ui/multi_report_page.py

import gradio as gr
import pandas as pd
import plotly.graph_objects as go

from ui.multi_report_functions import (
    load_multi_pdfs,
    run_coverage_scores,
    run_commitment_scores,
    run_positioning_chart,
    recalculate_from_revised_file,
    run_revised_positioning_chart
)


def build_multi_report_page():
    """
    Builds the Multi-Report Comparative Analysis page.
    Returns back_btn so interface.py can wire the Back navigation.
    This page is placed inside a gr.Tab by build_interface() rather than
    toggled via Column visibility, to avoid Gradio's known rendering bug
    where complex children (Dataframe, Plot, Gallery) can break after a
    parent Column's visibility is toggled off and back on.
    """
    back_btn = gr.Button("⬅ Back to Home", size="sm")

    gr.Markdown("# Multi-Report Comparative Analysis")
    gr.Markdown(
        "Upload multiple company sustainability reports to compare "
        "their ESG disclosure coverage and commitment levels."
    )

    # =========================
    # UPLOAD
    # =========================
    gr.Markdown("## Step 1: Upload Reports")

    multi_file = gr.File(
        label="Upload PDF Reports (you can select multiple files)",
        file_count="multiple"
    )
    load_multi_btn = gr.Button("Load Reports", variant="primary")
    upload_status = gr.Markdown()
    extracted_download = gr.File(label="Download Combined Extracted CSV")

    # =========================
    # COVERAGE SCORES
    # =========================
    gr.Markdown("## Step 2: Coverage Scores")
    gr.Markdown(
        "Measures how much of each report is sustainability-relevant: "
        "**Relevant ÷ (Relevant + Irrelevant) × 100**"
    )

    coverage_btn = gr.Button("Run Coverage Scores", variant="primary")
    coverage_status = gr.Markdown()
    coverage_table = gr.Dataframe(value=pd.DataFrame(), label="Coverage Scores (ranked)", interactive=False)
    coverage_download = gr.File(label="Download Coverage Scores CSV (per-report summary)")
    coverage_detail_download = gr.File(label="Download Per-Paragraph Detail CSV")

    # =========================
    # COMMITMENT SCORES
    # =========================
    gr.Markdown("## Step 3: Commitment Scores")
    gr.Markdown(
        "Measures the depth of disclosure among relevant paragraphs "
        "(quantitative claims are weighted higher than qualitative ones). "
        "Requires Coverage Scores to be run first."
    )

    commitment_btn = gr.Button("Run Commitment Scores", variant="primary")
    commitment_status = gr.Markdown()
    commitment_table = gr.Dataframe(value=pd.DataFrame(), label="Commitment Scores (ranked)", interactive=False)
    commitment_download = gr.File(label="Download Commitment Scores CSV (per-report summary)")
    commitment_detail_download = gr.File(label="Download Per-Paragraph Detail CSV (PM/FC retrieval + classification)")

    # ---------- Positioning Quadrant Chart ----------
    gr.Markdown("### Positioning")
    gr.Markdown(
        "Plots each company by Commitment Score (x-axis) vs Coverage Score "
        "(y-axis), split into four quadrants at the average of each score "
        "across all uploaded reports."
    )

    positioning_btn = gr.Button("Run Positioning", variant="primary")
    positioning_status = gr.Markdown()
    positioning_chart = gr.Plot(value=go.Figure(), label="Positioning")

    # =========================
    # REVISE & RECALCULATE
    # =========================
    gr.Markdown("## Step 4: Revise & Recalculate (Optional)")
    gr.Markdown(
        "If you reviewed the downloaded per-paragraph detail CSVs and "
        "corrected some predictions by hand, upload the revised file here "
        "to recompute Coverage Scores, Commitment Scores, and a new "
        "Positioning chart from it. The original results above are left "
        "untouched, so you can compare both side by side.\n\n"
        "Expected columns: `Doc_name`, `SA_label` (Relevant/Irrelevant), "
        "`QQ_label` (Qualitative/Quantitative), `PMFC_Label` "
        "(`Answer: PM` / `Answer: FC`). Rows with missing or unrecognized "
        "values in these columns are excluded from the relevant counts "
        "rather than causing an error."
    )

    revised_file = gr.File(label="Upload Revised Per-Paragraph CSV")
    recalculate_btn = gr.Button("Recalculate Scores", variant="primary")
    recalculate_status = gr.Markdown()

    revised_coverage_table = gr.Dataframe(value=pd.DataFrame(), label="Revised Coverage Scores (ranked)", interactive=False)
    revised_commitment_table = gr.Dataframe(value=pd.DataFrame(), label="Revised Commitment Scores (ranked)", interactive=False)

    revised_positioning_btn = gr.Button("Run Revised Positioning", variant="primary")
    revised_positioning_status = gr.Markdown()
    revised_positioning_chart = gr.Plot(value=go.Figure(), label="Positioning (Revised)")

    # =========================
    # CONNECTIONS
    # =========================
    load_multi_btn.click(
        load_multi_pdfs,
        inputs=multi_file,
        outputs=[extracted_download, upload_status]
    )

    coverage_btn.click(
        run_coverage_scores,
        outputs=[coverage_table, coverage_download, coverage_detail_download, coverage_status]
    )

    commitment_btn.click(
        run_commitment_scores,
        outputs=[commitment_table, commitment_download, commitment_detail_download, commitment_status]
    )

    positioning_btn.click(
        run_positioning_chart,
        outputs=[positioning_chart, positioning_status]
    )

    recalculate_btn.click(
        recalculate_from_revised_file,
        inputs=revised_file,
        outputs=[revised_coverage_table, revised_commitment_table, recalculate_status]
    )

    revised_positioning_btn.click(
        run_revised_positioning_chart,
        outputs=[revised_positioning_chart, revised_positioning_status]
    )

    return back_btn