Self-Healing Terminal
The Self-Healing Terminal is one of OmniIDE's defining features. When a command fails — whether it is a compilation error, a missing dependency, or a runtime exception — the terminal does not just display the error. It actively fixes the problem.
How Self-Healing Works
- Error Detection — The terminal monitors stdout and stderr for error patterns, non-zero exit codes, and stack traces
- Root Cause Analysis — The error output is sent to the LLM (Gemini or Ollama) with full workspace context
- Fix Generation — The LLM generates a corrective patch — code changes, dependency installs, or configuration updates
- Automatic Re-execution — The fix is applied and the original command is re-run
Example Scenario
$ npm start
> node server.js
Error: Cannot find module 'express'
// Agent detects the error
// Agent runs: npm install express
// Agent re-runs: npm start
// Server starts successfully ✓Safety: All healing operations run inside the Hardened Sandbox. The agent cannot modify files outside your workspace or execute system-level commands.
Supported Error Types
- Module not found / import errors
- Syntax and compilation errors
- Runtime exceptions and crashes
- Missing dependencies
- Configuration errors
- Permission issues (within sandbox scope)
- Test failures
Configuration
{
"omni.selfHealing.enabled": true,
"omni.selfHealing.maxAttempts": 3,
"omni.selfHealing.autoApply": true
}Related
- Agentic Loop — The execution cycle that drives self-healing
- Verification Engine — How errors are detected
- Recovery System — The fix generation pipeline