Initial commit: Split Macha autonomous system into separate flake
Macha is now a standalone NixOS flake that can be imported into other systems. This provides: - Independent versioning - Easier reusability - Cleaner separation of concerns - Better development workflow Includes: - Complete autonomous system code - NixOS module with full configuration options - Queue-based architecture with priority system - Chunked map-reduce for large outputs - ChromaDB knowledge base - Tool calling system - Multi-host SSH management - Gotify notification integration All capabilities from DESIGN.md are preserved.
This commit is contained in:
Lily Miller
522
chat.py
Normal file
522
chat.py
Normal file
@@ -0,0 +1,522 @@
|
||||
#!/usr/bin/env python3
|
||||
"""
|
||||
Interactive chat interface with Macha AI agent.
|
||||
Allows conversational interaction and directive execution.
|
||||
"""
|
||||
|
||||
import json
|
||||
import os
|
||||
import subprocess
|
||||
import sys
|
||||
from datetime import datetime
|
||||
from pathlib import Path
|
||||
from typing import List, Dict, Any
|
||||
|
||||
# Add parent directory to path for imports
|
||||
sys.path.insert(0, str(Path(__file__).parent))
|
||||
|
||||
from agent import MachaAgent
|
||||
|
||||
|
||||
class MachaChatSession:
|
||||
"""Interactive chat session with Macha"""
|
||||
|
||||
def __init__(self):
|
||||
self.agent = MachaAgent(use_queue=True, priority="INTERACTIVE")
|
||||
self.conversation_history: List[Dict[str, str]] = []
|
||||
self.session_start = datetime.now().isoformat()
|
||||
|
||||
def _create_chat_prompt(self, user_message: str) -> str:
|
||||
"""Create a prompt for the chat session"""
|
||||
|
||||
# Build conversation context
|
||||
context = ""
|
||||
if self.conversation_history:
|
||||
context = "\n\nCONVERSATION HISTORY:\n"
|
||||
for entry in self.conversation_history[-10:]: # Last 10 messages
|
||||
role = entry['role'].upper()
|
||||
msg = entry['message']
|
||||
context += f"{role}: {msg}\n"
|
||||
|
||||
prompt = f"""{MachaAgent.SYSTEM_PROMPT}
|
||||
|
||||
TASK: INTERACTIVE CHAT SESSION
|
||||
|
||||
You are in an interactive chat session with the system administrator.
|
||||
You can have a natural conversation and execute commands when directed.
|
||||
|
||||
CAPABILITIES:
|
||||
- Answer questions about system status
|
||||
- Explain configurations and issues
|
||||
- Execute commands when explicitly asked
|
||||
- Provide guidance and recommendations
|
||||
|
||||
COMMAND EXECUTION:
|
||||
When the user asks you to run a command or perform an action that requires execution:
|
||||
1. Respond with a JSON object containing the command to execute
|
||||
2. Format: {{"action": "execute", "command": "the command", "explanation": "why you're running it"}}
|
||||
3. After seeing the output, continue the conversation naturally
|
||||
|
||||
RESPONSE FORMAT:
|
||||
- For normal conversation: Respond naturally in plain text
|
||||
- For command execution: Respond with JSON containing action/command/explanation
|
||||
- Keep responses concise but informative
|
||||
|
||||
RULES:
|
||||
- Only execute commands when explicitly asked or when it's clearly needed
|
||||
- Explain what you're about to do before executing
|
||||
- Never execute destructive commands without explicit confirmation
|
||||
- If unsure, ask for clarification
|
||||
{context}
|
||||
|
||||
USER: {user_message}
|
||||
|
||||
MACHA:"""
|
||||
|
||||
return prompt
|
||||
|
||||
def _execute_command(self, command: str) -> Dict[str, Any]:
|
||||
"""Execute a shell command and return results"""
|
||||
try:
|
||||
result = subprocess.run(
|
||||
command,
|
||||
shell=True,
|
||||
capture_output=True,
|
||||
text=True,
|
||||
timeout=30
|
||||
)
|
||||
|
||||
# Check if command failed due to permissions
|
||||
needs_sudo = False
|
||||
permission_errors = [
|
||||
'Interactive authentication required',
|
||||
'Permission denied',
|
||||
'Operation not permitted',
|
||||
'Must be root',
|
||||
'insufficient privileges',
|
||||
'authentication is required'
|
||||
]
|
||||
|
||||
if result.returncode != 0:
|
||||
error_text = (result.stderr + result.stdout).lower()
|
||||
for perm_error in permission_errors:
|
||||
if perm_error.lower() in error_text:
|
||||
needs_sudo = True
|
||||
break
|
||||
|
||||
# Retry with sudo if permission error detected
|
||||
if needs_sudo and not command.strip().startswith('sudo'):
|
||||
print(f"\n⚠️ Permission denied, retrying with sudo...")
|
||||
sudo_command = f"sudo {command}"
|
||||
result = subprocess.run(
|
||||
sudo_command,
|
||||
shell=True,
|
||||
capture_output=True,
|
||||
text=True,
|
||||
timeout=30
|
||||
)
|
||||
|
||||
return {
|
||||
'success': result.returncode == 0,
|
||||
'exit_code': result.returncode,
|
||||
'stdout': result.stdout,
|
||||
'stderr': result.stderr,
|
||||
'command': sudo_command,
|
||||
'retried_with_sudo': True
|
||||
}
|
||||
|
||||
return {
|
||||
'success': result.returncode == 0,
|
||||
'exit_code': result.returncode,
|
||||
'stdout': result.stdout,
|
||||
'stderr': result.stderr,
|
||||
'command': command,
|
||||
'retried_with_sudo': False
|
||||
}
|
||||
except subprocess.TimeoutExpired:
|
||||
return {
|
||||
'success': False,
|
||||
'exit_code': -1,
|
||||
'stdout': '',
|
||||
'stderr': 'Command timed out after 30 seconds',
|
||||
'command': command,
|
||||
'retried_with_sudo': False
|
||||
}
|
||||
except Exception as e:
|
||||
return {
|
||||
'success': False,
|
||||
'exit_code': -1,
|
||||
'stdout': '',
|
||||
'stderr': str(e),
|
||||
'command': command,
|
||||
'retried_with_sudo': False
|
||||
}
|
||||
|
||||
def _parse_response(self, response: str) -> Dict[str, Any]:
|
||||
"""Parse AI response to determine if it's a command or text"""
|
||||
try:
|
||||
# Try to parse as JSON
|
||||
parsed = json.loads(response.strip())
|
||||
if isinstance(parsed, dict) and 'action' in parsed:
|
||||
return parsed
|
||||
except json.JSONDecodeError:
|
||||
pass
|
||||
|
||||
# It's plain text conversation
|
||||
return {'action': 'chat', 'message': response}
|
||||
|
||||
def _auto_diagnose_ollama(self) -> str:
|
||||
"""Automatically diagnose Ollama issues"""
|
||||
diagnostics = []
|
||||
|
||||
diagnostics.append("🔍 AUTO-DIAGNOSIS: Investigating Ollama failure...\n")
|
||||
|
||||
# Check if Ollama service is running
|
||||
try:
|
||||
result = subprocess.run(
|
||||
['systemctl', 'is-active', 'ollama.service'],
|
||||
capture_output=True,
|
||||
text=True,
|
||||
timeout=5
|
||||
)
|
||||
if result.returncode == 0:
|
||||
diagnostics.append("✅ Ollama service is active")
|
||||
else:
|
||||
diagnostics.append(f"❌ Ollama service is NOT active: {result.stdout.strip()}")
|
||||
# Get service status
|
||||
status_result = subprocess.run(
|
||||
['systemctl', 'status', 'ollama.service', '--no-pager', '-l'],
|
||||
capture_output=True,
|
||||
text=True,
|
||||
timeout=5
|
||||
)
|
||||
diagnostics.append(f"\nService status:\n```\n{status_result.stdout[-500:]}\n```")
|
||||
except Exception as e:
|
||||
diagnostics.append(f"⚠️ Could not check service status: {e}")
|
||||
|
||||
# Check memory usage
|
||||
try:
|
||||
result = subprocess.run(['free', '-h'], capture_output=True, text=True, timeout=5)
|
||||
lines = result.stdout.split('\n')
|
||||
for line in lines[:3]: # First 3 lines
|
||||
diagnostics.append(f" {line}")
|
||||
except Exception as e:
|
||||
diagnostics.append(f"⚠️ Could not check memory: {e}")
|
||||
|
||||
# Check which models are loaded
|
||||
try:
|
||||
import requests
|
||||
response = requests.get(f"{self.agent.ollama_host}/api/tags", timeout=5)
|
||||
if response.status_code == 200:
|
||||
models = response.json().get('models', [])
|
||||
diagnostics.append(f"\n📦 Loaded models ({len(models)}):")
|
||||
for model in models:
|
||||
name = model.get('name', 'unknown')
|
||||
size = model.get('size', 0) / (1024**3)
|
||||
is_current = "← TARGET" if name == self.agent.model else ""
|
||||
diagnostics.append(f" • {name} ({size:.1f} GB) {is_current}")
|
||||
|
||||
# Check if target model is loaded
|
||||
model_names = [m.get('name') for m in models]
|
||||
if self.agent.model not in model_names:
|
||||
diagnostics.append(f"\n❌ TARGET MODEL NOT LOADED: {self.agent.model}")
|
||||
diagnostics.append(f" Available models: {', '.join(model_names)}")
|
||||
else:
|
||||
diagnostics.append(f"❌ Ollama API returned {response.status_code}")
|
||||
except Exception as e:
|
||||
diagnostics.append(f"⚠️ Could not query Ollama API: {e}")
|
||||
|
||||
# Check recent Ollama logs
|
||||
try:
|
||||
result = subprocess.run(
|
||||
['journalctl', '-u', 'ollama.service', '-n', '10', '--no-pager'],
|
||||
capture_output=True,
|
||||
text=True,
|
||||
timeout=5
|
||||
)
|
||||
if result.stdout:
|
||||
diagnostics.append(f"\n📋 Recent Ollama logs (last 10 lines):\n```\n{result.stdout}\n```")
|
||||
except Exception as e:
|
||||
diagnostics.append(f"⚠️ Could not check logs: {e}")
|
||||
|
||||
return "\n".join(diagnostics)
|
||||
|
||||
def process_message(self, user_message: str) -> str:
|
||||
"""Process a user message and return Macha's response"""
|
||||
|
||||
# Add user message to history
|
||||
self.conversation_history.append({
|
||||
'role': 'user',
|
||||
'message': user_message,
|
||||
'timestamp': datetime.now().isoformat()
|
||||
})
|
||||
|
||||
# Build chat messages for tool-calling API
|
||||
messages = []
|
||||
|
||||
# Query relevant knowledge based on user message
|
||||
knowledge_context = self.agent._query_relevant_knowledge(user_message, limit=3)
|
||||
|
||||
# Add recent conversation history (last 15 messages to stay within context limits)
|
||||
# With tool calling, messages grow quickly, so we limit more aggressively
|
||||
recent_history = self.conversation_history[-15:] # Last ~7 exchanges
|
||||
for entry in recent_history:
|
||||
content = entry['message']
|
||||
# Truncate very long messages (e.g., command outputs)
|
||||
if len(content) > 3000:
|
||||
content = content[:1500] + "\n... [message truncated] ...\n" + content[-1500:]
|
||||
# Add knowledge context to first user message if available
|
||||
if entry == recent_history[-1] and knowledge_context:
|
||||
content += knowledge_context
|
||||
messages.append({
|
||||
"role": entry['role'],
|
||||
"content": content
|
||||
})
|
||||
|
||||
try:
|
||||
# Use tool-aware chat API
|
||||
ai_response = self.agent._query_ollama_with_tools(messages)
|
||||
except Exception as e:
|
||||
error_msg = (
|
||||
f"❌ CRITICAL: Failed to communicate with Ollama inference engine\n\n"
|
||||
f"Error Type: {type(e).__name__}\n"
|
||||
f"Error Message: {str(e)}\n\n"
|
||||
)
|
||||
# Auto-diagnose the issue
|
||||
diagnostics = self._auto_diagnose_ollama()
|
||||
return error_msg + "\n" + diagnostics
|
||||
|
||||
if not ai_response:
|
||||
error_msg = (
|
||||
f"❌ Empty response from Ollama inference engine\n\n"
|
||||
f"The request succeeded but returned no data. This usually means:\n"
|
||||
f" • The model ({self.agent.model}) is still loading\n"
|
||||
f" • Ollama ran out of memory during generation\n"
|
||||
f" • The prompt was too large for the context window\n\n"
|
||||
)
|
||||
# Auto-diagnose the issue
|
||||
diagnostics = self._auto_diagnose_ollama()
|
||||
return error_msg + "\n" + diagnostics
|
||||
|
||||
# Check if Ollama returned an error
|
||||
try:
|
||||
error_check = json.loads(ai_response)
|
||||
if isinstance(error_check, dict) and 'error' in error_check:
|
||||
error_msg = (
|
||||
f"❌ Ollama API Error\n\n"
|
||||
f"Error: {error_check.get('error', 'Unknown error')}\n"
|
||||
f"Diagnosis: {error_check.get('diagnosis', 'No details')}\n\n"
|
||||
)
|
||||
# Auto-diagnose the issue
|
||||
diagnostics = self._auto_diagnose_ollama()
|
||||
return error_msg + "\n" + diagnostics
|
||||
except json.JSONDecodeError:
|
||||
# Not JSON, it's a normal response
|
||||
pass
|
||||
|
||||
# Parse response
|
||||
parsed = self._parse_response(ai_response)
|
||||
|
||||
if parsed.get('action') == 'execute':
|
||||
# AI wants to execute a command
|
||||
command = parsed.get('command', '')
|
||||
explanation = parsed.get('explanation', '')
|
||||
|
||||
# Show what we're about to do
|
||||
response = f"🔧 {explanation}\n\nExecuting: `{command}`\n\n"
|
||||
|
||||
# Execute the command
|
||||
result = self._execute_command(command)
|
||||
|
||||
# Show if we retried with sudo
|
||||
if result.get('retried_with_sudo'):
|
||||
response += f"⚠️ Permission denied, retried as: `{result['command']}`\n\n"
|
||||
|
||||
if result['success']:
|
||||
response += "✅ Command succeeded:\n"
|
||||
if result['stdout']:
|
||||
response += f"```\n{result['stdout']}\n```"
|
||||
else:
|
||||
response += "(no output)"
|
||||
else:
|
||||
response += f"❌ Command failed (exit code {result['exit_code']}):\n"
|
||||
if result['stderr']:
|
||||
response += f"```\n{result['stderr']}\n```"
|
||||
elif result['stdout']:
|
||||
response += f"```\n{result['stdout']}\n```"
|
||||
|
||||
# Add command execution to history
|
||||
self.conversation_history.append({
|
||||
'role': 'macha',
|
||||
'message': response,
|
||||
'timestamp': datetime.now().isoformat(),
|
||||
'command_result': result
|
||||
})
|
||||
|
||||
# Now ask AI to respond to the command output
|
||||
followup_prompt = f"""The command completed. Here's what happened:
|
||||
|
||||
Command: {command}
|
||||
Success: {result['success']}
|
||||
Output: {result['stdout'][:500] if result['stdout'] else '(none)'}
|
||||
Error: {result['stderr'][:500] if result['stderr'] else '(none)'}
|
||||
|
||||
Please provide a brief analysis or next steps."""
|
||||
|
||||
followup_response = self.agent._query_ollama(followup_prompt)
|
||||
|
||||
if followup_response:
|
||||
response += f"\n\n{followup_response}"
|
||||
|
||||
return response
|
||||
|
||||
else:
|
||||
# Normal conversation response
|
||||
message = parsed.get('message', ai_response)
|
||||
|
||||
self.conversation_history.append({
|
||||
'role': 'macha',
|
||||
'message': message,
|
||||
'timestamp': datetime.now().isoformat()
|
||||
})
|
||||
|
||||
return message
|
||||
|
||||
def run(self):
|
||||
"""Run the interactive chat session"""
|
||||
print("=" * 70)
|
||||
print("🌐 MACHA INTERACTIVE CHAT")
|
||||
print("=" * 70)
|
||||
print("Type your message and press Enter. Commands:")
|
||||
print(" /exit or /quit - End the chat session")
|
||||
print(" /clear - Clear conversation history")
|
||||
print(" /history - Show conversation history")
|
||||
print(" /debug - Show Ollama connection status")
|
||||
print("=" * 70)
|
||||
print()
|
||||
|
||||
while True:
|
||||
try:
|
||||
# Get user input
|
||||
user_input = input("\n💬 YOU: ").strip()
|
||||
|
||||
if not user_input:
|
||||
continue
|
||||
|
||||
# Handle special commands
|
||||
if user_input.lower() in ['/exit', '/quit']:
|
||||
print("\n👋 Ending chat session. Goodbye!")
|
||||
break
|
||||
|
||||
elif user_input.lower() == '/clear':
|
||||
self.conversation_history.clear()
|
||||
print("🧹 Conversation history cleared.")
|
||||
continue
|
||||
|
||||
elif user_input.lower() == '/history':
|
||||
print("\n" + "=" * 70)
|
||||
print("CONVERSATION HISTORY")
|
||||
print("=" * 70)
|
||||
for entry in self.conversation_history:
|
||||
role = entry['role'].upper()
|
||||
msg = entry['message'][:100] + "..." if len(entry['message']) > 100 else entry['message']
|
||||
print(f"{role}: {msg}")
|
||||
print("=" * 70)
|
||||
continue
|
||||
|
||||
elif user_input.lower() == '/debug':
|
||||
import os
|
||||
import subprocess
|
||||
|
||||
print("\n" + "=" * 70)
|
||||
print("MACHA ARCHITECTURE & STATUS")
|
||||
print("=" * 70)
|
||||
|
||||
print("\n🏗️ SYSTEM ARCHITECTURE:")
|
||||
print(f" Hostname: macha.coven.systems")
|
||||
print(f" Service: macha-autonomous.service (systemd)")
|
||||
print(f" Working Directory: /var/lib/macha")
|
||||
|
||||
print("\n👤 EXECUTION CONTEXT:")
|
||||
current_user = os.getenv('USER') or os.getenv('USERNAME') or 'unknown'
|
||||
print(f" Current User: {current_user}")
|
||||
print(f" UID: {os.getuid()}")
|
||||
|
||||
# Check if user has sudo access
|
||||
try:
|
||||
result = subprocess.run(['sudo', '-n', 'true'],
|
||||
capture_output=True, timeout=1)
|
||||
if result.returncode == 0:
|
||||
print(f" Sudo Access: ✓ Yes (passwordless)")
|
||||
else:
|
||||
print(f" Sudo Access: ⚠ Requires password")
|
||||
except:
|
||||
print(f" Sudo Access: ❌ No")
|
||||
|
||||
print(f" Note: Chat runs as invoking user (you), not as macha-autonomous")
|
||||
|
||||
print("\n🧠 INFERENCE ENGINE:")
|
||||
print(f" Backend: Ollama")
|
||||
print(f" Host: {self.agent.ollama_host}")
|
||||
print(f" Model: {self.agent.model}")
|
||||
print(f" Service: ollama.service (systemd)")
|
||||
|
||||
print("\n💾 DATABASE:")
|
||||
print(f" Backend: ChromaDB")
|
||||
print(f" Host: http://localhost:8000")
|
||||
print(f" Data: /var/lib/chromadb")
|
||||
print(f" Service: chromadb.service (systemd)")
|
||||
|
||||
print("\n🔍 OLLAMA STATUS:")
|
||||
# Try to query Ollama status
|
||||
try:
|
||||
import requests
|
||||
# Check if Ollama is running
|
||||
response = requests.get(f"{self.agent.ollama_host}/api/tags", timeout=5)
|
||||
if response.status_code == 200:
|
||||
models = response.json().get('models', [])
|
||||
print(f" Status: ✓ Running")
|
||||
print(f" Loaded models: {len(models)}")
|
||||
for model in models:
|
||||
name = model.get('name', 'unknown')
|
||||
size = model.get('size', 0) / (1024**3) # GB
|
||||
is_current = "← ACTIVE" if name == self.agent.model else ""
|
||||
print(f" • {name} ({size:.1f} GB) {is_current}")
|
||||
else:
|
||||
print(f" Status: ❌ Error (HTTP {response.status_code})")
|
||||
except Exception as e:
|
||||
print(f" Status: ❌ Cannot connect: {e}")
|
||||
print(f" Hint: Check 'systemctl status ollama.service'")
|
||||
|
||||
print("\n💡 CONVERSATION:")
|
||||
print(f" History: {len(self.conversation_history)} messages")
|
||||
print(f" Session started: {self.session_start}")
|
||||
|
||||
print("=" * 70)
|
||||
continue
|
||||
|
||||
# Process the message
|
||||
print("\n🤖 MACHA: ", end='', flush=True)
|
||||
response = self.process_message(user_input)
|
||||
print(response)
|
||||
|
||||
except KeyboardInterrupt:
|
||||
print("\n\n👋 Chat interrupted. Use /exit to quit properly.")
|
||||
continue
|
||||
except EOFError:
|
||||
print("\n\n👋 Ending chat session. Goodbye!")
|
||||
break
|
||||
except Exception as e:
|
||||
print(f"\n❌ Error: {e}")
|
||||
continue
|
||||
|
||||
|
||||
def main():
|
||||
"""Main entry point"""
|
||||
session = MachaChatSession()
|
||||
session.run()
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
|
||||
Reference in New Issue
Block a user