import os
import requests
from datetime import datetime, timedelta
from dotenv import load_dotenv
from langchain_openai import ChatOpenAI
from langchain.prompts import ChatPromptTemplate, SystemMessagePromptTemplate
from langchain_core.prompts import PromptTemplate
from fin_interpreter import analyze_article

# === Load environment ===
load_dotenv()
OPENAI_KEY = os.getenv("OPENAI_API_KEY")
TAVILY_KEY = os.getenv("TAVILY_API_KEY")

# === Get OpenAI client ===
def get_llm():
    if not OPENAI_KEY:
        raise ValueError("OPENAI_API_KEY not found.")
    return ChatOpenAI(model_name="gpt-4.1", openai_api_key=OPENAI_KEY)

# === Related Terms (cleaned)
def get_related_terms(topic):
    llm = get_llm()
    prompt = f"What are 5 short financial or industry keywords closely related to '{topic}'? Only return a comma-separated list."
    response = llm.invoke(prompt)
    raw = response.content.strip().split("\n")[0]
    return [term.strip() for term in raw.split(",") if term.strip()][:5]

# === Tavily Search
def tavily_search(query, days, max_results=10):
    headers = {"Authorization": f"Bearer {TAVILY_KEY}"}
    payload = {
        "query": query,
        "search_depth": "advanced",
        "topic": "news",
        "days": int(days),
        "max_results": max_results,
        "include_answer": False,
        "include_raw_content": False
    }
    response = requests.post("https://api.tavily.com/search", json=payload, headers=headers)
    if response.status_code != 200:
        print(f"⚠️ Tavily API error: {response.status_code} - {response.text}")
        return {}
    return response.json()

# === Smart News Search
def fetch_deep_news(topic, days):
    all_results = []
    seen_urls = set()
    cutoff = datetime.now() - timedelta(days=days)

    base_queries = [
        topic,
        f"{topic} AND startup",
        f"{topic} AND acquisition OR merger OR funding",
        f"{topic} AND CEO OR executive OR leadership",
        f"{topic} AND venture capital OR Series A OR Series B",
        f"{topic} AND government grant OR approval OR contract",
        f"{topic} AND underrated OR small-cap OR micro-cap"
    ]

    investor_queries = [
        f"{topic} AND BlackRock OR Vanguard OR SoftBank",
        f"{topic} AND Elon Musk OR Sam Altman OR Peter Thiel",
        f"{topic} AND Berkshire Hathaway OR Warren Buffett",
        f"{topic} AND institutional investor OR hedge fund",
    ]

    related_terms = get_related_terms(topic)
    synonym_queries = [f"{term} AND {kw}" for term in related_terms for kw in ["startup", "funding", "merger", "acquisition"]]
    all_queries = base_queries + investor_queries + synonym_queries

    for query in all_queries:
        print(f"🔍 Tavily query: {query}")
        response = tavily_search(query, days)

        for item in response.get("results", []):
            url = item.get("url")
            content = item.get("content", "") or item.get("summary", "") or item.get("title", "")
            pub_date = item.get("published_date")

            if not url or url in seen_urls or len(content) < 150:
                continue

            # Filter out old news
            if pub_date:
                try:
                    date_obj = datetime.fromisoformat(pub_date.rstrip("Z"))
                    if date_obj < cutoff:
                        continue
                except Exception:
                    pass

            # Filter out non-financial content
            finance_keywords = ["valuation", "IPO", "Series A", "revenue", "funding", "merger", "acquisition", "earnings"]
            if not any(kw in content.lower() for kw in finance_keywords):
                continue

            all_results.append({
                "title": item.get("title"),
                "url": url,
                "content": content
            })
            seen_urls.add(url)

    print(f"📰 Total articles collected: {len(all_results)}")
    return all_results

# === Generate Markdown Report
def generate_value_investor_report(topic, news_results, max_articles=20, max_chars_per_article=400):
    news_results = sorted(news_results, key=lambda x: len(x.get("content", "")), reverse=True)
    news_results = news_results[:max_articles]

    for item in news_results:
        text = item.get("summary") or item.get("content", "")
        result = analyze_article(text)
        item["fin_sentiment"] = result.get("sentiment", "neutral")
        item["fin_confidence"] = result.get("confidence", 0.0)
        item["investment_decision"] = result.get("investment_decision", "Watch")

    article_summary = "".join(
        f"- **{item['title']}**: {item['content'][:max_chars_per_article]}... "
        f"(Sentiment: {item['fin_sentiment'].title()}, Confidence: {item['fin_confidence']:.2f}, "
        f"Decision: {item['investment_decision']}) [link]({item['url']})\n"
        for item in news_results
    )

    prompt = PromptTemplate.from_template("""
You're a highly focused value investor. Today is {Today}. Analyze this week's news on "{Topic}".

Your goal is to uncover:
- Meaningful events (e.g., CEO joining a startup, insider buys, big-name partnerships)
- Startups or small caps that may signal undervalued opportunity
- Connections to key individuals or institutions (e.g., Elon Musk investing, Sam Altman joining)
- Companies with strong fundamentals: low P/E, low P/B, high ROE, recent IPOs, moats, or high free cash flow

### News
{ArticleSummaries}

Write a markdown memo with:
1. **Key Value Signals**
2. **Stocks or Startups to Watch** — MUST include rationale and for each: P/E, P/B, Debt-to-Equity, FCF, PEG
3. **What Smart Money Might Be Acting On**
4. **References**
5. **Investment Hypothesis**

---

### 📌 Executive Summary

Summarize the topic's current investment environment in 3–4 bullet points. Include sentiment, risks, and catalysts.

---

### 📊 Signals and Analysis (Include Sources)

For each important news item, write a short paragraph with:
- What happened
- Why it matters (financially)
- Embedded source as `[source title](url)`
- Bold any key financial terms (e.g., **Series A**, **merger**, **valuation**)

---

### 🧠 Investment Thesis

Give a reasoned conclusion:
- Is this a buy/sell/watch opportunity?
- What’s the risk/reward?
- What signals or themes matter most?
""")

    chat_prompt = ChatPromptTemplate.from_messages([
        SystemMessagePromptTemplate(prompt=prompt)
    ])
    prompt_value = chat_prompt.format_prompt(
        Topic=topic,
        ArticleSummaries=article_summary,
        Today=datetime.now().strftime("%B %d, %Y")
    ).to_messages()

    llm = get_llm()
    result = llm.invoke(prompt_value)
    return result.content