Data formats and locations
All Nerve-home paths move with NERVE_HOME; default is ~/.nerve.
| Data | Default location / format |
|---|---|
| Nerve-home marker | <NERVE_HOME>/manifest.json; nerve-home v1 |
| Canonical domain records | <NERVE_HOME>/data/nerve.sqlite plus SQLite WAL/SHM |
| Conversation/tool payloads | <NERVE_HOME>/data/conversations |
| Durable images and plans | <NERVE_HOME>/data/images and <NERVE_HOME>/data/plans |
| Explore reports | <NERVE_HOME>/data/reports |
| RPC, maintenance, and trust state | Tables/documents in <NERVE_HOME>/data/nerve.sqlite |
| Rebuildable query cache | <NERVE_HOME>/cache/query-cache.sqlite plus SQLite sidecars |
| Other disposable caches | Other content under <NERVE_HOME>/cache |
| Task records and logs | <NERVE_HOME>/data/tasks |
| Agent resources | <NERVE_HOME>/agent |
| Desktop/daemon/event logs | <NERVE_HOME>/logs, JSONL |
| Crash/Node reports | <NERVE_HOME>/crashes; age-retained diagnostics |
| Configuration and credentials | <NERVE_HOME>/config and encrypted <NERVE_HOME>/secrets |
| TLS CA/leaf material | <NERVE_HOME>/tls when mobile HTTPS is enabled |
| Active daemon discovery/lease | <NERVE_HOME>/daemon.json; absent after clean shutdown |
| Migration history and backups | <NERVE_HOME>/migrations and <NERVE_HOME>/backups |
| Python/large tool artifacts | Nerve artifact paths returned by the tool |
| Pasted clipboard images | OS temp directory under nerve/; not durable attachments |
Conversation exports
Section titled “Conversation exports”JSON uses bundle format nerve.conversation.v1 and includes all stored entries, project, conversation, and agents. Markdown and escaped HTML are reading formats. Imports remap IDs and can skip malformed optional records.
Electron profile
Section titled “Electron profile”Electron browser cache, cookies, and local/session storage live in the platform Electron userData profile outside NERVE_HOME. Back up/isolate both only when browser state is required.
Sensitive data
Section titled “Sensitive data”Logs, exports, reports, artifacts, and task output can contain prompts, source, paths, commands, remote URLs, or provider details. Redact before sharing.