) | Every 30 min | LOW | | **Device State** | Which devices are active/connected | On change | LOW |