One service, one folder
There is no cluster, no worker, no database server. One process holds the plan and answers everyone.
One process, one port
vigile serve starts the whole service: the MCP endpoint, the HTTP API and the page, on one address and port. It prints where each is:
vigile 0.1.0 main
MCP http://127.0.0.1:8700/mcp
Web http://127.0.0.1:8700/
API http://127.0.0.1:8700/api
It runs in the foreground. Vigile does not install itself as a service: run it under whatever keeps processes alive on your machine. Ctrl-C, or a SIGTERM, finishes the write in flight and stops.
One service per data folder
The service keeps the items and links of the plan in memory. Two services on the same data would each work from their own copy, and each would be wrong about what the other did. So the running service holds the data folder, through a lock the operating system keeps for as long as the process lives:
- a second
serveon the same folder refuses to start, and says which process has it; - a service that was killed leaves nothing behind that keeps the next one out;
restore-backuprefuses while a service holds the folder.
For the same reason the command line never changes the data itself. A command such as vigile create-task is a client: it calls the running service. With no service running it says so, and has sent nothing to anyone.
The command line knows its service
Before a command sends a key, or anything of the plan, it makes sure that what answers on the config's port is the service of this config and not another program that took the port. Security says how.
Several instances
An instance is one config file, one data folder and one port. Several can run side by side and share nothing:
vigile config init --config ~/vigile/second/config.toml --name second --port 8701
vigile serve --config ~/vigile/second/config.toml
The second instance takes a terminal of its own, set up as Install and first run says: serve keeps it.
There is no default location. Every command takes --config <file> or reads VIGILE_CONFIG, so it is always plain which instance a command acts on.
The log
serve writes a line for each change and for each call that fails: which operation, who called, how long it took and how it ended. A setting adds the reads that succeed. Titles, descriptions and keys are never logged. Troubleshooting shows the lines.
Reference
- Install and first run.
- Configuration: every setting.
- Troubleshooting: the doctor, and what to do when a command cannot reach its service.