Skip to content

livesync-bridge: restore bidirectional sync at 60s poll (agent write workflow) #176

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
AlienWalker1995 merged 1 commit into from
Aug 12, 2026
Merged

livesync-bridge: restore bidirectional sync at 60s poll (agent write workflow) #176

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 10 additions & 12 deletions services/obsidian-livesync/plugin.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -59,18 +59,16 @@ services:
image: ordo/livesync-bridge:latest
wants_secrets: true
env:
# ONE-WAY materializer. The bridge's job here is CouchDB -> notes/ files (network-driven,
# always live) so the AI stack (rag-ingestion + memory-vault MCP) can read the vault. The
# REVERSE direction (disk -> CouchDB) needs a file WATCHER, but 9p bind mounts emit no inotify
# events, so the only way to catch host/AI disk writes is chokidar POLLING every file every few
# seconds — on a multi-thousand-file vault that stats ~100k files/min and pegs a core forever
# (~128% CPU idle). The AI-RAG use case never needs disk -> CouchDB (humans edit on devices,
# which sync to CouchDB over the network), so polling is OFF: chokidar sits on inotify and gets
# no events on 9p → ~0 idle CPU, while CouchDB -> disk keeps working. Set USEPOLLING="1" ONLY
# if you edit notes directly on the host filesystem and need those pushed back up to CouchDB
# (INTERVAL is then the poll period; 15s balances CPU vs sync latency on a large vault).
CHOKIDAR_USEPOLLING: "0"
CHOKIDAR_INTERVAL: "15000"
# BIDIRECTIONAL sync. CouchDB -> notes/ is network-driven (always live) so the AI stack
# (rag-ingestion + memory-vault MCP) reads the vault. disk -> CouchDB lets agents (Hermes /
# Claude) and host edits to notes/ flow UP to CouchDB -> every device. That upward direction
# needs a file WATCHER, but 9p bind mounts emit no inotify events, so it must POLL. Polling
# stats every file each tick (O(files)): at 3s on a ~5k-file vault it pegged a core (~128% CPU
# idle). 60s keeps the agent-write workflow (updates reach CouchDB within a minute) at ~6% CPU.
# Lower INTERVAL = faster sync, more CPU; set USEPOLLING="0" for a read-only (RAG-only) mirror
# at ~0 idle CPU if nothing writes to notes/ on the host side.
CHOKIDAR_USEPOLLING: "1"
CHOKIDAR_INTERVAL: "60000"
volumes:
# The vault's notes/ subfolder <-> the bridge's storage root. Same MEMORY_VAULT_PATH the
# memory-vault MCP (RW) and rag-ingestion (RO) mount, so notes land in the ONE shared vault
Expand Down
Loading