app.py

"""
EBP Research Tool for Student Nurses
Streamlit app — deployable to Hugging Face Spaces (free CPU tier).
"""

import streamlit as st

from search.pubmed import search_pubmed, fetch_abstract, fetch_summaries
from search.clinical_trials import search_trials
from summarizer import summarize, infer_evidence_level
from utils.citations import format_apa7, format_ama

# ---------------------------------------------------------------------------
# Page config (must be first Streamlit call)
# ---------------------------------------------------------------------------
st.set_page_config(
    page_title="EBP Research Tool — Student Nurses",
    page_icon="🩺",
    layout="wide",
    initial_sidebar_state="expanded",
)

# ---------------------------------------------------------------------------
# Minimal custom CSS
# ---------------------------------------------------------------------------
st.markdown(
    """
    <style>
    /* Card container */
    .ebp-card {
        border: 1px solid #d9e2ec;
        border-radius: 10px;
        padding: 1rem 1.2rem;
        margin-bottom: 1rem;
        background: #ffffff;
    }
    /* Source badges */
    .badge-pubmed  { background:#e8f4fd; color:#1558b0; padding:2px 8px;
                     border-radius:4px; font-size:0.72em; font-weight:600; }
    .badge-trials  { background:#fff3e0; color:#e65100; padding:2px 8px;
                     border-radius:4px; font-size:0.72em; font-weight:600; }
    /* Evidence level colours */
    .ev-I   { color:#1b7c1b; font-weight:700; }
    .ev-II  { color:#7c6b00; font-weight:700; }
    .ev-III { color:#b04e00; font-weight:700; }
    .ev-IV  { color:#b04e00; font-weight:700; }
    .ev-V   { color:#005f8c; font-weight:700; }
    .ev-VI  { color:#005f8c; font-weight:700; }
    .ev-VII { color:#333;    font-weight:700; }
    /* Slightly tighter headings */
    h1 { margin-bottom: 0 !important; }
    </style>
    """,
    unsafe_allow_html=True,
)

# ---------------------------------------------------------------------------
# Session state initialisation
# ---------------------------------------------------------------------------
_DEFAULTS = {
    "saved_articles":   [],
    "search_results":   [],
    "search_query":     "",
    "abstracts":        {},   # pmid → abstract text
    "summaries":        {},   # pmid → summary dict
}
for _k, _v in _DEFAULTS.items():
    if _k not in st.session_state:
        st.session_state[_k] = _v


# ---------------------------------------------------------------------------
# Helper: render one article card
# ---------------------------------------------------------------------------
def render_card(article: dict, card_idx: int, tab_key: str) -> None:
    pmid    = article.get("pmid", str(card_idx))
    title   = article.get("title", "No title")
    authors = article.get("authors", [])
    journal = article.get("journal", "")
    year    = article.get("year", "")
    url     = article.get("url", "")
    source  = article.get("source", "PubMed")

    badge_cls = "badge-pubmed" if source == "PubMed" else "badge-trials"
    badge_lbl = "PubMed" if source == "PubMed" else "ClinicalTrials"

    # Infer evidence level from whatever we already have
    cached_abstract = st.session_state.abstracts.get(pmid, "")
    ev_code, ev_emoji, ev_desc = infer_evidence_level(title, cached_abstract)

    with st.container():
        st.markdown(
            f'<span class="{badge_cls}">{badge_lbl}</span> '
            f'<span style="font-size:0.8em; color:#555;">{ev_emoji} {ev_code} — {ev_desc}</span>',
            unsafe_allow_html=True,
        )
        st.markdown(f"**{title}**")
        meta_parts = []
        if authors:
            meta_parts.append(", ".join(authors[:3]) + (" …" if len(authors) > 3 else ""))
        if journal:
            meta_parts.append(f"*{journal}*")
        if year:
            meta_parts.append(year)
        if meta_parts:
            st.caption(" · ".join(meta_parts))

        # Action row
        col_a, col_b, col_c, col_d, _ = st.columns([1.5, 1.5, 1.5, 1.5, 4])

        with col_a:
            load_key = f"load_{tab_key}_{card_idx}"
            if st.button("📄 Abstract", key=load_key, use_container_width=True):
                if pmid not in st.session_state.abstracts or not st.session_state.abstracts[pmid]:
                    with st.spinner("Loading…"):
                        ab = article.get("abstract") or fetch_abstract(pmid)
                        st.session_state.abstracts[pmid] = ab
                # toggle — if summary shown, clear it so only abstract shows
                if pmid in st.session_state.summaries:
                    del st.session_state.summaries[pmid]

        with col_b:
            sum_key = f"sum_{tab_key}_{card_idx}"
            if st.button("🧠 Summarise", key=sum_key, use_container_width=True):
                if pmid not in st.session_state.abstracts or not st.session_state.abstracts[pmid]:
                    with st.spinner("Loading abstract…"):
                        ab = article.get("abstract") or fetch_abstract(pmid)
                        st.session_state.abstracts[pmid] = ab
                with st.spinner("Summarising…"):
                    st.session_state.summaries[pmid] = summarize(
                        st.session_state.abstracts[pmid], title
                    )

        with col_c:
            save_key = f"save_{tab_key}_{card_idx}"
            already_saved = any(a.get("pmid") == pmid for a in st.session_state.saved_articles)
            if already_saved:
                st.button("✅ Saved", key=save_key, disabled=True, use_container_width=True)
            else:
                if st.button("💾 Save", key=save_key, use_container_width=True):
                    st.session_state.saved_articles.append(article)
                    st.rerun()

        with col_d:
            if url:
                st.link_button("🔗 Full Article", url, use_container_width=True)

        # Abstract pane
        if pmid in st.session_state.abstracts and st.session_state.abstracts[pmid]:
            if pmid not in st.session_state.summaries:
                with st.expander("Abstract", expanded=True):
                    raw = st.session_state.abstracts[pmid]
                    # Render **LABEL:** markdown nicely
                    st.markdown(raw)

        # Summary / nursing card pane
        if pmid in st.session_state.summaries:
            s = st.session_state.summaries[pmid]
            ev_c, ev_e, ev_d = s["evidence_level"]
            with st.expander("Nursing Summary Card", expanded=True):
                cols2 = st.columns(2)
                with cols2[0]:
                    st.markdown("**Overview / Background**")
                    st.write(s["overview"] or "—")
                    if s.get("methods"):
                        st.markdown("**Methods**")
                        st.write(s["methods"])
                with cols2[1]:
                    st.markdown("**Key Findings**")
                    st.write(s["key_findings"] or "—")
                    st.markdown("**Nursing Implications**")
                    st.info(s["nursing_implications"] or "—")

                st.markdown(f"**Evidence Level:** {ev_e} {ev_c} — {ev_d}")

                # APA citation
                st.markdown("**APA 7th Citation**")
                st.code(format_apa7(article), language=None)

        st.divider()


# ---------------------------------------------------------------------------
# Sidebar — filters
# ---------------------------------------------------------------------------
with st.sidebar:
    st.markdown("## 🔬 Search Filters")
    st.divider()

    databases = st.multiselect(
        "Databases",
        ["PubMed", "ClinicalTrials.gov"],
        default=["PubMed"],
    )

    date_range = st.radio(
        "Publication date",
        ["Last 1 year", "Last 2 years", "Last 5 years", "Last 10 years", "All time"],
        index=2,
    )

    study_types = st.multiselect(
        "Study design",
        ["Any", "Systematic Review", "Meta-Analysis", "RCT", "Cohort Study", "Case Study"],
        default=["Any"],
    )

    nursing_focus = st.toggle("Nursing-focused results only", value=True)
    max_results   = st.slider("Results per database", min_value=5, max_value=30, value=15)

    st.divider()
    st.markdown(
        """
**Evidence Level Guide**
🟢 Level I — Systematic Review / Meta-Analysis
🟡 Level II — RCT
🟠 Level III–IV — Quasi / Cohort
🔵 Level V–VI — Qualitative
⚫ Level VII — Expert Opinion
⚪ Unclassified
        """
    )


# ---------------------------------------------------------------------------
# Header
# ---------------------------------------------------------------------------
st.title("🩺 EBP Research Tool")
st.caption(
    "Evidence-Based Practice search for student nurses · "
    "PubMed (30 M+ articles) · ClinicalTrials.gov · Free & open"
)

# Quick-topic shortcuts
QUICK_TOPICS = [
    ("Wound Care",         "wound care nursing interventions"),
    ("Pain Management",    "pain management nursing practice"),
    ("Fall Prevention",    "fall prevention hospital nursing"),
    ("Med Safety",         "medication safety nursing errors"),
    ("Infection Control",  "hand hygiene infection control nursing"),
    ("Pt Education",       "patient education nursing outcomes"),
    ("Mental Health",      "mental health nursing interventions"),
    ("Paediatric Care",    "paediatric nursing care outcomes"),
]

qt_cols = st.columns(4)
for i, (label, query) in enumerate(QUICK_TOPICS):
    if qt_cols[i % 4].button(label, use_container_width=True, key=f"qt_{i}"):
        st.session_state.search_query = query
        st.rerun()

st.divider()

# ---------------------------------------------------------------------------
# Tabs
# ---------------------------------------------------------------------------
n_saved = len(st.session_state.saved_articles)
tab_search, tab_library, tab_cite = st.tabs(
    ["🔍 Search", f"📚 My Library ({n_saved})", "📝 Citation Builder"]
)

# ============================= SEARCH TAB ==================================
with tab_search:
    col_q, col_btn = st.columns([5, 1])
    with col_q:
        query = st.text_input(
            "search_box",
            value=st.session_state.search_query,
            placeholder="e.g. pressure injury prevention nursing ICU",
            label_visibility="collapsed",
        )
    with col_btn:
        do_search = st.button("Search", type="primary", use_container_width=True)

    if do_search and query.strip():
        st.session_state.search_query = query.strip()
        with st.spinner(f"Searching {', '.join(databases)}…"):
            results: list[dict] = []
            if "PubMed" in databases:
                results += search_pubmed(
                    query, date_range, study_types, nursing_focus, max_results
                )
            if "ClinicalTrials.gov" in databases:
                results += search_trials(query, max(5, max_results // 2))
            st.session_state.search_results = results

    results = st.session_state.search_results
    if results:
        st.success(f"**{len(results)}** results")
        for idx, art in enumerate(results):
            render_card(art, idx, "search")
    elif st.session_state.search_query:
        st.warning("No results found — try broader keywords or adjust filters.")
    else:
        st.info("Enter a topic above or choose a Quick Topic to begin.")

# ============================= LIBRARY TAB =================================
with tab_library:
    if not st.session_state.saved_articles:
        st.info("Your library is empty. Search and click 💾 Save to add articles here.")
    else:
        st.write(f"**{n_saved} saved article{'s' if n_saved != 1 else ''}**")

        # Export all citations
        if st.button("📋 Copy All Citations (APA 7th)"):
            all_cites = "\n\n".join(format_apa7(a) for a in st.session_state.saved_articles)
            st.code(all_cites, language=None)

        if st.button("🗑️ Clear Library", type="secondary"):
            st.session_state.saved_articles = []
            st.rerun()

        st.divider()
        for idx, art in enumerate(st.session_state.saved_articles):
            render_card(art, idx, "library")

# ============================= CITE BUILDER TAB ============================
with tab_cite:
    st.subheader("Citation Builder")
    st.write(
        "Enter a PubMed ID to instantly generate a formatted reference "
        "ready to paste into your assignment."
    )

    col1, col2, col3 = st.columns([2, 2, 1])
    with col1:
        pmid_input = st.text_input("PubMed ID (PMID)", placeholder="e.g. 33721172")
    with col2:
        cite_style = st.selectbox("Citation style", ["APA 7th Edition", "AMA"])
    with col3:
        st.write("")  # vertical align
        st.write("")
        gen_cite = st.button("Generate", type="primary", use_container_width=True)

    if gen_cite and pmid_input.strip():
        with st.spinner("Fetching article metadata…"):
            arts = fetch_summaries([pmid_input.strip()])
        if arts:
            art = arts[0]
            citation = format_apa7(art) if cite_style == "APA 7th Edition" else format_ama(art)
            st.success("Citation ready — copy the text below:")
            st.code(citation, language=None)
            st.markdown(
                f"**Title:** {art['title']}  \n"
                f"**Journal:** {art.get('journal','')} · **Year:** {art.get('year','')}  \n"
                f"[View on PubMed]({art.get('url','')})"
            )
        else:
            st.error("Article not found. Check the PMID and try again.")

    st.divider()
    st.markdown(
        """
**Formatting guide — APA 7th (nursing standard)**

> Author, F. M., & Author, F. M. (Year). Title of article. *Journal Name*, *Volume*(Issue), Pages. https://doi.org/xxxxx

**Tips for nursing assignments:**
- Always verify DOI links are active before submitting
- For articles with no DOI, include the journal homepage URL
- Use the exact journal abbreviation from the article, not your own
        """
    )

# ---------------------------------------------------------------------------
# Footer
# ---------------------------------------------------------------------------
st.divider()
st.caption(
    "Data sourced from PubMed / NCBI (NIH) and ClinicalTrials.gov — "
    "both public, freely available databases. "
    "Always verify information against original sources before clinical application."
)