Update rag.py

rag.py

@@ -533,136 +533,3 @@ class NurembergScholar:
     def clear_cache(self) -> None:
         self._cache.clear()
         print("  Cache cleared.")
-
-# ── Gradio UI ─────────────────────────────────────────────────────────────────
-
-def build_gradio_app(scholar: NurembergScholar):
-    import gradio as gr
-
-    def gradio_query(query: str, top_k: int):
-        if not query.strip():
-            return "Please enter a question.", "", ""
-        result  = scholar.answer(query, top_k=int(top_k))
-        answer  = result["answer"]
-        sources = _format_sources(result["sources"])
-        report  = _format_citation_report(
-            result["citation_report"],
-            result.get("cache_hit", False),
-        )
-        return answer, sources, report
-
-    def _format_sources(results) -> str:
-        if not results:
-            return "No sources retrieved."
-        lines = []
-        for i, r in enumerate(results, 1):
-            lines.append(
-                f"**[SOURCE {i}]** `{r.collection}` | {r.date_iso or '?'} | "
-                f"speaker: *{r.speaker or '-'}* | page {r.page_number or '?'} | "
-                f"rerank: `{f'{r.rerank_score:.4f}' if r.rerank_score is not None else 'n/a'}`\n\n"
-                f"> {r.body[:300]}..."
-            )
-        return "\n\n---\n\n".join(lines)
-
-    def _format_citation_report(report: dict, cache_hit: bool = False) -> str:
-        if not report:
-            return ""
-        cache_label = "HIT" if cache_hit else "MISS"
-        status      = "All citations valid" if report.get("clean") else "Issues found"
-        lines = [
-            f"**Citation check:** {status}   |   "
-            f"**Backend:** Groq ({GROQ_MODEL})   |   **Cache:** {cache_label}",
-        ]
-        if report.get("cited"):
-            lines.append(f"- Referenced: SOURCE {report['cited']}")
-        if report.get("hallucinated"):
-            lines.append(f"- Hallucinated refs stripped: {report['hallucinated']}")
-        if report.get("uncited_sources"):
-            lines.append(
-                f"- Retrieved but not cited: SOURCE {report['uncited_sources']}")
-        if report.get("uncited_sentences"):
-            lines.append(
-                f"- Paragraphs without citation: "
-                f"{len(report['uncited_sentences'])}")
-        stats = scholar.cache_stats
-        lines.append(
-            f"- Cache: {stats['size']} entries | "
-            f"{stats['hits']} hits / {stats['hits']+stats['misses']} queries "
-            f"({stats['hit_rate']:.0%})"
-        )
-        return "\n".join(lines)
-
-    example_queries = [
-        ["What did Goering say in his defense about the Luftwaffe?", 5],
-        ["How did the Tribunal define crimes against humanity under Article 6(c)?", 5],
-        ["What was Ohlendorf's confession regarding Einsatzgruppen killings?", 5],
-        ["What evidence was presented about the Final Solution?", 5],
-        ["What was the London Agreement and why was it significant?", 5],
-        ["How were the defendants sentenced on 1 October 1946?", 5],
-    ]
-
-    with gr.Blocks(title="Nuremberg Scholar", theme=gr.themes.Soft()) as app:
-        gr.Markdown(
-            """
-            # ⚖️ Nuremberg Scholar
-            **AI research assistant for the Nuremberg Trials (1945–1946)**
-
-            Answers are grounded exclusively in primary source documents.
-            Every factual claim is cited to a specific source passage.
-
-            *Llama-3.1-8B via Groq · BGE-M3 hybrid retrieval · 46,325 indexed passages*
-            """
-        )
-        with gr.Row():
-            with gr.Column(scale=3):
-                query_box = gr.Textbox(
-                    label       = "Your question",
-                    placeholder = "e.g. What did Speer claim about his knowledge of the Holocaust?",
-                    lines       = 2,
-                )
-                with gr.Row():
-                    top_k_slider = gr.Slider(
-                        minimum = 1, maximum = 10, value = 5, step = 1,
-                        label   = "Sources to retrieve",
-                    )
-                    submit_btn = gr.Button("Ask", variant="primary")
-                gr.Examples(
-                    examples = example_queries,
-                    inputs   = [query_box, top_k_slider],
-                    label    = "Example questions",
-                )
-            with gr.Column(scale=1):
-                gr.Markdown(
-                    """
-                    **Corpus**
-                    - 221 trial session transcripts
-                    - Full Tribunal judgment (Oct 1946)
-                    - Key prosecution documents
-                    - 46,325 indexed passages
-
-                    **Retrieval**
-                    Dense + sparse hybrid search,
-                    re-ranked by cross-encoder.
-
-                    **Citation policy**
-                    All claims must cite a SOURCE.
-                    Hallucinated citations are stripped.
-                    """
-                )
-
-        answer_box = gr.Markdown(label="Answer")
-
-        with gr.Accordion("Citation verification", open=False):
-            citation_box = gr.Markdown()
-
-        with gr.Accordion("Retrieved sources", open=False):
-            sources_box = gr.Markdown()
-
-        for trigger in (submit_btn.click, query_box.submit):
-            trigger(
-                fn      = gradio_query,
-                inputs  = [query_box, top_k_slider],
-                outputs = [answer_box, sources_box, citation_box],
-            )
-
-    return app