main.py

from fastapi import FastAPI, HTTPException, Depends, Security, BackgroundTasks
from fastapi.security import APIKeyHeader
from fastapi.responses import StreamingResponse
from pydantic import BaseModel, Field
from typing import Literal, List, Dict
import os
from functools import lru_cache
from openai import OpenAI
from uuid import uuid4
import tiktoken
import sqlite3
import time
from datetime import datetime, timedelta
import asyncio
import requests
from prompts import CODING_ASSISTANT_PROMPT, NEWS_ASSISTANT_PROMPT, generate_news_prompt, SEARCH_ASSISTANT_PROMPT, generate_search_prompt
from fastapi_cache import FastAPICache
from fastapi_cache.backends.inmemory import InMemoryBackend
from fastapi_cache.decorator import cache
import logging

# Configure logging
logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
    handlers=[
        logging.FileHandler("app.log"),
        logging.StreamHandler()
    ]
)
logger = logging.getLogger(__name__)

app = FastAPI()

API_KEY_NAME = "X-API-Key"
API_KEY = os.environ.get("CHAT_AUTH_KEY", "default_secret_key")
api_key_header = APIKeyHeader(name=API_KEY_NAME, auto_error=False)

ModelID = Literal[
    "openai/gpt-4o-mini",
    "meta-llama/llama-3-70b-instruct",
    "anthropic/claude-3.5-sonnet",
    "deepseek/deepseek-coder",
    "anthropic/claude-3-haiku",
    "openai/gpt-3.5-turbo-instruct",
    "qwen/qwen-72b-chat",
    "google/gemma-2-27b-it"
]

class QueryModel(BaseModel):
    user_query: str = Field(..., description="User's coding query")
    model_id: ModelID = Field(
        default="meta-llama/llama-3-70b-instruct",
        description="ID of the model to use for response generation"
    )
    conversation_id: str = Field(default_factory=lambda: str(uuid4()), description="Unique identifier for the conversation")
    user_id: str = Field(..., description="Unique identifier for the user")

    class Config:
        schema_extra = {
            "example": {
                "user_query": "How do I implement a binary search in Python?",
                "model_id": "meta-llama/llama-3-70b-instruct",
                "conversation_id": "123e4567-e89b-12d3-a456-426614174000",
                "user_id": "user123"
            }
        }

class NewsQueryModel(BaseModel):
    query: str = Field(..., description="News topic to search for")
    model_id: ModelID = Field(
        default="openai/gpt-4o-mini",
        description="ID of the model to use for response generation"
    )
    class Config:
        schema_extra = {
            "example": {
                "query": "Latest developments in AI",
                "model_id": "openai/gpt-4o-mini"
            }
        }

@lru_cache()
def get_api_keys():
    logger.info("Loading API keys")
    return {
        "OPENROUTER_API_KEY": f"sk-or-v1-{os.environ['OPENROUTER_API_KEY']}",
        "BRAVE_API_KEY": os.environ['BRAVE_API_KEY']
    }

api_keys = get_api_keys()
or_client = OpenAI(api_key=api_keys["OPENROUTER_API_KEY"], base_url="https://openrouter.ai/api/v1")

# In-memory storage for conversations
conversations: Dict[str, List[Dict[str, str]]] = {}
last_activity: Dict[str, float] = {}

# Token encoding
encoding = tiktoken.encoding_for_model("gpt-3.5-turbo")

def limit_tokens(input_string, token_limit=6000):
    return encoding.decode(encoding.encode(input_string)[:token_limit])

def calculate_tokens(msgs):
    return sum(len(encoding.encode(str(m))) for m in msgs)

def chat_with_llama_stream(messages, model="openai/gpt-4o-mini", max_llm_history=4, max_output_tokens=2500):
    logger.info(f"Starting chat with model: {model}")
    while calculate_tokens(messages) > (8000 - max_output_tokens):
        if len(messages) > max_llm_history:
            messages = [messages[0]] + messages[-max_llm_history:]
        else:
            max_llm_history -= 1
            if max_llm_history < 2:
                error_message = "Token limit exceeded. Please shorten your input or start a new conversation."
                logger.error(error_message)
                raise HTTPException(status_code=400, detail=error_message)

    try:
        response = or_client.chat.completions.create(
            model=model,
            messages=messages,
            max_tokens=max_output_tokens,
            stream=True
        )
        
        full_response = ""
        for chunk in response:
            if chunk.choices[0].delta.content is not None:
                content = chunk.choices[0].delta.content
                full_response += content
                yield content
        
        # After streaming, add the full response to the conversation history
        messages.append({"role": "assistant", "content": full_response})
        logger.info("Chat completed successfully")
    except Exception as e:
        logger.error(f"Error in model response: {str(e)}")
        raise HTTPException(status_code=500, detail=f"Error in model response: {str(e)}")

async def verify_api_key(api_key: str = Security(api_key_header)):
    if api_key != API_KEY:
        logger.warning("Invalid API key used")
        raise HTTPException(status_code=403, detail="Could not validate credentials")
    return api_key

# SQLite setup
DB_PATH = '/app/data/conversations.db'

def init_db():
    logger.info("Initializing database")
    os.makedirs(os.path.dirname(DB_PATH), exist_ok=True)
    conn = sqlite3.connect(DB_PATH)
    c = conn.cursor()
    c.execute('''CREATE TABLE IF NOT EXISTS conversations
                 (id INTEGER PRIMARY KEY AUTOINCREMENT,
                  user_id TEXT,
                  conversation_id TEXT,
                  message TEXT,
                  response TEXT,
                  timestamp DATETIME DEFAULT CURRENT_TIMESTAMP)''')
    conn.commit()
    conn.close()
    logger.info("Database initialized successfully")

init_db()

def update_db(user_id, conversation_id, message, response):
    logger.info(f"Updating database for conversation: {conversation_id}")
    conn = sqlite3.connect(DB_PATH)
    c = conn.cursor()
    c.execute('''INSERT INTO conversations (user_id, conversation_id, message, response)
                 VALUES (?, ?, ?, ?)''', (user_id, conversation_id, message, response))
    conn.commit()
    conn.close()
    logger.info("Database updated successfully")

async def clear_inactive_conversations():
    while True:
        logger.info("Clearing inactive conversations")
        current_time = time.time()
        inactive_convos = [conv_id for conv_id, last_time in last_activity.items() 
                           if current_time - last_time > 1800]  # 30 minutes
        for conv_id in inactive_convos:
            if conv_id in conversations:
                del conversations[conv_id]
            if conv_id in last_activity:
                del last_activity[conv_id]
        logger.info(f"Cleared {len(inactive_convos)} inactive conversations")
        await asyncio.sleep(60)  # Check every minute

@app.on_event("startup")
async def startup_event():
    logger.info("Starting up the application")
    FastAPICache.init(InMemoryBackend(), prefix="fastapi-cache")
    asyncio.create_task(clear_inactive_conversations())

@app.post("/coding-assistant")
async def coding_assistant(query: QueryModel, background_tasks: BackgroundTasks, api_key: str = Depends(verify_api_key)):
    """
    Coding assistant endpoint that provides programming help based on user queries.
    Available models:
    - meta-llama/llama-3-70b-instruct (default)
    - anthropic/claude-3.5-sonnet
    - deepseek/deepseek-coder
    - anthropic/claude-3-haiku
    - openai/gpt-3.5-turbo-instruct
    - qwen/qwen-72b-chat
    - google/gemma-2-27b-it
    - openai/gpt-4o-mini
    Requires API Key authentication via X-API-Key header.
    """
    logger.info(f"Received coding assistant query: {query.user_query}")
    if query.conversation_id not in conversations:
        conversations[query.conversation_id] = [
            {"role": "system", "content": "You are a helpful assistant proficient in coding tasks. Help the user in understanding and writing code."}
        ]
    
    conversations[query.conversation_id].append({"role": "user", "content": query.user_query})
    last_activity[query.conversation_id] = time.time()
    
    # Limit tokens in the conversation history
    limited_conversation = conversations[query.conversation_id]

    def process_response():
        full_response = ""
        for content in chat_with_llama_stream(limited_conversation, model=query.model_id):
            full_response += content
            yield content
        background_tasks.add_task(update_db, query.user_id, query.conversation_id, query.user_query, full_response)
        logger.info(f"Completed coding assistant response for query: {query.user_query}")

    return StreamingResponse(process_response(), media_type="text/event-stream")

# New functions for news assistant

def internet_search(query, type = "web", num_results=20):
    logger.info(f"Performing internet search for query: {query}, type: {type}")
    if type == "web":
        url = "https://api.search.brave.com/res/v1/web/search"
    else:
        url = "https://api.search.brave.com/res/v1/news/search"

    headers = {
        "Accept": "application/json",
        "Accept-Encoding": "gzip",
        "X-Subscription-Token": api_keys["BRAVE_API_KEY"]
    }
    params = {"q": query}

    response = requests.get(url, headers=headers, params=params)

    if response.status_code != 200:
        logger.error(f"Failed to fetch search results. Status code: {response.status_code}")
        return []
    
    if type == "web":
        search_data = response.json()["web"]["results"]
    else:
        search_data = response.json()["results"]
    processed_results = []

    for item in search_data:
        if not item.get("extra_snippets"):
            continue
        
        result = {
            "title": item["title"],
            "snippet": item["extra_snippets"][0],
            "last_updated": item.get("age", "")
        }
        processed_results.append(result)

    logger.info(f"Retrieved {len(processed_results)} search results")
    return processed_results[:num_results]

@lru_cache(maxsize=100)
def cached_internet_search(query: str):
    logger.info(f"Performing cached internet search for query: {query}")
    return internet_search(query, type = "news")


def analyze_news(query):
    logger.info(f"Analyzing news for query: {query}")
    news_data = cached_internet_search(query)
    
    if not news_data:
        logger.error("Failed to fetch news data")
        return "Failed to fetch news data.", []

    # Prepare the prompt for the AI
    # Use the imported function to generate the prompt (now includes today's date)
    prompt = generate_news_prompt(query, news_data)

    messages = [
        {"role": "system", "content": NEWS_ASSISTANT_PROMPT},
        {"role": "user", "content": prompt}
    ]

    logger.info("News analysis completed")
    return messages

@app.post("/news-assistant")
async def news_assistant(query: NewsQueryModel, api_key: str = Depends(verify_api_key)):
    """
    News assistant endpoint that provides summaries and analysis of recent news based on user queries.
    Requires API Key authentication via X-API-Key header.
    """
    logger.info(f"Received news assistant query: {query.query}")
    messages = analyze_news(query.query)
    
    if not messages:
        logger.error("Failed to fetch news data")
        raise HTTPException(status_code=500, detail="Failed to fetch news data")
        
    def process_response():
        for content in chat_with_llama_stream(messages, model=query.model_id):
            yield content
        logger.info(f"Completed news assistant response for query: {query.query}")

    return StreamingResponse(process_response(), media_type="text/event-stream")

class SearchQueryModel(BaseModel):
    query: str = Field(..., description="Search query")
    model_id: ModelID = Field(
        default="openai/gpt-4o-mini",
        description="ID of the model to use for response generation"
    )
    class Config:
        schema_extra = {
            "example": {
                "query": "What are the latest advancements in quantum computing?",
                "model_id": "meta-llama/llama-3-70b-instruct"
            }
        }

def analyze_search_results(query):
    logger.info(f"Analyzing search results for query: {query}")
    search_data = internet_search(query, type="web")
    
    if not search_data:
        logger.error("Failed to fetch search data")
        return "Failed to fetch search data.", []

    # Prepare the prompt for the AI
    prompt = generate_search_prompt(query, search_data)

    messages = [
        {"role": "system", "content": SEARCH_ASSISTANT_PROMPT},
        {"role": "user", "content": prompt}
    ]

    logger.info("Search results analysis completed")
    return messages

@app.post("/search-assistant")
async def search_assistant(query: SearchQueryModel, api_key: str = Depends(verify_api_key)):
    """
    Search assistant endpoint that provides summaries and analysis of web search results based on user queries.
    Requires API Key authentication via X-API-Key header.
    """
    logger.info(f"Received search assistant query: {query.query}")
    messages = analyze_search_results(query.query)
    
    if not messages:
        logger.error("Failed to fetch search data")
        raise HTTPException(status_code=500, detail="Failed to fetch search data")
        
    def process_response():
        logger.info(f"Generating response using LLM: {messages}")
        full_response = ""
        for content in chat_with_llama_stream(messages, model=query.model_id):
            full_response+=content
            yield content
        logger.info(f"Completed search assistant response for query: {query.query}")
        logger.info(f"LLM Response: {full_response}")

    return StreamingResponse(process_response(), media_type="text/event-stream")


from pydantic import BaseModel, Field
import yaml
import json
from yaml.loader import SafeLoader

class FollowupQueryModel(BaseModel):
    query: str = Field(..., description="User's query for the followup agent")
    model_id: ModelID = Field(
        default="openai/gpt-4o-mini",
        description="ID of the model to use for response generation"
    )
    conversation_id: str = Field(default_factory=lambda: str(uuid4()), description="Unique identifier for the conversation")
    user_id: str = Field(..., description="Unique identifier for the user")

    class Config:
        schema_extra = {
            "example": {
                "query": "How can I improve my productivity?",
                "model_id": "openai/gpt-4o-mini",
                "conversation_id": "123e4567-e89b-12d3-a456-426614174000",
                "user_id": "user123"
            }
        }

FOLLOWUP_AGENT_PROMPT = """
You are a helpful, intelligent assistant who can create interactive buttons and markdown responses.

If the user request needs further clarification, provide clarifying questions using <interact> to assist the user.
Else respond with a helpful answer using <response>.   
The options in <interact> tags will be rendered as buttons so that user can interact with it. Hence make sure to use it for engaging with the user with Next Steps, followup questions, quizzes etc. whichever appropriate 

Your output format: # Use the following <response>,<interact> tags in any order as many times required.
<response>response to user request formatted using markdown</response>
<interact>
questions:
  - text: [First interaction question]
    options:
      - [Option 1]
      - [Option 2]
      - [Option 3]
      - [Option 4 (if needed)]
  - text: [Second interaction question]
    options:
      - [Option 1]
      - [Option 2]
      - [Option 3]
  # Add more questions as needed 
  # make sure this section is in valid YAML format
</interact>
"""

import re

def parse_followup_response(input_text):
    # Define patterns for response and interact
    response_pattern = re.compile(r'<response>(.*?)<\/response>', re.DOTALL)
    interact_pattern = re.compile(r'<interact>(.*?)<\/interact>', re.DOTALL)

    # Find all matches for response and interact
    response_matches = response_pattern.finditer(input_text)
    interact_matches = interact_pattern.finditer(input_text)

    # Initialize variables to keep track of the position
    last_end = 0
    combined_response = ""
    parsed_interacts = []

    # Combine responses and capture everything in between
    for response_match in response_matches:
        # Capture text before the current response tag
        combined_response += input_text[last_end:response_match.start()].strip() + "\n"
        # Add the response content
        combined_response += response_match.group(1).strip() + "\n"
        # Update the last end position
        last_end = response_match.end()

    # Check for interacts and parse them
    for interact_match in interact_matches:
        # Capture text before the current interact tag
        combined_response += input_text[last_end:interact_match.start()].strip() + "\n"
        # Process the interact block
        interact_text = interact_match.group(1).strip()
        if interact_text:
            # Split by "text:" to separate each question block
            question_blocks = interact_text.split("- text:")
            
            # Loop through each block and extract the question and its options
            for block in question_blocks[1:]:
                # Extract the question using regex (up to the "options:" part)
                question_match = re.search(r'^(.*?)\s*options:', block, re.DOTALL)
                if question_match:
                    question = question_match.group(1).strip().strip('"')
                    
                    # Extract the options using regex
                    options_match = re.search(r'options:\s*(.*?)$', block, re.DOTALL)
                    if options_match:
                        options = [option.strip().strip('"') for option in options_match.group(1).split('-') if option.strip()]
                        
                        # Add the parsed question and options to the list
                        parsed_interacts.append({'question': question, 'options': options})
        # Update the last end position
        last_end = interact_match.end()

    # Capture any remaining text after the last tag
    combined_response += input_text[last_end:].strip()

    return combined_response.strip(), parsed_interacts


@app.post("/followup-agent")
async def followup_agent(query: FollowupQueryModel, background_tasks: BackgroundTasks, api_key: str = Depends(verify_api_key)):
    """
    Followup agent endpoint that provides helpful responses or generates clarifying questions based on user queries.
    Requires API Key authentication via X-API-Key header.
    """
    logger.info(f"Received followup agent query: {query.query}")

    if query.conversation_id not in conversations:
        conversations[query.conversation_id] = [
            {"role": "system", "content": FOLLOWUP_AGENT_PROMPT}
        ]
    
    conversations[query.conversation_id].append({"role": "user", "content": query.query})
    last_activity[query.conversation_id] = time.time()
    
    # Limit tokens in the conversation history
    limited_conversation = conversations[query.conversation_id]

    def process_response():
        full_response = ""
        for content in chat_with_llama_stream(limited_conversation, model=query.model_id):
            full_response += content
            yield content

        logger.info(f"LLM RAW response for query: {query.query}: {full_response}")
        response_content, interact = parse_followup_response(full_response)
        
        result = {
            "response": response_content,
            "clarification": interact
        }
        
        yield "\n\n" + json.dumps(result)
        
        # Add the assistant's response to the conversation history
        conversations[query.conversation_id].append({"role": "assistant", "content": full_response})
        
        background_tasks.add_task(update_db, query.user_id, query.conversation_id, query.query, full_response)
        logger.info(f"Completed followup agent response for query: {query.query}, send result: {result}")

    return StreamingResponse(process_response(), media_type="text/event-stream")

@app.post("/v2/followup-agent")
async def followup_agent(query: FollowupQueryModel, background_tasks: BackgroundTasks, api_key: str = Depends(verify_api_key)):
    """
    Followup agent endpoint that provides helpful responses or generates clarifying questions based on user queries.
    Requires API Key authentication via X-API-Key header.
    """
    logger.info(f"Received followup agent query: {query.query}")

    if query.conversation_id not in conversations:
        conversations[query.conversation_id] = [
            {"role": "system", "content": FOLLOWUP_AGENT_PROMPT}
        ]
    
    conversations[query.conversation_id].append({"role": "user", "content": query.query})
    last_activity[query.conversation_id] = time.time()
    
    # Limit tokens in the conversation history
    limited_conversation = conversations[query.conversation_id]

    def process_response():
        full_response = ""
        for content in chat_with_llama_stream(limited_conversation, model=query.model_id):
            full_response += content
            yield content

        logger.info(f"LLM RAW response for query: {query.query}: {full_response}")
        response_content, interact = parse_followup_response(full_response)
        
        result = {
            "clarification": interact
        }
        
        yield "<json>" + json.dumps(result)
        
        # Add the assistant's response to the conversation history
        conversations[query.conversation_id].append({"role": "assistant", "content": full_response})
        
        background_tasks.add_task(update_db, query.user_id, query.conversation_id, query.query, full_response)
        logger.info(f"Completed followup agent response for query: {query.query}, send result: {result}")

    return StreamingResponse(process_response(), media_type="text/event-stream")

@app.post("/v3/followup-agent")
async def followup_agent(query: FollowupQueryModel, background_tasks: BackgroundTasks, api_key: str = Depends(verify_api_key)):
    """
    Followup agent endpoint that provides helpful responses or generates clarifying questions based on user queries.
    Requires API Key authentication via X-API-Key header.
    """
    logger.info(f"Received followup agent query: {query.query}")

    if query.conversation_id not in conversations:
        conversations[query.conversation_id] = [
            {"role": "system", "content": FOLLOWUP_AGENT_PROMPT}
        ]
    
    conversations[query.conversation_id].append({"role": "user", "content": query.query})
    last_activity[query.conversation_id] = time.time()
    
    # Limit tokens in the conversation history
    limited_conversation = conversations[query.conversation_id]

    def process_response():
        full_response = ""
        for content in chat_with_llama_stream(limited_conversation, model=query.model_id):
            full_response += content
            yield content

        logger.info(f"LLM RAW response for query: {query.query}: {full_response}")
        response_content, interact = parse_followup_response(full_response)
        
        result = {
            "clarification": interact
        }
        
        yield "<json>" +"[[["+ json.dumps(result)+"]]]"+"</json>"
        
        # Add the assistant's response to the conversation history
        conversations[query.conversation_id].append({"role": "assistant", "content": full_response})
        
        background_tasks.add_task(update_db, query.user_id, query.conversation_id, query.query, full_response)
        logger.info(f"Completed followup agent response for query: {query.query}, send result: {result}")

    return StreamingResponse(process_response(), media_type="text/event-stream")


if __name__ == "__main__":
    import uvicorn
    logger.info("Starting the application")
    uvicorn.run(app, host="0.0.0.0", port=7860)