import gradio as gr

from src.pipeline import SigmaPipeline

# Initialize pipeline once during startup
pipeline = SigmaPipeline()


def analyze_text(text):
    """
    Run full SIGMA analysis pipeline and format results.
    """
    result = pipeline.analyze(text)

    formatted_output = []

    formatted_output.append("## EXPLICIT FACTS\n")

    for fact in result["explicit_facts"]:
        formatted_output.append(
            f"- **{fact['subject']}** → `{fact['action']}` → {fact['object']}"
        )

    formatted_output.append("\n## IMPLICIT INFERENCES\n")

    for inference in result["implicit_inferences"]:
        formatted_output.append(
            f"- **[{inference['label']}]** "
            f"{inference['hypothesis']}  \n"
            f"  Confidence: `{inference['confidence']}` "
            f"({inference['confidence_tier']})"
        )

    formatted_output.append("\n## ANALYTIC SUMMARY\n")

    formatted_output.append(result["summary"])

    return "\n".join(formatted_output)


demo = gr.Interface(
    fn=analyze_text,

    inputs=gr.Textbox(
        lines=12,
        label="Intelligence Report Input",
        placeholder="Paste intelligence-style narrative text here.\nClick Submit and wait for results to appear on the right panel."
    ),

    outputs=gr.Markdown(),

    title="SIGMA: Structured Intelligence & Grounded Meaning Analyzer",

    description=(
        "SIGMA extracts explicit facts, evaluates implicit inferences using Natural Language Inference (MNLI), and generates "
        "concise intelligence-style summaries using pretrained NLP models. "
        "For details, refer to [README.md](https://huggingface.co/spaces/morinousagi/nlp-intelligence-analyzer/blob/main/README.md)"
    ),

    show_progress="full"
)


if __name__ == "__main__":
    demo.queue()
    demo.launch()