Upload app.py with huggingface_hub

app.py

@@ -0,0 +1,116 @@
+
+import os
+import requests
+import gradio as gr
+from langchain_community.vectorstores import FAISS
+from langchain_community.embeddings import HuggingFaceEmbeddings
+from langchain_text_splitters import RecursiveCharacterTextSplitter
+from langchain_core.documents import Document
+from crewai import Agent, Task, Crew, Process
+
+os.environ["CEREBRAS_API_KEY"] = "csk-pk4m4ntefdvyt3vm495tre6jvenx9r88dytcyjt26mvv263m"
+
+documents = [
+    "Apple Inc. is a multinational technology company headquartered in Cupertino, California. Key products include iPhone, Mac, iPad, Apple Watch, and Apple TV. Revenue in FY2024 was approximately $391 billion.",
+    "Microsoft Corporation is headquartered in Redmond, Washington. Key products include Windows, Azure, Office 365, and LinkedIn. Revenue in FY2024 was approximately $245 billion.",
+    "NVIDIA Corporation designs graphics processing units and AI chips. Key products include GeForce GPUs and CUDA platform. Revenue in FY2024 was approximately $60 billion.",
+    "Tesla Inc. designs and manufactures electric vehicles and clean energy products. Key models include Model S, Model 3, Model X, and Model Y.",
+    "Amazon.com Inc. focuses on e-commerce, cloud computing, and AI. AWS is the world leading cloud platform. Revenue in FY2024 was approximately $620 billion.",
+    "Google LLC specializes in internet services including search, advertising, cloud computing, and AI. Parent company is Alphabet Inc.",
+    "Meta Platforms operates Facebook, Instagram, WhatsApp, and Threads. Revenue in FY2024 was approximately $165 billion.",
+    "Anthropic is an AI safety company that builds Claude. Founded in 2021 by former OpenAI researchers.",
+    "OpenAI develops GPT models, ChatGPT, DALL-E, and Sora. Microsoft is a major investor.",
+]
+
+splitter = RecursiveCharacterTextSplitter(chunk_size=200, chunk_overlap=30)
+docs = []
+for text in documents:
+    for chunk in splitter.split_text(text):
+        docs.append(Document(page_content=chunk))
+
+embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
+vector_store = FAISS.from_documents(docs, embeddings)
+
+def search_company_news(company_name):
+    try:
+        url = f"https://en.wikipedia.org/api/rest_v1/page/summary/{company_name.replace(' ', '_')}"
+        response = requests.get(url, timeout=10)
+        if response.status_code == 200:
+            data = response.json()
+            return f"Wikipedia: {data.get('extract', '')[:500]}"
+        return f"No data found for {company_name}"
+    except Exception as e:
+        return f"Search error: {str(e)}"
+
+def retrieve_from_knowledge_base(company_name):
+    results = vector_store.similarity_search(company_name, k=3)
+    if results:
+        return "\n".join([r.page_content for r in results])
+    return f"No info found for {company_name}"
+
+web_researcher = Agent(
+    role="Web Research Analyst",
+    goal="Gather latest information about a company from public sources",
+    backstory="You are an expert financial research analyst.",
+    llm="cerebras/llama3.1-8b",
+    verbose=False
+)
+
+kb_analyst = Agent(
+    role="Knowledge Base Analyst",
+    goal="Retrieve relevant company information from the internal knowledge base",
+    backstory="You are a specialist in querying structured knowledge bases.",
+    llm="cerebras/llama3.1-8b",
+    verbose=False
+)
+
+synthesiser = Agent(
+    role="Senior Investment Research Analyst",
+    goal="Synthesise information into a clear structured investment research report",
+    backstory="You are a senior analyst at a leading investment bank.",
+    llm="cerebras/llama3.1-8b",
+    verbose=False
+)
+
+def research_company(company_name):
+    if not company_name.strip():
+        return "Please enter a company name."
+    try:
+        web_task = Task(
+            description=f"Research {company_name}. Data: {search_company_news(company_name)[:400]}. Give 3 bullet points.",
+            expected_output="3 bullet points about the company",
+            agent=web_researcher
+        )
+        kb_task = Task(
+            description=f"Find info on {company_name}. Data: {retrieve_from_knowledge_base(company_name)[:400]}. Give 3 bullet points.",
+            expected_output="3 bullet points from knowledge base",
+            agent=kb_analyst
+        )
+        synth_task = Task(
+            description=f"Write a 200-word research report on {company_name} with sections: 1. Overview 2. Financials 3. Investment Considerations.",
+            expected_output="A 200-word structured research report",
+            agent=synthesiser,
+            context=[web_task, kb_task]
+        )
+        crew = Crew(
+            agents=[web_researcher, kb_analyst, synthesiser],
+            tasks=[web_task, kb_task, synth_task],
+            process=Process.sequential,
+            verbose=False
+        )
+        result = crew.kickoff()
+        return str(result)
+    except Exception as e:
+        return f"Error: {str(e)}"
+
+app = gr.Interface(
+    fn=research_company,
+    inputs=gr.Textbox(lines=1, placeholder="Enter company name e.g. Apple, Microsoft...", label="Company Name"),
+    outputs=gr.Textbox(lines=20, label="Research Report"),
+    title="Multi-Agent Investment Research Assistant",
+    description="Three AI agents collaborate using CrewAI, LangChain, FAISS, and Llama 3.1 to produce structured investment research reports.",
+    examples=[["Apple"], ["Microsoft"], ["NVIDIA"]],
+    theme=gr.themes.Soft()
+)
+
+app.launch(server_name="0.0.0.0", server_port=7860, share=False)