import logging
from mcp.types import Tool, TextContent
from gmail_client import GmailClient
from google_docs_helper import GoogleDocsHelper
from anthropic import Anthropic
from helpers.prompt_builder import fetch_guidelines, build_reply_prompt
logger = logging.getLogger("gmail-assistant")
def get_tool_definition() -> Tool:
"""Return the tool definition for MCP."""
return Tool(
name="create_draft_reply",
description="Generate an AI-powered draft reply to an email and save it in Gmail",
inputSchema={
"type": "object",
"properties": {
"email_id": {
"type": "string",
"description": "The ID of the email to reply to"
},
"email_content": {
"type": "string",
"description": "The content of the email to reply to"
},
"sender": {
"type": "string",
"description": "The sender of the original email"
},
"subject": {
"type": "string",
"description": "The subject of the original email"
}
},
"required": ["email_id", "email_content", "sender", "subject"]
}
)
async def handle(
gmail_client: GmailClient,
anthropic_client: Anthropic,
google_docs_helper: GoogleDocsHelper,
guidelines_doc_id: str,
arguments: dict
) -> list[TextContent]:
"""Handle the create_draft_reply tool call."""
email_id = arguments.get("email_id")
email_content = arguments.get("email_content")
sender = arguments.get("sender")
subject = arguments.get("subject")
try:
# Fetch guidelines
guidelines = await fetch_guidelines(google_docs_helper, guidelines_doc_id)
# Build the prompt
prompt = build_reply_prompt(sender, subject, email_content, guidelines)
# Generate reply
logger.info("Requesting AI-generated reply via Anthropic API...")
message = anthropic_client.messages.create(
model="claude-sonnet-4-20250514",
max_tokens=1000,
messages=[{"role": "user", "content": prompt}]
)
generated_reply = message.content[0].text
logger.info(f"Generated reply (first 100 chars): {generated_reply[:100]}...")
# Create draft in Gmail
gmail_client.create_draft_reply(
to_email=sender,
subject=subject,
body=generated_reply
)
return [
TextContent(
type="text",
text=f"✅ Draft reply created successfully!\n\nGenerated reply:\n{generated_reply}"
)
]
except Exception as e:
logger.error(f"Error creating draft reply: {e}")
return [
TextContent(
type="text",
text=f"Error creating draft reply: {str(e)}"
)
]
