Initial RAG & LangChain Explorer Space with Gradio 5.50.0

README.md

@@ -1,12 +1,36 @@
 ---
-title: Rag Langchain Explorer
-emoji: 🏃
-colorFrom: green
-colorTo: yellow
+title: RAG & LangChain Explorer
+emoji: 🔍
+colorFrom: blue
+colorTo: purple
 sdk: gradio
-sdk_version: 6.5.1
+sdk_version: "5.50.0"
 app_file: app.py
 pinned: false
+datasets:
+  - AYI-NEDJIMI/rag-langchain-fr
+  - AYI-NEDJIMI/rag-langchain-en
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# RAG & LangChain Explorer
+
+An interactive explorer for RAG (Retrieval-Augmented Generation) and LangChain components.
+
+Browse, search, and compare document loaders, text splitters, embedding models, vector stores, retrievers, and chains — in both French and English.
+
+## Features
+
+- **Explorer**: Searchable table with category filtering
+- **Details**: View full details for any component
+- **Q&A**: Quick answers about RAG & LangChain concepts
+- **Statistics**: Interactive Plotly charts by category and type
+- **Bilingual**: Toggle between French and English
+
+## Datasets
+
+- [AYI-NEDJIMI/rag-langchain-fr](https://huggingface.co/datasets/AYI-NEDJIMI/rag-langchain-fr)
+- [AYI-NEDJIMI/rag-langchain-en](https://huggingface.co/datasets/AYI-NEDJIMI/rag-langchain-en)
+
+---
+
+Built by [AYI-NEDJIMI Consultants](https://ayinedjimi-consultants.fr)

app.py

@@ -0,0 +1,338 @@
+import gradio as gr
+import pandas as pd
+import plotly.express as px
+import plotly.graph_objects as go
+from datasets import load_dataset
+
+# ---------------------------------------------------------------------------
+# Data loading
+# ---------------------------------------------------------------------------
+
+COLUMNS = [
+    "id", "type", "category", "name", "content",
+    "details", "pros", "cons", "tools", "source_url", "language",
+]
+
+def load_data():
+    """Load both FR and EN datasets and return as DataFrames."""
+    try:
+        ds_fr = load_dataset("AYI-NEDJIMI/rag-langchain-fr", split="train")
+        df_fr = ds_fr.to_pandas()
+    except Exception:
+        df_fr = pd.DataFrame(columns=COLUMNS)
+
+    try:
+        ds_en = load_dataset("AYI-NEDJIMI/rag-langchain-en", split="train")
+        df_en = ds_en.to_pandas()
+    except Exception:
+        df_en = pd.DataFrame(columns=COLUMNS)
+
+    return df_fr, df_en
+
+
+DF_FR, DF_EN = load_data()
+
+CATEGORIES = [
+    "document_loader", "text_splitter", "embedding_model",
+    "vector_store", "retriever", "chain",
+]
+
+LABELS = {
+    "FR": {
+        "title": "RAG & LangChain Explorer",
+        "explorer": "Explorateur",
+        "details": "Détails",
+        "qna": "Q&R",
+        "statistics": "Statistiques",
+        "search": "Rechercher…",
+        "category": "Catégorie",
+        "all": "Toutes",
+        "select_item": "Sélectionner un élément",
+        "no_results": "Aucun résultat.",
+        "by_category": "Répartition par catégorie",
+        "by_type": "Répartition par type",
+        "items_per_cat": "Nombre d'éléments par catégorie",
+        "ask": "Posez votre question sur RAG / LangChain",
+        "answer": "Réponse",
+        "qa_placeholder": "Ex : Qu'est-ce qu'un Text Splitter ?",
+        "no_match": "Aucune correspondance trouvée. Essayez un autre terme.",
+        "name_col": "Nom",
+        "type_col": "Type",
+        "category_col": "Catégorie",
+        "content_lbl": "Contenu",
+        "details_lbl": "Détails",
+        "pros_lbl": "Avantages",
+        "cons_lbl": "Inconvénients",
+        "tools_lbl": "Outils",
+        "source_lbl": "Source",
+    },
+    "EN": {
+        "title": "RAG & LangChain Explorer",
+        "explorer": "Explorer",
+        "details": "Details",
+        "qna": "Q&A",
+        "statistics": "Statistics",
+        "search": "Search…",
+        "category": "Category",
+        "all": "All",
+        "select_item": "Select an item",
+        "no_results": "No results.",
+        "by_category": "Distribution by category",
+        "by_type": "Distribution by type",
+        "items_per_cat": "Number of items per category",
+        "ask": "Ask a question about RAG / LangChain",
+        "answer": "Answer",
+        "qa_placeholder": "E.g. What is a Text Splitter?",
+        "no_match": "No match found. Try another term.",
+        "name_col": "Name",
+        "type_col": "Type",
+        "category_col": "Category",
+        "content_lbl": "Content",
+        "details_lbl": "Details",
+        "pros_lbl": "Pros",
+        "cons_lbl": "Cons",
+        "tools_lbl": "Tools",
+        "source_lbl": "Source",
+    },
+}
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+def get_df(lang: str) -> pd.DataFrame:
+    return DF_FR.copy() if lang == "FR" else DF_EN.copy()
+
+
+def filter_table(search: str, category: str, lang: str):
+    df = get_df(lang)
+    if df.empty:
+        return pd.DataFrame()
+    if category and category not in ("Toutes", "All"):
+        df = df[df["category"] == category]
+    if search:
+        mask = df.apply(
+            lambda r: search.lower() in " ".join(r.astype(str)).lower(), axis=1
+        )
+        df = df[mask]
+    display_cols = ["name", "type", "category"]
+    available = [c for c in display_cols if c in df.columns]
+    return df[available].reset_index(drop=True)
+
+
+def get_item_names(lang: str):
+    df = get_df(lang)
+    if df.empty:
+        return []
+    return sorted(df["name"].dropna().unique().tolist())
+
+
+def get_item_details(name: str, lang: str):
+    L = LABELS[lang]
+    df = get_df(lang)
+    if df.empty or not name:
+        return L["no_results"]
+    row = df[df["name"] == name]
+    if row.empty:
+        return L["no_results"]
+    r = row.iloc[0]
+    parts = []
+    parts.append(f"## {r.get('name', '')}")
+    parts.append(f"**{L['category_col']}**: {r.get('category', '')}")
+    parts.append(f"**{L['type_col']}**: {r.get('type', '')}")
+    if pd.notna(r.get("content")):
+        parts.append(f"\n### {L['content_lbl']}\n{r['content']}")
+    if pd.notna(r.get("details")):
+        parts.append(f"\n### {L['details_lbl']}\n{r['details']}")
+    if pd.notna(r.get("pros")):
+        parts.append(f"\n### {L['pros_lbl']}\n{r['pros']}")
+    if pd.notna(r.get("cons")):
+        parts.append(f"\n### {L['cons_lbl']}\n{r['cons']}")
+    if pd.notna(r.get("tools")):
+        parts.append(f"\n### {L['tools_lbl']}\n{r['tools']}")
+    if pd.notna(r.get("source_url")):
+        parts.append(f"\n### {L['source_lbl']}\n[Link]({r['source_url']})")
+    return "\n".join(parts)
+
+
+def answer_question(question: str, lang: str):
+    L = LABELS[lang]
+    if not question or not question.strip():
+        return ""
+    df = get_df(lang)
+    if df.empty:
+        return L["no_match"]
+    q = question.lower()
+    # Search across content-heavy columns
+    search_cols = ["name", "content", "details", "category", "type"]
+    scores = []
+    for _, row in df.iterrows():
+        text = " ".join(str(row.get(c, "")) for c in search_cols).lower()
+        score = sum(1 for w in q.split() if w in text)
+        scores.append(score)
+    df = df.copy()
+    df["_score"] = scores
+    best = df.sort_values("_score", ascending=False).head(3)
+    best = best[best["_score"] > 0]
+    if best.empty:
+        return L["no_match"]
+    parts = []
+    for _, r in best.iterrows():
+        parts.append(f"### {r.get('name', '')}")
+        parts.append(f"**{L['category_col']}**: {r.get('category', '')}")
+        if pd.notna(r.get("content")):
+            parts.append(f"{r['content'][:500]}")
+        parts.append("---")
+    return "\n".join(parts)
+
+
+def make_category_chart(lang: str):
+    L = LABELS[lang]
+    df = get_df(lang)
+    if df.empty:
+        return go.Figure()
+    counts = df["category"].value_counts().reset_index()
+    counts.columns = ["category", "count"]
+    fig = px.bar(
+        counts, x="category", y="count",
+        title=L["items_per_cat"],
+        color="category",
+        color_discrete_sequence=px.colors.qualitative.Set2,
+    )
+    fig.update_layout(showlegend=False, xaxis_title="", yaxis_title="")
+    return fig
+
+
+def make_type_chart(lang: str):
+    L = LABELS[lang]
+    df = get_df(lang)
+    if df.empty:
+        return go.Figure()
+    counts = df["type"].value_counts().reset_index()
+    counts.columns = ["type", "count"]
+    fig = px.pie(
+        counts, names="type", values="count",
+        title=L["by_type"],
+        color_discrete_sequence=px.colors.qualitative.Pastel,
+    )
+    return fig
+
+
+def make_category_pie(lang: str):
+    L = LABELS[lang]
+    df = get_df(lang)
+    if df.empty:
+        return go.Figure()
+    counts = df["category"].value_counts().reset_index()
+    counts.columns = ["category", "count"]
+    fig = px.pie(
+        counts, names="category", values="count",
+        title=L["by_category"],
+        color_discrete_sequence=px.colors.qualitative.Set2,
+    )
+    return fig
+
+
+# ---------------------------------------------------------------------------
+# UI
+# ---------------------------------------------------------------------------
+
+FOOTER_HTML = """
+<div style="text-align:center; padding:20px; margin-top:30px; border-top:1px solid #444; color:#888; font-size:0.9em;">
+    Built by <a href="https://ayinedjimi-consultants.fr" target="_blank"
+    style="color:#7c8aff; text-decoration:none;">AYI-NEDJIMI Consultants</a>
+</div>
+"""
+
+with gr.Blocks(
+    title="RAG & LangChain Explorer",
+    theme=gr.themes.Soft(primary_hue="blue", secondary_hue="purple"),
+) as demo:
+
+    gr.Markdown("# RAG & LangChain Explorer")
+
+    lang_toggle = gr.Radio(
+        choices=["FR", "EN"], value="FR", label="Language / Langue",
+        interactive=True,
+    )
+
+    with gr.Tabs():
+        # ---- Explorer tab ----
+        with gr.Tab("Explorer / Explorateur"):
+            with gr.Row():
+                search_box = gr.Textbox(
+                    label="Search…", placeholder="Search…", scale=3,
+                )
+                cat_filter = gr.Dropdown(
+                    choices=["Toutes"] + CATEGORIES,
+                    value="Toutes",
+                    label="Category",
+                    scale=1,
+                )
+            table_output = gr.Dataframe(
+                value=filter_table("", "Toutes", "FR"),
+                interactive=False,
+            )
+
+            search_box.change(
+                filter_table, [search_box, cat_filter, lang_toggle], table_output,
+            )
+            cat_filter.change(
+                filter_table, [search_box, cat_filter, lang_toggle], table_output,
+            )
+            lang_toggle.change(
+                filter_table, [search_box, cat_filter, lang_toggle], table_output,
+            )
+
+        # ---- Details tab ----
+        with gr.Tab("Details / Détails"):
+            item_dropdown = gr.Dropdown(
+                choices=get_item_names("FR"),
+                label="Select an item / Sélectionner un élément",
+                interactive=True,
+            )
+            detail_output = gr.Markdown()
+
+            item_dropdown.change(
+                get_item_details, [item_dropdown, lang_toggle], detail_output,
+            )
+
+            def refresh_names(lang):
+                return gr.update(choices=get_item_names(lang), value=None)
+
+            lang_toggle.change(refresh_names, lang_toggle, item_dropdown)
+
+        # ---- Q&A tab ----
+        with gr.Tab("Q&A / Q&R"):
+            qa_input = gr.Textbox(
+                label="Ask a question / Posez votre question",
+                placeholder="E.g. What is a Text Splitter?",
+                lines=2,
+            )
+            qa_btn = gr.Button("Search / Rechercher")
+            qa_output = gr.Markdown()
+
+            qa_btn.click(answer_question, [qa_input, lang_toggle], qa_output)
+            qa_input.submit(answer_question, [qa_input, lang_toggle], qa_output)
+
+        # ---- Statistics tab ----
+        with gr.Tab("Statistics / Statistiques"):
+            with gr.Row():
+                cat_bar = gr.Plot(value=make_category_chart("FR"))
+                cat_pie = gr.Plot(value=make_category_pie("FR"))
+            with gr.Row():
+                type_pie = gr.Plot(value=make_type_chart("FR"))
+
+            def refresh_stats(lang):
+                return (
+                    make_category_chart(lang),
+                    make_category_pie(lang),
+                    make_type_chart(lang),
+                )
+
+            lang_toggle.change(refresh_stats, lang_toggle, [cat_bar, cat_pie, type_pie])
+
+    gr.HTML(FOOTER_HTML)
+
+if __name__ == "__main__":
+    demo.launch()

requirements.txt

@@ -0,0 +1,4 @@
+gradio==5.50.0
+plotly
+pandas
+datasets