✅ TICKET-006: Wake-word Detection Service - Implemented wake-word detection using openWakeWord - HTTP/WebSocket server on port 8002 - Real-time detection with configurable threshold - Event emission for ASR integration - Location: home-voice-agent/wake-word/ ✅ TICKET-010: ASR Service - Implemented ASR using faster-whisper - HTTP endpoint for file transcription - WebSocket endpoint for streaming transcription - Support for multiple audio formats - Auto language detection - GPU acceleration support - Location: home-voice-agent/asr/ ✅ TICKET-014: TTS Service - Implemented TTS using Piper - HTTP endpoint for text-to-speech synthesis - Low-latency processing (< 500ms) - Multiple voice support - WAV audio output - Location: home-voice-agent/tts/ ✅ TICKET-047: Updated Hardware Purchases - Marked Pi5 kit, SSD, microphone, and speakers as purchased - Updated progress log with purchase status 📚 Documentation: - Added VOICE_SERVICES_README.md with complete testing guide - Each service includes README.md with usage instructions - All services ready for Pi5 deployment 🧪 Testing: - Created test files for each service - All imports validated - FastAPI apps created successfully - Code passes syntax validation 🚀 Ready for: - Pi5 deployment - End-to-end voice flow testing - Integration with MCP server Files Added: - wake-word/detector.py - wake-word/server.py - wake-word/requirements.txt - wake-word/README.md - wake-word/test_detector.py - asr/service.py - asr/server.py - asr/requirements.txt - asr/README.md - asr/test_service.py - tts/service.py - tts/server.py - tts/requirements.txt - tts/README.md - tts/test_service.py - VOICE_SERVICES_README.md Files Modified: - tickets/done/TICKET-047_hardware-purchases.md Files Moved: - tickets/backlog/TICKET-006_prototype-wake-word-node.md → tickets/done/ - tickets/backlog/TICKET-010_streaming-asr-service.md → tickets/done/ - tickets/backlog/TICKET-014_tts-service.md → tickets/done/
241 lines
8.1 KiB
Python
241 lines
8.1 KiB
Python
"""
|
|
Boundary enforcement policy.
|
|
|
|
Enforces separation between work and family agents through:
|
|
- Path whitelisting
|
|
- Config separation
|
|
- Network-level checks
|
|
- Static policy validation
|
|
"""
|
|
|
|
import logging
|
|
from pathlib import Path
|
|
from typing import List, Optional, Set
|
|
from dataclasses import dataclass
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
|
@dataclass
|
|
class BoundaryPolicy:
|
|
"""Policy for agent boundaries."""
|
|
agent_type: str # "work" or "family"
|
|
allowed_paths: Set[Path]
|
|
forbidden_paths: Set[Path]
|
|
allowed_networks: Set[str] # IP ranges or network names
|
|
forbidden_networks: Set[str]
|
|
allowed_tools: Set[str]
|
|
forbidden_tools: Set[str]
|
|
|
|
|
|
class BoundaryEnforcer:
|
|
"""Enforces boundaries between work and family agents."""
|
|
|
|
def __init__(self):
|
|
"""Initialize boundary enforcer with policies."""
|
|
# Base paths
|
|
self.base_dir = Path(__file__).parent.parent.parent
|
|
|
|
# Family agent policy
|
|
self.family_policy = BoundaryPolicy(
|
|
agent_type="family",
|
|
allowed_paths={
|
|
self.base_dir / "data" / "tasks" / "home",
|
|
self.base_dir / "data" / "notes" / "home",
|
|
self.base_dir / "data" / "conversations.db",
|
|
self.base_dir / "data" / "timers.db",
|
|
},
|
|
forbidden_paths={
|
|
# Work-related paths (if they exist)
|
|
self.base_dir.parent / "work-repos", # Example
|
|
self.base_dir / "data" / "work", # If work data exists
|
|
},
|
|
allowed_networks={
|
|
"localhost",
|
|
"127.0.0.1",
|
|
"10.0.0.0/8", # Local network (can be restricted further)
|
|
},
|
|
forbidden_networks={
|
|
# Work-specific networks (if configured)
|
|
},
|
|
allowed_tools={
|
|
"get_current_time",
|
|
"get_date",
|
|
"get_timezone_info",
|
|
"convert_timezone",
|
|
"get_weather",
|
|
"create_timer",
|
|
"create_reminder",
|
|
"list_timers",
|
|
"cancel_timer",
|
|
"add_task",
|
|
"update_task_status",
|
|
"list_tasks",
|
|
"create_note",
|
|
"read_note",
|
|
"append_to_note",
|
|
"search_notes",
|
|
"list_notes",
|
|
},
|
|
forbidden_tools={
|
|
# Work-specific tools (if any)
|
|
}
|
|
)
|
|
|
|
# Work agent policy
|
|
self.work_policy = BoundaryPolicy(
|
|
agent_type="work",
|
|
allowed_paths={
|
|
# Work agent can access more paths
|
|
self.base_dir / "data" / "tasks" / "home", # Can read home tasks
|
|
self.base_dir / "data" / "work", # Work-specific data
|
|
},
|
|
forbidden_paths={
|
|
# Work agent should not modify family data
|
|
self.base_dir / "data" / "notes" / "home", # Family notes
|
|
},
|
|
allowed_networks={
|
|
"localhost",
|
|
"127.0.0.1",
|
|
"10.0.0.0/8",
|
|
"10.0.30.63", # GPU VM
|
|
},
|
|
forbidden_networks=set(),
|
|
allowed_tools={
|
|
# Work agent can use all tools
|
|
},
|
|
forbidden_tools=set()
|
|
)
|
|
|
|
def check_path_access(self, agent_type: str, path: Path) -> tuple[bool, str]:
|
|
"""
|
|
Check if agent can access a path.
|
|
|
|
Args:
|
|
agent_type: "work" or "family"
|
|
path: Path to check
|
|
|
|
Returns:
|
|
(allowed, reason)
|
|
"""
|
|
policy = self.family_policy if agent_type == "family" else self.work_policy
|
|
|
|
# Resolve path
|
|
try:
|
|
resolved_path = path.resolve()
|
|
except Exception as e:
|
|
return False, f"Invalid path: {e}"
|
|
|
|
# Check forbidden paths first
|
|
for forbidden in policy.forbidden_paths:
|
|
try:
|
|
if resolved_path.is_relative_to(forbidden.resolve()):
|
|
return False, f"Path is in forbidden area: {forbidden}"
|
|
except Exception:
|
|
continue
|
|
|
|
# Check allowed paths
|
|
for allowed in policy.allowed_paths:
|
|
try:
|
|
if resolved_path.is_relative_to(allowed.resolve()):
|
|
return True, f"Path is in allowed area: {allowed}"
|
|
except Exception:
|
|
continue
|
|
|
|
# Default: deny for family agent, allow for work agent
|
|
if agent_type == "family":
|
|
return False, "Path not in allowed whitelist for family agent"
|
|
else:
|
|
return True, "Work agent has broader access"
|
|
|
|
def check_tool_access(self, agent_type: str, tool_name: str) -> tuple[bool, str]:
|
|
"""
|
|
Check if agent can use a tool.
|
|
|
|
Args:
|
|
agent_type: "work" or "family"
|
|
tool_name: Name of tool
|
|
|
|
Returns:
|
|
(allowed, reason)
|
|
"""
|
|
policy = self.family_policy if agent_type == "family" else self.work_policy
|
|
|
|
# Check forbidden tools
|
|
if tool_name in policy.forbidden_tools:
|
|
return False, f"Tool '{tool_name}' is forbidden for {agent_type} agent"
|
|
|
|
# Check allowed tools (if specified)
|
|
if policy.allowed_tools and tool_name not in policy.allowed_tools:
|
|
return False, f"Tool '{tool_name}' is not in allowed list for {agent_type} agent"
|
|
|
|
return True, f"Tool '{tool_name}' is allowed for {agent_type} agent"
|
|
|
|
def check_network_access(self, agent_type: str, target: str) -> tuple[bool, str]:
|
|
"""
|
|
Check if agent can access a network target.
|
|
|
|
Args:
|
|
agent_type: "work" or "family"
|
|
target: Network target (IP, hostname, or network range)
|
|
|
|
Returns:
|
|
(allowed, reason)
|
|
"""
|
|
policy = self.family_policy if agent_type == "family" else self.work_policy
|
|
|
|
# Check forbidden networks
|
|
for forbidden in policy.forbidden_networks:
|
|
if self._matches_network(target, forbidden):
|
|
return False, f"Network '{target}' is forbidden for {agent_type} agent"
|
|
|
|
# Check allowed networks
|
|
for allowed in policy.allowed_networks:
|
|
if self._matches_network(target, allowed):
|
|
return True, f"Network '{target}' is allowed for {agent_type} agent"
|
|
|
|
# Default: deny for family agent, allow for work agent
|
|
if agent_type == "family":
|
|
return False, f"Network '{target}' is not in allowed whitelist for family agent"
|
|
else:
|
|
return True, "Work agent has broader network access"
|
|
|
|
def _matches_network(self, target: str, network: str) -> bool:
|
|
"""Check if target matches network pattern."""
|
|
# Simple matching - can be enhanced with proper CIDR matching
|
|
if network == target:
|
|
return True
|
|
if network.endswith("/8") and target.startswith(network.split("/")[0].rsplit(".", 1)[0]):
|
|
return True
|
|
return False
|
|
|
|
def validate_config_separation(self, agent_type: str, config_path: Path) -> tuple[bool, str]:
|
|
"""
|
|
Validate that config is properly separated.
|
|
|
|
Args:
|
|
agent_type: "work" or "family"
|
|
config_path: Path to config file
|
|
|
|
Returns:
|
|
(valid, reason)
|
|
"""
|
|
# Family agent config should not contain work-related paths
|
|
if agent_type == "family":
|
|
config_content = config_path.read_text()
|
|
work_indicators = ["work-repos", "work/", "work_"]
|
|
for indicator in work_indicators:
|
|
if indicator in config_content:
|
|
return False, f"Family config contains work indicator: {indicator}"
|
|
|
|
return True, "Config separation validated"
|
|
|
|
|
|
# Global enforcer instance
|
|
_enforcer = BoundaryEnforcer()
|
|
|
|
|
|
def get_enforcer() -> BoundaryEnforcer:
|
|
"""Get the global boundary enforcer instance."""
|
|
return _enforcer
|