phase1_agents.py

import os
import faiss
import numpy as np
from phi.agent import Agent
from phi.tools.firecrawl import FirecrawlTools
from phi.tools.duckduckgo import DuckDuckGo
from phi.model.openai import OpenAIChat
from phi.embedder.openai import OpenAIEmbedder
from typing import List
from pydantic import BaseModel, Field
from fastapi import UploadFile

# Load environment variables (API keys, etc.)
from dotenv import load_dotenv
load_dotenv()

#####################################################################################
#                                    PHASE 1                                        #
#####################################################################################

##############################
# 1️⃣ Company Search Agent   #
##############################
company_search_agent = Agent(
    name="Company Search Agent",
    model=OpenAIChat(id="gpt-4o"),
    tools=[DuckDuckGo()],
    description="Finds company details based on name using web search.",
    instructions=["Always include sources in search results."],
    show_tool_calls=True,
    markdown=True,
)

def search_company(company_name: str):
    query = f"Find detailed company information for {company_name}. Extract its official website, mission, services, and any AI-related initiatives. Prioritize official sources and provide links where available."
    response = company_search_agent.run(query)
    return response.content


##############################
# 2️⃣ Website Scraper Agent   #
##############################
firecrawl_agent = Agent(
    name="Website Scraper Agent",
    tools=[FirecrawlTools(scrape=True, crawl=False)],
    description="Extracts content from company websites.",
    show_tool_calls=True,
    markdown=True,
)

def scrape_website(url: str):
    response = firecrawl_agent.run(f"Extract all relevant business information from {url}, including mission statement, services, case studies, and AI-related content. Provide structured output.")
    return response.content

##############################
# 3️⃣ Text Processing Agent   #
##############################
class CompanySummary(BaseModel):
    summary: str = Field(..., description="Summarized company details based on user input.")

text_processing_agent = Agent(
    model=OpenAIChat(id="gpt-4o"),
    description="Summarizes user-written company descriptions.",
    response_model=CompanySummary,
)

def process_company_description(text: str):
    response = text_processing_agent.run(f"Summarize the following company description: {text}. Focus on key services, mission, industry, and potential AI use cases where applicable.")
    return response.content

#################################
# 4️⃣ Document Processing Agent  #
#################################
# FAISS Index for storing extracted knowledge
embedding_model = OpenAIEmbedder(model="text-embedding-3-small")
dimension = 1536  # OpenAI's embedding dimension
faiss_index = faiss.IndexFlatL2(dimension)



def process_uploaded_document(file: UploadFile):
    file_path = f"tmp/{file.filename}"
    with open(file_path, "wb") as buffer:
        buffer.write(file.file.read())
    
    with open(file_path, "r", encoding="utf-8") as f:
        document_text = f.read()
    
    # Optionally, you can generate an embedding if needed, but return the raw content for further phases.
    return document_text  # Return the processed document text


# def process_uploaded_document(file: UploadFile):
#     file_path = f"tmp/{file.filename}"
#     with open(file_path, "wb") as buffer:
#         buffer.write(file.file.read())
    
#     with open(file_path, "r", encoding="utf-8") as f:
#         document_text = f.read()
    
#     # Generate embedding
#     embedding = np.array(embedding_model.embed([document_text])).astype("float32")
#     faiss_index.add(embedding)
    
#     return f"Document processed and stored in FAISS index: {file.filename}"