Architecture

pane-awareness is a coordination layer that runs alongside AI coding sessions. Each session writes its state to shared JSON files, and reads other sessions' state to make coordination decisions. No server, no database, no daemon. Sessions coordinate through locked JSON files on the local filesystem. This makes the system: - **Zero-dependency** — nothing to install beyond Python - **Zero-configuration** — works out of the box - **Crash-resilient** — stale panes auto-expire, no orphan processes - **Portable** — works on macOS, Linux, and Windows 1. Lowercase and tokenize the prompt 2. Remove stop words (configurable set of ~80 common words + convergence noise words) 3. Remove identity noise (username, hostname, configurable extras) 4. Deduplicate and take top N by frequency in the prompt 5. Store in trajectory window (rolling last 10 prompt snapshots) - **deepening**: Appears in 60%+ of recent prompts AND frequency is increasing - **emerging**: Appears only in the last 3 prompts (new topic) - **fading**: Was present in older prompts but absent from the last 3 - **stable**: Consistent presence without significant trend