main.py

#!/usr/bin/env python3
"""
Job Search MCP Server - Smart job matching and instant application helper

Main entry point for the MCP server that exposes the four core endpoints:
1. profile.upsert - Store user résumé, skills, salary wish, and career goals
2. jobs.search - Pull fresh job posts, rank with GPU embeddings, return fit scores
3. letter.generate - Create personalized cover letters using LLM
4. qa.reply - Draft concise answers to client questions

This server uses both GPU processing for embeddings and LLM APIs for text generation,
demonstrating efficient use of both credit pools.
"""

import asyncio
import gradio as gr
from typing import Dict, Any

from src.tools import ProfileTool, JobSearchTool, CoverLetterTool, QATool
from src.config import get_settings


class JobSearchMCPServer:
    """Main MCP server class integrating all job search tools."""

    def __init__(self):
        self.settings = get_settings()

        # Initialize all tools
        self.profile_tool = ProfileTool()
        self.job_search_tool = JobSearchTool()
        self.cover_letter_tool = CoverLetterTool()
        self.qa_tool = QATool()

        print(f"🚀 Job Search MCP Server initialized")
        print(f"📊 GPU Embeddings: {self.settings.embedding_model}")
        print(f"🤖 LLM Provider: {self.settings.llm_provider}")

    # Core MCP Endpoints

    def profile_upsert(self, user_id: str, profile_data: str) -> Dict[str, Any]:
        """
        MCP Endpoint: profile.upsert

        Stores the user résumé, skills, salary expectations, and career goals.
        Keeps personal context so every later call is tailored.
        """
        return self.profile_tool.upsert(user_id, profile_data)

    def jobs_search(
        self, user_id: str, query: str = "", location: str = "", job_type: str = ""
    ) -> Dict[str, Any]:
        """
        MCP Endpoint: jobs.search

        Pulls fresh job posts, ranks them with GPU embeddings, and returns a fit score.
        Users see the most relevant roles first, no endless scrolling.
        """
        return self.job_search_tool.search(user_id, query, location, job_type)

    def letter_generate(
        self, user_id: str, job_description: str, tone: str = "professional"
    ) -> Dict[str, Any]:
        """
        MCP Endpoint: letter.generate

        Calls an LLM to create a short, personalized cover letter in any tone.
        Saves time and improves response quality.
        """
        return self.cover_letter_tool.generate(user_id, job_description, tone)

    def qa_reply(
        self, user_id: str, question: str, context: str = ""
    ) -> Dict[str, Any]:
        """
        MCP Endpoint: qa.reply

        Drafts concise answers to client questions like "Why should we hire you?"
        Speeds up Upwork, Fiverr, or LinkedIn chats.
        """
        return self.qa_tool.reply(user_id, question, context)

    # Additional Helper Endpoints

    def get_server_stats(self) -> Dict[str, Any]:
        """Get server statistics and health information."""
        try:
            from src.services import EmbeddingService

            embedding_service = EmbeddingService()
            embed_stats = embedding_service.get_index_stats()

            return {
                "success": True,
                "server_info": {
                    "app_name": self.settings.app_name,
                    "embedding_model": self.settings.embedding_model,
                    "llm_provider": self.settings.llm_provider,
                    "llm_model": self.settings.llm_model,
                },
                "embedding_stats": embed_stats,
                "endpoints": [
                    "profile.upsert",
                    "jobs.search",
                    "letter.generate",
                    "qa.reply",
                ],
            }
        except Exception as e:
            return {
                "success": False,
                "message": f"Error getting server stats: {str(e)}",
            }


# Initialize the MCP server
mcp_server = JobSearchMCPServer()


# Create Gradio interface for easy testing and demonstration
def create_gradio_interface():
    """Create a Gradio interface for the MCP server."""

    with gr.Blocks(
        title="Job Search MCP Server",
        theme=gr.themes.Soft(),
        css="""
        .gradio-container {
            max-width: 1200px !important;
        }
        .main-header {
            text-align: center;
            background: linear-gradient(45deg, #667eea 0%, #764ba2 100%);
            color: white;
            padding: 20px;
            border-radius: 10px;
            margin-bottom: 20px;
        }
        """,
    ) as demo:
        # Header
        gr.HTML("""
        <div class="main-header">
            <h1>🔍 Job Search MCP Server</h1>
            <p>Smart job matching and instant application helper</p>
            <p><strong>4 Core Endpoints:</strong> profile.upsert | jobs.search | letter.generate | qa.reply</p>
        </div>
        """)

        # Server Stats
        with gr.Row():
            with gr.Column():
                stats_button = gr.Button("📊 Get Server Stats", variant="secondary")
                stats_output = gr.JSON(label="Server Statistics")

                stats_button.click(fn=mcp_server.get_server_stats, outputs=stats_output)

        # Main endpoints in tabs
        with gr.Tabs():
            # Profile Management Tab
            with gr.Tab("👤 Profile Management (profile.upsert)"):
                gr.Markdown("### Store and update your professional profile")
                gr.Markdown(
                    "*This endpoint keeps personal context so every later call is tailored*"
                )

                with gr.Row():
                    with gr.Column():
                        profile_user_id = gr.Textbox(
                            label="User ID",
                            placeholder="Enter your unique user ID (e.g., john_doe_2024)",
                            value="demo_user",
                        )
                        profile_data = gr.TextArea(
                            label="Profile Data (JSON)",
                            placeholder='{\n  "resume": "Full resume text here...",\n  "skills": ["Python", "JavaScript", "React", "Node.js"],\n  "salary_wish": "$80,000 - $120,000 annually",\n  "career_goals": "Looking to transition into a senior full-stack developer role at a tech company",\n  "experience_level": "Mid-level",\n  "location": "Remote",\n  "education": "BS Computer Science"\n}',
                            lines=8,
                        )
                        profile_submit = gr.Button(
                            "💾 Update Profile", variant="primary"
                        )

                    with gr.Column():
                        profile_output = gr.JSON(label="Response")

                        # Quick profile actions
                        with gr.Row():
                            get_profile_btn = gr.Button(
                                "👁️ View Profile", variant="secondary"
                            )
                            delete_profile_btn = gr.Button(
                                "🗑️ Delete Profile", variant="secondary"
                            )

                profile_submit.click(
                    fn=mcp_server.profile_upsert,
                    inputs=[profile_user_id, profile_data],
                    outputs=profile_output,
                )

                get_profile_btn.click(
                    fn=mcp_server.profile_tool.get,
                    inputs=[profile_user_id],
                    outputs=profile_output,
                )

                delete_profile_btn.click(
                    fn=mcp_server.profile_tool.delete,
                    inputs=[profile_user_id],
                    outputs=profile_output,
                )

            # Job Search Tab
            with gr.Tab("🔍 Job Search (jobs.search)"):
                gr.Markdown(
                    "### Find and rank relevant job opportunities with GPU embeddings"
                )
                gr.Markdown(
                    "*Pulls fresh job posts, ranks them with GPU embeddings, and returns fit scores*"
                )

                with gr.Row():
                    with gr.Column():
                        search_user_id = gr.Textbox(label="User ID", value="demo_user")
                        search_query = gr.Textbox(
                            label="Search Query",
                            placeholder="e.g., Python developer, Data scientist, Frontend engineer",
                        )

                        with gr.Row():
                            search_location = gr.Textbox(
                                label="Location",
                                placeholder="e.g., Remote, New York, San Francisco",
                            )
                            search_job_type = gr.Dropdown(
                                label="Job Type",
                                choices=[
                                    "full-time",
                                    "part-time",
                                    "contract",
                                    "freelance",
                                    "remote",
                                ],
                                value="full-time",
                            )

                        search_submit = gr.Button("🔍 Search Jobs", variant="primary")

                    with gr.Column():
                        search_output = gr.JSON(label="Job Results with Fit Scores")

                        # Additional job search features
                        suggestions_btn = gr.Button(
                            "💡 Get Search Suggestions", variant="secondary"
                        )
                        clear_cache_btn = gr.Button(
                            "🗑️ Clear Job Cache", variant="secondary"
                        )

                search_submit.click(
                    fn=mcp_server.jobs_search,
                    inputs=[
                        search_user_id,
                        search_query,
                        search_location,
                        search_job_type,
                    ],
                    outputs=search_output,
                )

                suggestions_btn.click(
                    fn=mcp_server.job_search_tool.get_search_suggestions,
                    inputs=[search_user_id],
                    outputs=search_output,
                )

                clear_cache_btn.click(
                    fn=mcp_server.job_search_tool.clear_job_cache, outputs=search_output
                )

            # Cover Letter Generator Tab
            with gr.Tab("📝 Cover Letter (letter.generate)"):
                gr.Markdown("### Generate personalized cover letters using LLM")
                gr.Markdown(
                    "*Creates short, personalized cover letters in any tone - saves time and improves quality*"
                )

                with gr.Row():
                    with gr.Column():
                        letter_user_id = gr.Textbox(label="User ID", value="demo_user")
                        letter_tone = gr.Dropdown(
                            label="Tone",
                            choices=[
                                "professional",
                                "casual",
                                "enthusiastic",
                                "formal",
                            ],
                            value="professional",
                        )
                        letter_job_desc = gr.TextArea(
                            label="Job Description",
                            placeholder="Paste the complete job description here...",
                            lines=6,
                        )
                        letter_submit = gr.Button(
                            "📝 Generate Cover Letter", variant="primary"
                        )

                    with gr.Column():
                        letter_output = gr.JSON(label="Generated Cover Letter")

                        # Additional cover letter features
                        multiple_tones_btn = gr.Button(
                            "🎭 Generate Multiple Tones", variant="secondary"
                        )
                        template_btn = gr.Button("📋 Get Template", variant="secondary")

                letter_submit.click(
                    fn=mcp_server.letter_generate,
                    inputs=[letter_user_id, letter_job_desc, letter_tone],
                    outputs=letter_output,
                )

                multiple_tones_btn.click(
                    fn=mcp_server.cover_letter_tool.generate_multiple_tones,
                    inputs=[letter_user_id, letter_job_desc],
                    outputs=letter_output,
                )

                template_btn.click(
                    fn=mcp_server.cover_letter_tool.get_cover_letter_template,
                    inputs=[letter_tone],
                    outputs=letter_output,
                )

            # Q&A Assistant Tab
            with gr.Tab("💬 Q&A Assistant (qa.reply)"):
                gr.Markdown(
                    "### Get help with interview questions and client responses"
                )
                gr.Markdown(
                    "*Drafts concise answers to speed up Upwork, Fiverr, or LinkedIn chats*"
                )

                with gr.Row():
                    with gr.Column():
                        qa_user_id = gr.Textbox(label="User ID", value="demo_user")
                        qa_question = gr.TextArea(
                            label="Question",
                            placeholder="e.g., Why should we hire you?\nWhat's your experience with Python?\nHow much do you charge for this project?",
                            lines=4,
                        )
                        qa_context = gr.Textbox(
                            label="Context (optional)",
                            placeholder="Additional context about the role or conversation...",
                        )
                        qa_submit = gr.Button("💬 Generate Response", variant="primary")

                    with gr.Column():
                        qa_output = gr.JSON(label="Generated Response")

                        # Additional Q&A features
                        with gr.Row():
                            common_questions_btn = gr.Button(
                                "❓ Common Questions", variant="secondary"
                            )
                            practice_session_btn = gr.Button(
                                "🎯 Practice Session", variant="secondary"
                            )

                qa_submit.click(
                    fn=mcp_server.qa_reply,
                    inputs=[qa_user_id, qa_question, qa_context],
                    outputs=qa_output,
                )

                common_questions_btn.click(
                    fn=lambda: mcp_server.qa_tool.get_common_questions("developer"),
                    outputs=qa_output,
                )

                practice_session_btn.click(
                    fn=lambda uid: mcp_server.qa_tool.practice_session(
                        uid, "developer", 3
                    ),
                    inputs=[qa_user_id],
                    outputs=qa_output,
                )

        # Footer with usage information
        gr.HTML("""
        <div style="margin-top: 40px; padding: 20px; background-color: #f0f0f0; border-radius: 10px;">
            <h3>🎯 How It Works</h3>
            <p><strong>GPU Part (T4-small):</strong> The server embeds user profile text and each job post with a modern sentence-embedding model. A FAISS index runs similarity search in real time.</p>
            <p><strong>Inference-API Part:</strong> A hosted LLM writes cover letters and Q&A replies. Average call is under 300 tokens.</p>
            <p><strong>Typical User Flow:</strong></p>
            <ol>
                <li>Upload résumé and skills once using <code>profile.upsert</code></li>
                <li>Call <code>jobs.search</code> with a role keyword (e.g., "LLM engineer")</li>
                <li>Get a ranked list of matches with fit percentages</li>
                <li>Pick a job ID and call <code>letter.generate</code> to copy a ready cover letter</li>
                <li>When the recruiter asks something, send the question to <code>qa.reply</code> for an instant answer</li>
            </ol>
            <p><strong>Benefits:</strong> Cuts application time by 80%+, reduces copy-pasted cover letters, improves job-to-skill matching</p>
        </div>
        """)

    return demo


def main():
    """Main entry point for the application."""
    print("🚀 Starting Job Search MCP Server...")

    # Create and launch Gradio interface
    demo = create_gradio_interface()

    # Launch with MCP enabled
    demo.launch(
        server_name=mcp_server.settings.host,
        server_port=mcp_server.settings.port,
        mcp_server=True,
        share=False,
        show_error=True
        )


if __name__ == "__main__":
    main()