app.py

import os
import streamlit as st
import requests
import feedparser
import datetime
from dotenv import load_dotenv
from duckduckgo_search import DDGS

load_dotenv()
OPENROUTER_API_KEY = os.getenv("OPENROUTER_API_KEY")

# --- Call OpenRouter LLM ---
def call_llm(messages, model="deepseek/deepseek-chat-v3-0324:free", max_tokens=2048, temperature=0.7):
    url = "https://openrouter.ai/api/v1/chat/completions"
    headers = {
        "Authorization": f"Bearer {OPENROUTER_API_KEY}",
        "Content-Type": "application/json",
        "X-Title": "Autonomous Research Assistant"
    }
    data = {
        "model": model,
        "messages": messages,
        "max_tokens": max_tokens,
        "temperature": temperature
    }
    try:
        response = requests.post(url, headers=headers, json=data)
        result = response.json()
    except Exception as e:
        raise RuntimeError(f"Failed to connect or parse response: {e}")
    if response.status_code != 200:
        raise RuntimeError(result.get("error", {}).get("message", "LLM API error"))
    if "choices" not in result:
        raise RuntimeError(f"Invalid response: {result}")
    return result["choices"][0]["message"]["content"]

# --- Source Utilities ---
def get_arxiv_papers(query, max_results=3):
    from urllib.parse import quote_plus
    url = f"http://export.arxiv.org/api/query?search_query=all:{quote_plus(query)}&start=0&max_results={max_results}"
    feed = feedparser.parse(url)
    return [{
        "title": e.title or "Untitled",
        "summary": (e.summary or "No summary available").replace("\n", " ").strip(),
        "url": next((l.href for l in e.links if l.type == "application/pdf"), "")
    } for e in feed.entries]

def get_semantic_scholar_papers(query, max_results=3):
    url = "https://api.semanticscholar.org/graph/v1/paper/search"
    params = {"query": query, "limit": max_results, "fields": "title,abstract,url"}
    response = requests.get(url, params=params)
    papers = response.json().get("data", [])
    return [{
        "title": p.get("title") or "Untitled",
        "summary": (p.get("abstract") or "No abstract available").strip(),
        "url": p.get("url", "")
    } for p in papers]

def search_duckduckgo(query, max_results=3):
    with DDGS() as ddgs:
        return [{
            "title": r["title"] or "Untitled",
            "snippet": r["body"] or "",
            "url": r["href"] or ""
        } for r in ddgs.text(query, max_results=max_results)]

def get_image_urls(query, max_images=3):
    with DDGS() as ddgs:
        return [img["image"] for img in ddgs.images(query, max_results=max_images)]

def generate_apa_citation(title, url, source=""):
    current_year = datetime.datetime.now().year
    if source == "arxiv":
        return f"{title}. ({current_year}). *arXiv*. {url}"
    elif source == "semantic":
        return f"{title}. ({current_year}). *Semantic Scholar*. {url}"
    elif source == "web":
        return f"{title}. ({current_year}). *Web Source*. {url}"
    else:
        return f"{title}. ({current_year}). {url}"


# --- Research Agent ---
def autonomous_research_agent(topic):
    arxiv = get_arxiv_papers(topic)
    scholar = get_semantic_scholar_papers(topic)
    web = search_duckduckgo(topic)
    images = get_image_urls(topic)

    arxiv_md, arxiv_citations = "", []
    for p in arxiv:
        arxiv_md += f"- [{p['title']}]({p['url']})\n> {p['summary'][:300]}...\n\n"
        arxiv_citations.append(generate_apa_citation(p["title"], p["url"], source="arxiv"))

    scholar_md, scholar_citations = "", []
    for p in scholar:
        scholar_md += f"- [{p['title']}]({p['url']})\n> {p['summary'][:300]}...\n\n"
        scholar_citations.append(generate_apa_citation(p["title"], p["url"], source="semantic"))

    web_md, web_citations = "", []
    for w in web:
        web_md += f"- [{w['title']}]({w['url']})\n> {w['snippet']}\n\n"
        web_citations.append(generate_apa_citation(w["title"], w["url"], source="web"))

    prompt = f"""
# Research Topic: {topic}

## ArXiv:
{arxiv_md}

## Semantic Scholar:
{scholar_md}

## Web Insights:
{web_md}

Now synthesize this information into:
1. A research gap
2. A novel research direction
3. A full markdown-formatted research article (continuous, no section labels, academic tone)
"""

    response = call_llm([{"role": "user", "content": prompt}], max_tokens=3000)

    # Append Sources
    response += "\n\n---\n### Sources Cited\n"
    if arxiv_md:
        response += "**ArXiv:**\n" + arxiv_md
    if scholar_md:
        response += "**Semantic Scholar:**\n" + scholar_md
    if web_md:
        response += "**Web:**\n" + web_md

    # APA Citations Section
    all_citations = arxiv_citations + scholar_citations + web_citations
    response += "\n---\n### 📚 APA Citations\n"
    for cite in all_citations:
        response += f"- {cite}\n"

    return response, images

# --- Streamlit UI ---
st.set_page_config("Autonomous Research Assistant", layout="wide")
st.title("🤖 Autonomous AI Research Assistant")

if "chat_history" not in st.session_state:
    st.session_state.chat_history = []

topic = st.text_input("Enter a research topic:")
if st.button("Run Research Agent"):
    with st.spinner("Gathering sources & thinking..."):
        try:
            response, images = autonomous_research_agent(topic)

            # Display images
            if images:
                st.subheader("🖼️ Relevant Images")
                st.image(images, width=300)

            # Display markdown response
            st.session_state.chat_history.append({"role": "user", "content": topic})
            st.session_state.chat_history.append({"role": "assistant", "content": response})
            st.markdown(response)
        except Exception as e:
            st.error(f"Failed: {e}")

# --- Follow-up Chat ---
st.divider()
st.subheader("💬 Follow-up Q&A")
followup = st.text_input("Ask a follow-up question:")
if st.button("Ask"):
    if followup:
        try:
            chat = st.session_state.chat_history + [{"role": "user", "content": followup}]
            answer = call_llm(chat, max_tokens=1500)
            st.session_state.chat_history.append({"role": "user", "content": followup})
            st.session_state.chat_history.append({"role": "assistant", "content": answer})
            st.markdown(answer)
        except Exception as e:
            st.error(f"Follow-up error: {e}")