What's New in Hrida Terminal
Recent improvements to Hrida Terminal.
Deployment
Production Compose file
A ready-made docker-compose-terminal-prod.yaml with health checks, log rotation, CPU/memory/process limits, a clean-shutdown grace period and hardened security defaults. Per-server settings live in a single .env file, with an .env.example template to copy.
OpenShift image
A new Dockerfile.openshift builds an image that runs under OpenShift's arbitrary non-root UIDs. Features that need root (multi-user mode, egress filtering, startup package installs) are disabled there, with a clear error or warning.
Host and port from environment variables
HRIDA_TERMINAL_HOST and HRIDA_TERMINAL_PORT now set the bind address and port for both hrida-terminal run and hrida-terminal mcp, so containers can be configured without changing the start command.
Multi-user mode
Works with hardened containers
The new HRIDA_TERMINAL_NO_NEW_PRIVILEGES setting controls Docker's no-new-privileges option in the bundled Compose files. Set it to false when you enable multi-user mode, because creating per-user accounts needs sudo.
Clear startup check
If multi-user mode is enabled but the container can't use sudo, Hrida Terminal now stops at startup with a message saying exactly what to change. Previously it started normally and then failed on the first terminal request with a long error trace.
Readable account errors
When creating a user account fails, the log now includes the actual reason from the system (for example Permission denied), instead of only "returned non-zero exit status 1".
Files and the AI
~ paths work as expected
Paths like ~/report.py now resolve to the user's home folder. Previously the AI could create a literal folder named ~ that didn't show up where you expected it.
Clearer file errors
When a file write, move, delete or folder creation fails, the AI now gets the real reason, such as Permission denied: /etc/x. Write inside your home directory instead. That lets it correct itself rather than retrying blindly.
More reliable AI reporting
The built-in system prompt now tells the AI to run commands itself instead of asking you to, to report a file as created only after the tool confirms it, and to show the error when something fails instead of claiming success.