setInput(e.target.value)}
+ onChange={e => {
+ const val = e.target.value;
+ setInput(val);
+ setIsConfused(val.includes('???'));
+ setIsHappy(val.toLowerCase().includes('thanks') || val.toLowerCase().includes('thx'));
+ setTimeout(updateCaret, 0);
+ }}
onKeyPress={e => e.key === 'Enter' && sendMessage()}
+ onKeyUp={updateCaret}
+ onClick={updateCaret}
+ onFocus={updateCaret}
+ onBlur={() => setCaretPos(null)}
placeholder="Ask BuddAI to build something..."
autoFocus
/>
diff --git a/buddai_v3.2.py b/main.py
similarity index 78%
rename from buddai_v3.2.py
rename to main.py
index 3a2ae03..c1f83e5 100644
--- a/buddai_v3.2.py
+++ b/main.py
@@ -11,6 +11,7 @@ License: MIT
import sys
import os
import json
+import logging
import sqlite3
from datetime import datetime
from pathlib import Path
@@ -22,6 +23,13 @@ import shutil
import queue
import socket
import argparse
+import io
+from urllib.parse import urlparse
+
+try:
+ import qrcode
+except ImportError:
+ qrcode = None
try:
import psutil
@@ -30,10 +38,10 @@ except ImportError:
# Server dependencies
try:
- from fastapi import FastAPI, UploadFile, File, Header, WebSocket, WebSocketDisconnect
+ from fastapi import FastAPI, UploadFile, File, Header, WebSocket, WebSocketDisconnect, Request
from fastapi.middleware.cors import CORSMiddleware
from fastapi.staticfiles import StaticFiles
- from fastapi.responses import FileResponse, HTMLResponse
+ from fastapi.responses import FileResponse, HTMLResponse, JSONResponse, Response
from pydantic import BaseModel
import uvicorn
SERVER_AVAILABLE = True
@@ -41,7 +49,7 @@ except ImportError:
SERVER_AVAILABLE = False
# Configuration
-OLLAMA_HOST = os.getenv("OLLAMA_HOST", "localhost")
+OLLAMA_HOST = os.getenv("OLLAMA_HOST", "127.0.0.1")
OLLAMA_PORT = int(os.getenv("OLLAMA_PORT", "11434"))
DATA_DIR = Path(__file__).parent / "data"
DB_PATH = DATA_DIR / "conversations.db"
@@ -592,7 +600,8 @@ class BuddAI:
simple_triggers = [
"what is", "what's", "who is", "who's", "when is",
"how do i", "can you explain", "tell me about",
- "what are", "where is"
+ "what are", "where is", "hi", "hello", "hey",
+ "good morning", "good evening"
]
# Also check if it's just a question without code keywords
@@ -684,7 +693,6 @@ class BuddAI:
def call_model(self, model_name: str, message: str, stream: bool = False) -> Union[str, Generator[str, None, None]]:
"""Call specified model"""
- conn = None
try:
current_time = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
user_context = self.get_user_status()
@@ -734,35 +742,92 @@ float applyForge(float current, float target, float k) {{ return target + (curre
"model": MODELS[model_name],
"messages": messages,
"stream": stream,
- "options": {"temperature": 0.7, "num_ctx": 4096}
+ "options": {"temperature": 0.7, "num_ctx": 1024} # Default options
}
- conn = OLLAMA_POOL.get_connection()
headers = {"Content-Type": "application/json"}
json_body = json.dumps(body)
- conn.request("POST", "/api/chat", json_body, headers)
- response = conn.getresponse()
-
- if stream:
- return self._stream_response(response, conn)
-
- if response.status == 200:
- data = json.loads(response.read().decode('utf-8'))
- OLLAMA_POOL.return_connection(conn)
- return data.get("message", {}).get("content", "No response")
- else:
- conn.close()
- return f"Error: {response.status}"
-
+ # Retry logic for connection stability
+ # Attempts: 0=Normal, 1=Retry/CPU Fallback, 2=Final Retry
+ for attempt in range(3):
+ conn = None
+ try:
+ # Re-serialize body in case options changed (CPU fallback)
+ json_body = json.dumps(body)
+
+ conn = OLLAMA_POOL.get_connection()
+ conn.request("POST", "/api/chat", json_body, headers)
+ response = conn.getresponse()
+
+ if stream:
+ if response.status != 200:
+ error_text = response.read().decode('utf-8')
+ conn.close()
+
+ # GPU OOM Detection -> CPU Fallback
+ if "CUDA" in error_text or "buffer" in error_text:
+ if "num_gpu" not in body["options"]:
+ print("⚠️ GPU OOM detected. Switching to CPU mode...")
+ body["options"]["num_gpu"] = 0 # Force CPU
+ continue # Retry immediately
+
+ try:
+ err_msg = f"Error {response.status}: {json.loads(error_text).get('error', error_text)}"
+ except:
+ err_msg = f"Error {response.status}: {error_text}"
+
+ if "num_gpu" in body["options"]:
+ err_msg += "\n\n(⚠️ CPU Mode also failed. System RAM might be full.)"
+ elif "CUDA" in err_msg or "buffer" in err_msg:
+ err_msg += "\n\n(⚠️ GPU Out of Memory. Retrying on CPU failed.)"
+
+ return (x for x in [err_msg])
+
+ return self._stream_response(response, conn)
+
+ if response.status == 200:
+ data = json.loads(response.read().decode('utf-8'))
+ OLLAMA_POOL.return_connection(conn)
+ return data.get("message", {}).get("content", "No response")
+ else:
+ error_text = response.read().decode('utf-8')
+ conn.close()
+
+ # GPU OOM Detection -> CPU Fallback (Non-stream)
+ if "CUDA" in error_text or "buffer" in error_text:
+ if "num_gpu" not in body["options"]:
+ print("⚠️ GPU OOM detected. Switching to CPU mode...")
+ body["options"]["num_gpu"] = 0 # Force CPU
+ continue # Retry immediately
+
+ try:
+ err_msg = f"Error {response.status}: {json.loads(error_text).get('error', error_text)}"
+ except:
+ err_msg = f"Error {response.status}: {error_text}"
+
+ if "num_gpu" in body["options"]:
+ err_msg += "\n\n(⚠️ CPU Mode also failed.)"
+ elif "CUDA" in err_msg or "buffer" in err_msg:
+ err_msg += "\n\n(⚠️ GPU Out of Memory.)"
+ return err_msg
+
+ except (http.client.NotConnected, BrokenPipeError, ConnectionResetError, socket.timeout) as e:
+ if conn: conn.close()
+ if attempt == 2: # Last attempt
+ return f"Error: Connection failed. {str(e)}"
+ continue # Retry
+ except Exception as e:
+ if conn: conn.close()
+ return f"Error: {str(e)}"
+
except Exception as e:
- if conn:
- conn.close()
return f"Error: {str(e)}"
def _stream_response(self, response, conn) -> Generator[str, None, None]:
"""Yield chunks from HTTP response"""
fully_consumed = False
+ has_content = False
try:
while True:
line = response.readline()
@@ -771,16 +836,23 @@ float applyForge(float current, float target, float k) {{ return target + (curre
data = json.loads(line.decode('utf-8'))
if "message" in data:
content = data["message"].get("content", "")
- if content: yield content
+ if content:
+ has_content = True
+ yield content
if data.get("done"):
fully_consumed = True
break
except: pass
+ except Exception as e:
+ yield f"\n[Stream Error: {str(e)}]"
finally:
if fully_consumed:
OLLAMA_POOL.return_connection(conn)
else:
conn.close()
+
+ if not has_content and not fully_consumed:
+ yield "\n[Error: Empty response from Ollama. Check if model is loaded.]"
def execute_modular_build(self, _: str, modules: List[str], plan: List[Dict[str, str]], forge_mode: str = "2") -> str:
"""Execute build plan step by step"""
@@ -1006,6 +1078,15 @@ float applyForge(float current, float target, float k) {{ return target + (curre
conn.commit()
conn.close()
+ def clear_current_session(self) -> None:
+ """Clear all messages from the current session"""
+ conn = sqlite3.connect(DB_PATH)
+ cursor = conn.cursor()
+ cursor.execute("DELETE FROM messages WHERE session_id = ?", (self.session_id,))
+ conn.commit()
+ conn.close()
+ self.context_messages = []
+
def load_session(self, session_id: str) -> List[Dict[str, str]]:
"""Load a specific session context"""
conn = sqlite3.connect(DB_PATH)
@@ -1035,6 +1116,113 @@ float applyForge(float current, float target, float k) {{ return target + (curre
self.context_messages = []
return self.session_id
+ def reset_gpu(self) -> str:
+ """Force unload models from GPU to free VRAM"""
+ try:
+ conn = http.client.HTTPConnection(OLLAMA_HOST, OLLAMA_PORT, timeout=10)
+ # Unload all known models
+ for model in MODELS.values():
+ body = json.dumps({"model": model, "keep_alive": 0})
+ conn.request("POST", "/api/generate", body)
+ resp = conn.getresponse()
+ resp.read() # Consume response
+ conn.close()
+ return "✅ GPU Memory Cleared (Models Unloaded)"
+ except Exception as e:
+ return f"❌ Error clearing GPU: {str(e)}"
+
+ def export_session_to_markdown(self, session_id: str = None) -> str:
+ """Export session history to a Markdown file"""
+ sid = session_id or self.session_id
+
+ conn = sqlite3.connect(DB_PATH)
+ cursor = conn.cursor()
+ cursor.execute("SELECT role, content, timestamp FROM messages WHERE session_id = ? ORDER BY id ASC", (sid,))
+ rows = cursor.fetchall()
+ conn.close()
+
+ if not rows:
+ return "No history found."
+
+ filename = f"session_{sid}.md"
+ filepath = DATA_DIR / filename
+
+ with open(filepath, "w", encoding="utf-8") as f:
+ f.write(f"# BuddAI Session: {sid}\n\n")
+ for role, content, ts in rows:
+ f.write(f"### {role.upper()} ({ts})\n\n{content}\n\n---\n\n")
+
+ return f"✅ Session exported to: {filepath}"
+
+ def get_session_export_data(self, session_id: str = None) -> Dict:
+ """Get session data as a dictionary for export"""
+ sid = session_id or self.session_id
+ conn = sqlite3.connect(DB_PATH)
+ cursor = conn.cursor()
+ cursor.execute("SELECT role, content, timestamp FROM messages WHERE session_id = ? ORDER BY id ASC", (sid,))
+ rows = cursor.fetchall()
+ conn.close()
+
+ return {
+ "session_id": sid,
+ "exported_at": datetime.now().isoformat(),
+ "messages": [{"role": r, "content": c, "timestamp": t} for r, c, t in rows]
+ }
+
+ def export_session_to_json(self, session_id: str = None) -> str:
+ """Export session history to a JSON file"""
+ data = self.get_session_export_data(session_id)
+ if not data["messages"]:
+ return "No history found."
+
+ filename = f"session_{data['session_id']}.json"
+ filepath = DATA_DIR / filename
+
+ with open(filepath, "w", encoding="utf-8") as f:
+ json.dump(data, f, indent=2, ensure_ascii=False)
+
+ return f"✅ Session exported to: {filepath}"
+
+ def import_session_from_json(self, data: Dict) -> str:
+ """Import a session from JSON data"""
+ session_id = data.get("session_id")
+ messages = data.get("messages", [])
+
+ if not session_id or not messages:
+ raise ValueError("Invalid session JSON format")
+
+ conn = sqlite3.connect(DB_PATH)
+ cursor = conn.cursor()
+
+ # Check if session exists to avoid collision
+ cursor.execute("SELECT 1 FROM sessions WHERE session_id = ? AND user_id = ?", (session_id, self.user_id))
+ if cursor.fetchone():
+ # Generate new ID
+ timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+ session_id = f"{session_id}_imp_{timestamp}"
+
+ # Determine start time
+ started_at = datetime.now().isoformat()
+ if messages and "timestamp" in messages[0]:
+ started_at = messages[0]["timestamp"]
+
+ cursor.execute(
+ "INSERT INTO sessions (session_id, user_id, started_at, title) VALUES (?, ?, ?, ?)",
+ (session_id, self.user_id, started_at, f"Imported: {data.get('session_id')}")
+ )
+
+ # Insert messages
+ for msg in messages:
+ cursor.execute(
+ "INSERT INTO messages (session_id, role, content, timestamp) VALUES (?, ?, ?, ?)",
+ (session_id, msg.get("role"), msg.get("content"), msg.get("timestamp", datetime.now().isoformat()))
+ )
+
+ conn.commit()
+ conn.close()
+
+ return session_id
+
def run(self) -> None:
"""Main loop"""
try:
@@ -1063,6 +1251,7 @@ float applyForge(float current, float target, float k) {{ return target + (curre
print("/balanced - Use balanced model")
print("/index
- Index local repositories")
print("/scan - Scan style signature (V3.0)")
+ print("/save - Export chat to Markdown")
print("/help - This message")
print("exit - End session\n")
continue
@@ -1076,6 +1265,12 @@ float applyForge(float current, float target, float k) {{ return target + (curre
elif cmd == '/scan':
self.scan_style_signature()
continue
+ elif cmd.startswith('/save'):
+ if 'json' in user_input.lower():
+ print(self.export_session_to_json())
+ else:
+ print(self.export_session_to_markdown())
+ continue
else:
print("\nUnknown command. Type /help")
continue
@@ -1118,6 +1313,9 @@ if SERVER_AVAILABLE:
class FeedbackRequest(BaseModel):
message_id: int
positive: bool
+
+ class ResetGpuRequest(BaseModel):
+ pass
# Multi-user support
class BuddAIManager:
@@ -1137,10 +1335,45 @@ if SERVER_AVAILABLE:
app.mount("/web", StaticFiles(directory=frontend_path, html=True), name="web")
@app.get("/", response_class=HTMLResponse)
- async def root():
+ async def root(request: Request):
server_buddai = buddai_manager.get_instance("default")
status = server_buddai.get_user_status()
+ public_url = getattr(request.app.state, "public_url", "")
+ qr_section = ""
+ ip_section = ""
+
+ if public_url:
+ parsed = urlparse(public_url)
+ host = parsed.hostname
+ label = "Server Address"
+ color = "#fff"
+
+ if host:
+ if host.startswith("100."):
+ label = "Tailscale IP"
+ color = "#ff79c6" # Magenta
+ elif host.startswith("192.168.") or host.startswith("10.") or host.startswith("172."):
+ label = "LAN IP"
+ color = "#50fa7b" # Green
+ elif "ngrok" in public_url:
+ label = "Public Tunnel"
+ color = "#8be9fd" # Cyan
+
+ ip_section = f"""
+
+ """
+
+ qr_section = f"""
+
+
Scan to Connect
+
+
+