Deploy from GitHub (a639361)

.gitattributes

@@ -1,3 +1,3 @@
+data/*.faiss filter=lfs diff=lfs merge=lfs -text
 data/*.json filter=lfs diff=lfs merge=lfs -text
 data/*.db filter=lfs diff=lfs merge=lfs -text
-data/*.faiss filter=lfs diff=lfs merge=lfs -text

Dockerfile

@@ -10,10 +10,12 @@ RUN uv sync --frozen --no-dev
 COPY config.py ./
 COPY rag/ rag/
 
-# Pre-export ONNX models at build time (cached in image layer)
+# Pre-download and load quantized ONNX models at build time (cached in layer)
 RUN uv run python -c "\
 from rag.embeddings import load_embedding_model, load_cross_encoder; \
-load_embedding_model(); load_cross_encoder()"
+print('Loading embedding model...'); load_embedding_model(); \
+print('Loading cross-encoder...'); load_cross_encoder(); \
+print('Models cached.')"
 
 COPY app.py ./
 COPY templates/ templates/

README.md

@@ -9,7 +9,7 @@ pinned: false
 preload_from_hub:
   - sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2
   - cross-encoder/mmarco-mMiniLMv2-L12-H384-v1
-startup_duration_timeout: 5m
+startup_duration_timeout: 10m
 ---
 
 <!-- HuggingFace Spaces frontmatter above -- do not remove -->
@@ -79,6 +79,7 @@ The entire system runs locally with no external API calls, no paid dependencies,
 - **Fast** -- sub-2s response times on CPU
 - **35,000+ verses** -- complete French Bible (AELF translation)
 - **PWA-ready** -- offline support via service worker, installable on mobile
+- **Per-verse feedback** -- thumbs up/down on results, synced to HuggingFace Dataset
 - **Self-contained** -- no external APIs, runs entirely on your machine
 
 ## Live Demo
@@ -112,6 +113,7 @@ Open [http://localhost:8000](http://localhost:8000) in your browser.
 | GET    | `/health`      | Health check (200 `ok` or 503 `loading`) |
 | GET    | `/robots.txt`  | Robots.txt for crawlers                  |
 | GET    | `/sitemap.xml` | XML sitemap for crawlers                 |
+| POST   | `/feedback`    | Per-verse feedback (fire-and-forget, returns 204) |
 
 ## Architecture
 

config.py

@@ -1,6 +1,7 @@
 """Central configuration for the RAG Bible pipeline."""
 
 import os
+import platform
 from pathlib import Path
 
 # Paths
@@ -16,6 +17,15 @@ EMBEDDING_DIMENSION: int = 384
 # Cross-encoder model
 CROSS_ENCODER_MODEL: str = "cross-encoder/mmarco-mMiniLMv2-L12-H384-v1"
 
+# ONNX quantized model file (architecture-specific)
+_machine = platform.machine()
+_onnx_map: dict[str, str] = {
+    "x86_64": "onnx/model_qint8_avx512.onnx",
+    "AMD64": "onnx/model_qint8_avx512.onnx",
+    "arm64": "onnx/model_qint8_arm64.onnx",
+}
+ONNX_FILE_NAME: str = _onnx_map.get(_machine, "onnx/model.onnx")
+
 # Retrieval parameters
 FAISS_TOP_K: int = 20
 RERANK_TOP_K: int = 5

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "rag-bible"
-version = "1.1.0"
+version = "1.1.1"
 description = "French Bible RAG system with FAISS + cross-encoder reranking"
 requires-python = ">=3.12"
 dependencies = [

rag/embeddings.py

@@ -20,7 +20,11 @@ def load_embedding_model(model_name: str | None = None) -> SentenceTransformer:
         Loaded embedding model.
     """
     name = model_name or config.EMBEDDING_MODEL
-    return SentenceTransformer(name, backend="onnx")
+    return SentenceTransformer(
+        name,
+        backend="onnx",
+        model_kwargs={"file_name": config.ONNX_FILE_NAME},
+    )
 
 
 def encode_texts(
@@ -71,5 +75,9 @@ def load_cross_encoder(model_name: str | None = None) -> CrossEncoder:
         Loaded cross-encoder model.
     """
     name = model_name or config.CROSS_ENCODER_MODEL
-    model: CrossEncoder = CrossEncoder(name, backend="onnx")
+    model: CrossEncoder = CrossEncoder(
+        name,
+        backend="onnx",
+        model_kwargs={"file_name": config.ONNX_FILE_NAME},
+    )
     return model

rag/retrieve.py

@@ -1,7 +1,6 @@
 """Two-stage retrieval: FAISS vector search + cross-encoder reranking."""
 
 import json
-from concurrent.futures import ThreadPoolExecutor
 from pathlib import Path
 from typing import Any
 
@@ -50,16 +49,10 @@ def load_pipeline(
     idx_path = index_path or config.INDEX_PATH
     map_path = mapping_path or config.MAPPING_PATH
 
-    with ThreadPoolExecutor(max_workers=4) as pool:
-        f_index = pool.submit(lambda: faiss.read_index(str(idx_path)))
-        f_mapping = pool.submit(lambda: _load_mapping(map_path))
-        f_embed = pool.submit(load_embedding_model)
-        f_cross = pool.submit(load_cross_encoder)
-
-        index = f_index.result()
-        mapping = f_mapping.result()
-        embed_model = f_embed.result()
-        cross_encoder = f_cross.result()
+    index = faiss.read_index(str(idx_path))
+    mapping = _load_mapping(map_path)
+    embed_model = load_embedding_model()
+    cross_encoder = load_cross_encoder()
 
     return index, mapping, embed_model, cross_encoder
 

static/index.html

@@ -65,7 +65,7 @@
     <h2 class="sidebar-title">Historique</h2>
     <div class="history-list"></div>
     <p class="history-empty">Vos recherches apparaîtront ici</p>
-    <p class="sidebar-version">v1.1.0</p>
+    <p class="sidebar-version">v1.1.1</p>
   </aside>
 
   <div class="hero-section">

static/styles.css

@@ -564,43 +564,51 @@ h1 {
   color: var(--color-text-card);
 }
 
-/* --- Card Footer / Feedback --- */
+/* --- Header Actions / Feedback --- */
 
-.card-footer {
+.card-header-actions {
   display: flex;
-  justify-content: flex-end;
+  align-items: center;
   gap: var(--space-xs);
-  padding-top: var(--space-sm);
-  border-top: 1px solid var(--color-border);
-  margin-top: var(--space-sm);
+}
+
+.feedback-group {
+  display: flex;
+  align-items: center;
+  gap: 2px;
 }
 
 .feedback-btn {
   display: flex;
   align-items: center;
   justify-content: center;
-  width: var(--touch-target-desktop);
-  height: var(--touch-target-desktop);
+  width: 28px;
+  height: 28px;
   padding: 0;
   border: none;
   border-radius: var(--radius-sm);
   background: transparent;
   color: var(--color-text-muted);
   cursor: pointer;
+  opacity: 0.5;
   transition: color var(--transition-quick),
-              background-color var(--transition-quick);
+              background-color var(--transition-quick),
+              opacity var(--transition-quick);
 }
 
 .feedback-btn:hover {
   background: var(--color-highlight);
+  opacity: 1;
 }
 
 .feedback-up[aria-pressed="true"] {
   color: var(--color-score-high);
+  opacity: 1;
 }
 
 .feedback-down[aria-pressed="true"] {
   color: var(--color-feedback-error);
+  opacity: 1;
 }
 
 /* --- Status Messages --- */
@@ -1018,8 +1026,8 @@ textarea:focus-visible {
   }
 
   .feedback-btn {
-    width: var(--touch-target-mobile);
-    height: var(--touch-target-mobile);
+    width: 32px;
+    height: 32px;
   }
 
   .example-query {

templates/results.html

@@ -19,32 +19,34 @@
                data-score="{{ r.score }}">
             <div class="card-header">
               <span><strong>{{ r.book_title }}</strong> — {{ r.chapter }}{% if r.verse %}:{{ r.verse }}{% endif %}</span>
-              <span class="score-badge
-                {%- if r.pct >= 80 %} score-high
-                {%- elif r.pct >= 50 %} score-mid
-                {%- elif r.pct >= 30 %} score-low
-                {%- else %} score-minimal
-                {%- endif %}">{{ r.label }} ({{ r.pct }}%)</span>
+              <div class="card-header-actions">
+                <span class="score-badge
+                  {%- if r.pct >= 80 %} score-high
+                  {%- elif r.pct >= 50 %} score-mid
+                  {%- elif r.pct >= 30 %} score-low
+                  {%- else %} score-minimal
+                  {%- endif %}">{{ r.label }} ({{ r.pct }}%)</span>
+                <div class="feedback-group">
+                  <button type="button" class="feedback-btn feedback-up"
+                          data-feedback="up" aria-label="Pertinent" aria-pressed="false">
+                    <svg width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor"
+                         stroke-width="2.5" stroke-linecap="round" stroke-linejoin="round" aria-hidden="true">
+                      <path d="M14 9V5a3 3 0 0 0-3-3l-4 9v11h11.28a2 2 0 0 0 2-1.7l1.38-9a2 2 0 0 0-2-2.3H14z"/>
+                      <path d="M7 22H4a2 2 0 0 1-2-2v-7a2 2 0 0 1 2-2h3"/>
+                    </svg>
+                  </button>
+                  <button type="button" class="feedback-btn feedback-down"
+                          data-feedback="down" aria-label="Non pertinent" aria-pressed="false">
+                    <svg width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor"
+                         stroke-width="2.5" stroke-linecap="round" stroke-linejoin="round" aria-hidden="true">
+                      <path d="M10 15v4a3 3 0 0 0 3 3l4-9V2H5.72a2 2 0 0 0-2 1.7l-1.38 9a2 2 0 0 0 2 2.3H10z"/>
+                      <path d="M17 2h3a2 2 0 0 1 2 2v7a2 2 0 0 1-2 2h-3"/>
+                    </svg>
+                  </button>
+                </div>
+              </div>
             </div>
             {% include "context_verses.html" %}
-            <div class="card-footer">
-              <button type="button" class="feedback-btn feedback-up"
-                      data-feedback="up" aria-label="Pertinent" aria-pressed="false">
-                <svg width="18" height="18" viewBox="0 0 24 24" fill="none" stroke="currentColor"
-                     stroke-width="2" stroke-linecap="round" stroke-linejoin="round" aria-hidden="true">
-                  <path d="M14 9V5a3 3 0 0 0-3-3l-4 9v11h11.28a2 2 0 0 0 2-1.7l1.38-9a2 2 0 0 0-2-2.3H14z"/>
-                  <path d="M7 22H4a2 2 0 0 1-2-2v-7a2 2 0 0 1 2-2h3"/>
-                </svg>
-              </button>
-              <button type="button" class="feedback-btn feedback-down"
-                      data-feedback="down" aria-label="Non pertinent" aria-pressed="false">
-                <svg width="18" height="18" viewBox="0 0 24 24" fill="none" stroke="currentColor"
-                     stroke-width="2" stroke-linecap="round" stroke-linejoin="round" aria-hidden="true">
-                  <path d="M10 15v4a3 3 0 0 0 3 3l4-9V2H5.72a2 2 0 0 0-2 1.7l-1.38 9a2 2 0 0 0 2 2.3H10z"/>
-                  <path d="M17 2h3a2 2 0 0 1 2 2v7a2 2 0 0 1-2 2h-3"/>
-                </svg>
-              </button>
-            </div>
           </div>
         </div>
         {% endfor %}