gradio.chat.app-HFIPs / chat_handler.py
ysharma's picture
ysharma HF Staff
Update chat_handler.py
87bbbcd verified
raw
history blame
25 kB
"""
Chat handling logic for Universal MCP Client - Enhanced with Inference Provider Support
"""
import re
import logging
import traceback
import asyncio
from datetime import datetime
from typing import Dict, Any, List, Tuple, Optional
import gradio as gr
from gradio import ChatMessage
import time
from config import AppConfig
from mcp_client import UniversalMCPClient
logger = logging.getLogger(__name__)
class ChatHandler:
"""Handles chat interactions with multiple LLM backends and MCP servers using ChatMessage dataclass"""
def __init__(self, mcp_client: UniversalMCPClient):
self.mcp_client = mcp_client
def process_multimodal_message(self, message: Dict[str, Any], history: List) -> Tuple[List[ChatMessage], Dict[str, Any]]:
"""Enhanced MCP chat function with multimodal input support and multiple LLM backends"""
# Check if any LLM backend is configured
backend_configured = False
if self.mcp_client.anthropic_client and AppConfig.ANTHROPIC_API_KEY:
backend_configured = True
backend_type = "anthropic"
elif self.mcp_client.hf_client and self.mcp_client.current_provider:
backend_configured = True
backend_type = "hf_inference"
if not backend_configured:
error_msg = "❌ No LLM backend configured. Please configure either Anthropic API key or HuggingFace Inference Provider."
history.append(ChatMessage(role="user", content=error_msg))
history.append(ChatMessage(role="assistant", content=error_msg))
return history, gr.MultimodalTextbox(value=None, interactive=False)
# Initialize variables for error handling
user_text = ""
user_files = []
try:
# Handle multimodal input - message is a dict with 'text' and 'files'
user_text = message.get("text", "") if message else ""
user_files = message.get("files", []) if message else []
# Handle case where message might be a string (backward compatibility)
if isinstance(message, str):
user_text = message
user_files = []
logger.info(f"💬 Processing multimodal message with {backend_type} backend:")
logger.info(f" 📝 Text: {user_text}")
logger.info(f" 📁 Files: {len(user_files)} files uploaded")
logger.info(f" 📋 History type: {type(history)}, length: {len(history)}")
# Convert history to ChatMessage objects if needed
converted_history = []
for i, msg in enumerate(history):
try:
if isinstance(msg, dict):
# Convert dict to ChatMessage for internal processing
logger.info(f" 📝 Converting dict message {i}: {msg.get('role', 'unknown')}")
converted_history.append(ChatMessage(
role=msg.get('role', 'assistant'),
content=msg.get('content', ''),
metadata=msg.get('metadata', None)
))
else:
# Already a ChatMessage
logger.info(f" ✅ ChatMessage {i}: {getattr(msg, 'role', 'unknown')}")
converted_history.append(msg)
except Exception as conv_error:
logger.error(f"Error converting message {i}: {conv_error}")
logger.error(f"Message content: {msg}")
# Skip problematic messages
continue
history = converted_history
# Add uploaded files to chat history first
for file_path in user_files:
logger.info(f" 📄 File: {file_path}")
history.append(ChatMessage(role="user", content={"path": file_path}))
# Add text message if provided
if user_text and user_text.strip():
history.append(ChatMessage(role="user", content=user_text))
# If no text and no files, return early
if not user_text.strip() and not user_files:
return history, gr.MultimodalTextbox(value=None, interactive=False)
# Create messages for LLM API
messages = self._prepare_llm_messages(history)
# Process the chat based on backend type
if backend_type == "anthropic":
response_messages = self._call_anthropic_api(messages, user_files)
else: # hf_inference
response_messages = self._call_hf_inference_api(messages, user_files)
# Add all response messages to history
history.extend(response_messages)
return history, gr.MultimodalTextbox(value=None, interactive=False)
except Exception as e:
error_msg = f"❌ Error: {str(e)}"
logger.error(f"Chat error: {e}")
logger.error(traceback.format_exc())
# Add user input to history if it exists
if user_text and user_text.strip():
history.append(ChatMessage(role="user", content=user_text))
if user_files:
for file_path in user_files:
history.append(ChatMessage(role="user", content={"path": file_path}))
history.append(ChatMessage(role="assistant", content=error_msg))
return history, gr.MultimodalTextbox(value=None, interactive=False)
def _prepare_llm_messages(self, history: List) -> List[Dict[str, Any]]:
"""Convert history (ChatMessage or dict) to LLM API format"""
messages = []
# Convert history to LLM API format (text only for context)
recent_history = history[-16:] if len(history) > 16 else history
for msg in recent_history:
# Handle both ChatMessage objects and dictionary format for backward compatibility
if hasattr(msg, 'role'): # ChatMessage object
role = msg.role
content = msg.content
elif isinstance(msg, dict) and 'role' in msg: # Dictionary format
role = msg.get('role')
content = msg.get('content')
else:
continue # Skip invalid messages
if role in ["user", "assistant"]:
# Convert any non-string content to string description for context
if isinstance(content, dict):
if "path" in content:
file_path = content.get('path', 'unknown')
# Determine file type for context
if AppConfig.is_image_file(file_path):
content = f"[User uploaded an image: {file_path}]"
elif AppConfig.is_audio_file(file_path):
content = f"[User uploaded an audio file: {file_path}]"
elif AppConfig.is_video_file(file_path):
content = f"[User uploaded a video file: {file_path}]"
else:
content = f"[User uploaded a file: {file_path}]"
else:
content = f"[Object: {str(content)[:50]}...]"
elif isinstance(content, (list, tuple)):
content = f"[List: {str(content)[:50]}...]"
elif content is None:
content = "[Empty]"
else:
content = str(content)
messages.append({
"role": role,
"content": content
})
return messages
def _call_anthropic_api(self, messages: List[Dict[str, Any]], user_files: List[str]) -> List[ChatMessage]:
"""Call Anthropic API (existing implementation)"""
# Check if we have MCP servers to use
if not self.mcp_client.servers:
return self._call_claude_without_mcp(messages)
else:
return self._call_claude_with_mcp(messages, user_files)
def _call_hf_inference_api(self, messages: List[Dict[str, Any]], user_files: List[str]) -> List[ChatMessage]:
"""Call HuggingFace Inference API with custom MCP implementation"""
# Run async call in sync context
def run_async():
loop = asyncio.new_event_loop()
asyncio.set_event_loop(loop)
try:
return loop.run_until_complete(
self.mcp_client.call_llm_with_mcp(messages, user_files)
)
finally:
loop.close()
try:
return run_async()
except Exception as e:
logger.error(f"HF Inference API error: {e}")
return [ChatMessage(role="assistant", content=f"❌ Error with HF Inference: {str(e)}")]
def _call_claude_without_mcp(self, messages: List[Dict[str, Any]]) -> List[ChatMessage]:
"""Call Claude API without MCP servers"""
logger.info("💬 No MCP servers available, using regular Claude chat")
system_prompt = self._get_native_system_prompt()
# Use regular messages API
response = self.mcp_client.anthropic_client.messages.create(
model=AppConfig.CLAUDE_MODEL,
max_tokens=AppConfig.MAX_TOKENS,
system=system_prompt,
messages=messages
)
response_text = ""
for content in response.content:
if content.type == "text":
response_text += content.text
if not response_text:
response_text = "I understand your request and I'm here to help."
return [ChatMessage(role="assistant", content=response_text)]
def _call_claude_with_mcp(self, messages: List[Dict[str, Any]], user_files: List[str]) -> List[ChatMessage]:
"""Call Claude API with MCP servers and return structured responses"""
mcp_servers = []
for server_name, config in self.mcp_client.servers.items():
mcp_servers.append({
"type": "url",
"url": config.url,
"name": server_name.replace(" ", "_").lower()
})
# Enhanced system prompt with multimodal and MCP instructions
system_prompt = self._get_mcp_system_prompt(user_files)
# Debug logging
logger.info(f"📤 Sending {len(messages)} messages to Claude API")
logger.info(f"🔧 Using {len(mcp_servers)} MCP servers")
start_time = time.time()
# Call Claude with MCP connector using the correct beta API
response = self.mcp_client.anthropic_client.beta.messages.create(
model=AppConfig.CLAUDE_MODEL,
max_tokens=AppConfig.MAX_TOKENS,
system=system_prompt,
messages=messages,
mcp_servers=mcp_servers,
betas=[AppConfig.MCP_BETA_VERSION]
)
return self._process_mcp_response(response, start_time)
def _process_mcp_response(self, response, start_time: float) -> List[ChatMessage]:
"""Process Claude's response with MCP tool calls into structured ChatMessage objects"""
chat_messages = []
current_tool_id = None
current_server_name = None
tool_start_time = None
text_segments = [] # Collect text segments separately
# Process Claude's response
for content in response.content:
if content.type == "text":
# Collect text segments but don't combine them yet
text_content = content.text
# Check if Claude indicated media was generated
if "MEDIA_GENERATED:" in text_content:
media_match = re.search(r"MEDIA_GENERATED:\s*([^\s]+)", text_content)
if media_match:
media_url = media_match.group(1)
# Clean up the response text
text_content = re.sub(r"MEDIA_GENERATED:\s*[^\s]+", "", text_content).strip()
logger.info(f"🎯 Claude indicated media generated: {media_url}")
# Add media as separate message
chat_messages.append(ChatMessage(
role="assistant",
content={"path": media_url}
))
if text_content.strip():
text_segments.append(text_content.strip())
elif hasattr(content, 'type') and content.type == "mcp_tool_use":
# Add any accumulated text before tool use
if text_segments:
combined_text = " ".join(text_segments)
if combined_text.strip():
chat_messages.append(ChatMessage(
role="assistant",
content=combined_text.strip()
))
text_segments = [] # Reset
tool_name = content.name
server_name = content.server_name
current_tool_id = getattr(content, 'id', 'unknown')
current_server_name = server_name
tool_start_time = time.time()
logger.info(f"🔧 Claude used MCP tool: {tool_name} on server: {server_name}")
# Create a "thinking" message for tool usage
chat_messages.append(ChatMessage(
role="assistant",
content="",
metadata={
"title": f"🔧 Using {tool_name}",
"id": current_tool_id,
"status": "pending",
"log": f"Server: {server_name}"
}
))
elif hasattr(content, 'type') and content.type == "mcp_tool_result":
tool_use_id = getattr(content, 'tool_use_id', 'unknown')
duration = time.time() - tool_start_time if tool_start_time else None
logger.info(f"📝 Processing MCP tool result (tool_use_id: {tool_use_id})")
# Update the pending tool message to completed
for msg in chat_messages:
if (msg.metadata and
msg.metadata.get("id") == current_tool_id and
msg.metadata.get("status") == "pending"):
msg.metadata["status"] = "done"
if duration:
msg.metadata["duration"] = round(duration, 2)
break
media_url = None
if content.content:
result_content = content.content[0]
result_text = result_content.text if hasattr(result_content, 'text') else str(result_content)
logger.info(f"📝 MCP tool result: {result_text[:200]}...")
# Try to extract media URL from the result
if current_server_name and current_server_name in self.mcp_client.servers:
config = self.mcp_client.servers[current_server_name]
extracted_media = self.mcp_client._extract_media_from_mcp_response(result_text, config)
if extracted_media:
media_url = extracted_media
logger.info(f"🎯 Extracted media from MCP result: {media_url}")
else:
# Fallback: try all servers to find media
for server_name, config in self.mcp_client.servers.items():
extracted_media = self.mcp_client._extract_media_from_mcp_response(result_text, config)
if extracted_media:
media_url = extracted_media
logger.info(f"🎯 Extracted media from MCP result (fallback): {media_url}")
break
# Always show the full tool result
chat_messages.append(ChatMessage(
role="assistant",
content=result_text,
metadata={
"title": "📋 Tool Result",
"parent_id": current_tool_id,
"status": "done"
}
))
# Only add separate media display if the tool result does NOT contain
# any Gradio file data structures that would be auto-rendered
if media_url and not self._contains_gradio_file_structure(result_text):
logger.info(f"🎯 Adding separate media display for: {media_url}")
chat_messages.append(ChatMessage(
role="assistant",
content={"path": media_url}
))
else:
if media_url:
logger.info(f"🚫 Skipping separate media - tool result contains Gradio file structure")
else:
logger.info(f"🚫 No media URL extracted")
else:
# Add error message for failed tool call
chat_messages.append(ChatMessage(
role="assistant",
content="Tool call failed: No content returned",
metadata={
"title": "❌ Tool Error",
"parent_id": current_tool_id,
"status": "done"
}
))
# Add any remaining text segments after all processing
if text_segments:
combined_text = " ".join(text_segments)
if combined_text.strip():
chat_messages.append(ChatMessage(
role="assistant",
content=combined_text.strip()
))
# Fallback if no content was processed
if not chat_messages:
chat_messages.append(ChatMessage(
role="assistant",
content="I understand your request and I'm here to help."
))
return chat_messages
def _contains_gradio_file_structure(self, text: str) -> bool:
"""Check if the text contains ANY Gradio file data structures that would be auto-rendered"""
# Check for key indicators of Gradio file structures
gradio_indicators = [
# Gradio FileData type indicators
"'_type': 'gradio.FileData'",
'"_type": "gradio.FileData"',
'gradio.FileData',
# File structure patterns
"'path':",
'"path":',
"'url':",
'"url":',
"'orig_name':",
'"orig_name":',
"'mime_type':",
'"mime_type":',
'is_stream',
'meta_type',
# Common file result patterns
"{'image':",
'{"image":',
"{'audio':",
'{"audio":',
"{'video':",
'{"video":',
"{'file':",
'{"file":',
# List patterns that typically contain file objects
"[{'image'",
'[{"image"',
"[{'audio'",
'[{"audio"',
"[{'video'",
'[{"video"',
"[{'file'",
'[{"file"'
]
# If we find multiple indicators, it's likely a Gradio file structure
indicator_count = sum(1 for indicator in gradio_indicators if indicator in text)
# Also check for simple URL patterns (for audio case)
is_simple_url = (text.strip().startswith('http') and
len(text.strip().split()) == 1 and
any(ext in text.lower() for ext in ['.wav', '.mp3', '.mp4', '.png', '.jpg', '.jpeg', '.gif', '.svg', '.webm', '.ogg']))
result = indicator_count >= 2 or is_simple_url
logger.debug(f"📋 File structure check: {indicator_count} indicators, simple_url: {is_simple_url}, result: {result}")
return result
def _get_native_system_prompt(self) -> str:
"""Get system prompt for Claude without MCP servers"""
return f"""You are Claude Sonnet 4, a helpful AI assistant with native multimodal capabilities. You can have conversations, answer questions, help with various tasks, and provide information on a wide range of topics.
YOUR NATIVE CAPABILITIES (Available right now):
- **Image Understanding**: You can directly see and describe images, analyze their content, read text in images, identify objects, people, scenes, etc.
- **Text Processing**: You can analyze, summarize, translate, and process text directly
- **General Knowledge**: You can answer questions, explain concepts, and have conversations
- **Code Analysis**: You can read, analyze, and explain code
Current time: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}
IMPORTANT: You DO NOT need MCP servers for:
- Describing or analyzing uploaded images
- Reading text in images
- Identifying objects, people, or scenes in images
- General conversation and knowledge questions
You DO need MCP servers for:
- Creating new images, audio, or video
- Editing or transforming existing media files
- Transcribing audio files
- Processing non-image files (audio, video, documents)
If users upload images and ask you to describe or analyze them, use your native vision capabilities immediately. Only mention MCP servers if they ask for creation or editing tasks."""
def _get_mcp_system_prompt(self, user_files: List[str]) -> str:
"""Get system prompt for Claude with MCP servers"""
uploaded_files_context = ""
if user_files:
uploaded_files_context = f"\n\nFILES UPLOADED BY USER:\n"
for i, file_path in enumerate(user_files, 1):
file_name = file_path.split('/')[-1] if '/' in file_path else file_path
if AppConfig.is_image_file(file_path):
file_type = "Image"
elif AppConfig.is_audio_file(file_path):
file_type = "Audio"
elif AppConfig.is_video_file(file_path):
file_type = "Video"
else:
file_type = "File"
uploaded_files_context += f"{i}. {file_type}: {file_name} (path: {file_path})\n"
return f"""You are Claude Sonnet 4, a helpful AI assistant with both native multimodal capabilities and access to various MCP tools.
YOUR NATIVE CAPABILITIES (No MCP tools needed):
- **Image Understanding**: You can directly see and describe images, analyze their content, read text in images, etc.
- **Text Processing**: You can analyze, summarize, translate, and process text directly
- **General Knowledge**: You can answer questions, explain concepts, and have conversations
- **Code Analysis**: You can read, analyze, and explain code
WHEN TO USE MCP TOOLS:
- **Image Generation**: Creating new images from text prompts
- **Image Editing**: Modifying, enhancing, or transforming existing images
- **Audio Processing**: Transcribing audio, generating speech, audio enhancement
- **Video Processing**: Creating or editing videos
- **Specialized Analysis**: Tasks requiring specific models or APIs
UPLOADED FILES HANDLING:
{uploaded_files_context}
IMPORTANT - For uploaded images:
- **Image Description/Analysis**: Use your NATIVE vision capabilities - you can see and describe images directly
- **Image Editing/Enhancement**: Use MCP image processing tools
- **Image Generation**: Use MCP image generation tools
IMPORTANT - GRADIO MEDIA DISPLAY:
When MCP tools return media, end your response with "MEDIA_GENERATED: [URL]" where [URL] is the actual media URL.
Current time: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}
Available MCP servers: {list(self.mcp_client.servers.keys())}"""