How m-awais-11/nova-neural-operations-virtual-assistant works
This is a local Windows desktop agent that turns natural-language requests into allowlisted computer actions. Users enter commands through the CLI or Tkinter GUI; the agent plans tool calls with a configured language-model service, executes them behind validation and confirmation gates, verifies outcomes, and returns a response. The product also includes file and application tools, browser tools, and SQLite history with sensitive-data redaction. README capabilities inform the browser/tool coverage; unsampled module internals are not expanded, and uncertain connections are omitted.
Main entry points: executor.py Tool executor · agent.py Agent orchestrator · planner.py Task planner · memory.py Interaction memory
Components 20
User interfaces
- CLI
main.py - Desktop GUI
gui.py
Agent orchestration
- Agent orchestrator
agent/agent.py - Task planner
agent/planner.py - Tool executor
agent/executor.py - Outcome verifier
agent/verifier.py
Allowlisted tools
- Tool registry
tools/registry.py - File tools
tools/file_tools.py - Application tools
tools/app_tools.py - Browser tools
tools/browser_tools.py
Security controls
- Path sandbox
security/path_validator.py - Permissions
security/permissions.py - Confirmation gate
security/confirmation.py
Configuration and memory
- Agent settings
config/settings.py - Interaction memory
memory/memory.py - SQLite database
Other components
- User
- LLM service
- Local files
- Desktop applications
Connections 28
- User → CLI: enters command
- User → Desktop GUI: enters command
- CLI → Agent orchestrator: runs task
- Desktop GUI → Agent orchestrator: runs task
- Agent orchestrator → Task planner: requests plan
- Task planner → Tool registry: gets tool specs
- Task planner → LLM service: sends request
- LLM service → Task planner: returns plan
- Agent orchestrator → Tool executor: dispatches calls
- Tool executor → Tool registry: looks up tool
- Tool executor → Confirmation gate: requests approval
- Tool executor → Outcome verifier: checks outcome
- Tool executor → File tools: executes tool
- Tool executor → Application tools: executes tool
- Tool executor → Browser tools: executes tool
- File tools → Path sandbox: validates paths
- Application tools → Path sandbox: validates target
- File tools → Local files: reads and writes
- Outcome verifier → Local files: checks disk state
- Application tools → Desktop applications: launches process
- Outcome verifier → Desktop applications: checks process state
- Path sandbox → Agent settings: reads allowed roots
- Confirmation gate → Agent settings: reads policy
- Tool executor → Permissions: handles access errors
- Agent orchestrator → Interaction memory: logs interactions
- CLI → Interaction memory: reads history
- Interaction memory → SQLite database: reads and writes
- Interaction memory → Agent settings: reads database path