visual-rag-toolkit 0.1.3__py3-none-any.whl → 0.1.5__py3-none-any.whl

demo/__init__.py

@@ -7,4 +7,4 @@ A Streamlit-based UI for:
 - Interactive playground for visual search
 """
 
-__version__ = "0.1.0"
+__version__ = "0.1.5"

demo/ui/sidebar.py

@@ -16,6 +16,28 @@ from demo.qdrant_utils import (
 
 
 def render_sidebar():
+    # CSS to make sidebar metrics smaller
+    st.markdown("""
+    <style>
+    /* Smaller metrics in sidebar */
+    [data-testid="stSidebar"] [data-testid="stMetricValue"] {
+        font-size: 1.2rem !important;
+    }
+    [data-testid="stSidebar"] [data-testid="stMetricLabel"] {
+        font-size: 0.75rem !important;
+    }
+    /* Smaller expander headers in sidebar */
+    [data-testid="stSidebar"] [data-testid="stExpander"] summary {
+        font-size: 0.9rem !important;
+    }
+    /* Compact subheaders */
+    [data-testid="stSidebar"] h3 {
+        font-size: 1rem !important;
+        margin-bottom: 0.5rem !important;
+    }
+    </style>
+    """, unsafe_allow_html=True)
+    
     with st.sidebar:
         st.subheader("🔑 Qdrant Credentials")
         

demo/ui/upload.py

@@ -449,7 +449,7 @@ def process_pdfs(uploaded_files, config):
         
         if total_uploaded > 0:
             st.session_state["upload_success"] = f"Uploaded {total_uploaded} pages to {collection_name}"
-            st.balloons()
+            st.rerun()  # Immediately refresh to show success toast + balloons
             
     except Exception as e:
         st.error(f"❌ Processing error: {e}")

visual_rag/__init__.py

@@ -33,17 +33,17 @@ Each component works independently - use only what you need.
 
 import logging
 
-__version__ = "0.1.3"
+__version__ = "0.1.5"
 
 
 def setup_logging(level: str = "INFO", format: str = None) -> None:
     """
     Configure logging for visual_rag package.
-    
+
     Args:
         level: Log level ("DEBUG", "INFO", "WARNING", "ERROR")
         format: Custom format string. Default shows time, level, and message.
-    
+
     Example:
         >>> import visual_rag
         >>> visual_rag.setup_logging("INFO")
@@ -51,13 +51,13 @@ def setup_logging(level: str = "INFO", format: str = None) -> None:
     """
     if format is None:
         format = "[%(asctime)s] %(levelname)s - %(message)s"
-    
+
     logging.basicConfig(
         level=getattr(logging, level.upper(), logging.INFO),
         format=format,
         datefmt="%H:%M:%S",
     )
-    
+
     # Also set the visual_rag logger specifically
     logger = logging.getLogger("visual_rag")
     logger.setLevel(getattr(logging, level.upper(), logging.INFO))
@@ -111,13 +111,16 @@ try:
 except ImportError:
     QdrantAdmin = None
 
-try:
-    from visual_rag.demo_runner import demo
-except ImportError:
-    demo = None
+# demo is lazily imported to avoid RuntimeWarning when running as __main__
+# Access via visual_rag.demo() which triggers __getattr__
 
 # Config utilities (always available)
-from visual_rag.config import get, get_section, load_config
+try:
+    from visual_rag.config import get, get_section, load_config
+except ImportError:
+    get = None
+    get_section = None
+    load_config = None
 
 __all__ = [
     # Version
@@ -138,3 +141,15 @@ __all__ = [
     # Logging
     "setup_logging",
 ]
+
+
+def __getattr__(name: str):
+    """Lazy import for demo to avoid RuntimeWarning when running as __main__."""
+    if name == "demo":
+        try:
+            from visual_rag.demo_runner import demo
+
+            return demo
+        except ImportError:
+            return None
+    raise AttributeError(f"module 'visual_rag' has no attribute {name!r}")

visual_rag/cli/main.py

@@ -288,7 +288,7 @@ def cmd_search(args):
         )
 
     # Search
-    query_np = query_embedding.detach().cpu().numpy()
+    query_np = query_embedding.detach().cpu().float().numpy()  # .float() for BFloat16
     if args.strategy == "single_full":
         results = single_stage.search(
             query_embedding=query_np,

visual_rag/demo_runner.py

@@ -52,13 +52,11 @@ def demo(
     cmd = [sys.executable, "-m", "streamlit", "run", str(app_path)]
     cmd += ["--server.address", str(host)]
     cmd += ["--server.port", str(int(port))]
-    cmd += ["--server.headless", "true" if headless else "false"]
+    # headless=true prevents browser from auto-opening; open_browser overrides
+    should_be_headless = headless and not open_browser
+    cmd += ["--server.headless", "true" if should_be_headless else "false"]
     cmd += ["--browser.gatherUsageStats", "false"]
     cmd += ["--server.runOnSave", "false"]
-    cmd += ["--browser.serverAddress", str(host)]
-    if not open_browser:
-        cmd += ["--browser.serverPort", str(int(port))]
-        cmd += ["--browser.open", "false"]
 
     if extra_args:
         cmd += list(extra_args)

visual_rag/indexing/qdrant_indexer.py

@@ -335,7 +335,9 @@ class QdrantIndexer:
                 return val.tolist()
             return val
 
-        def _build_qdrant_points(batch_points: List[Dict[str, Any]]) -> List[qdrant_models.PointStruct]:
+        def _build_qdrant_points(
+            batch_points: List[Dict[str, Any]],
+        ) -> List[qdrant_models.PointStruct]:
             qdrant_points: List[qdrant_models.PointStruct] = []
             for p in batch_points:
                 global_pooled = p.get("global_pooled_embedding")
@@ -382,7 +384,7 @@ class QdrantIndexer:
                 )
 
                 logger.info(f"   ✅ Uploaded {len(points)} points to Qdrant")
-                
+
                 if delay_between_batches > 0:
                     if _is_cancelled():
                         return 0

visual_rag/retrieval/multi_vector.py

@@ -56,14 +56,10 @@ class MultiVectorRetriever:
                 )
 
             qdrant_url = (
-                qdrant_url
-                or os.getenv("QDRANT_URL")
-                or os.getenv("SIGIR_QDRANT_URL")  # legacy
+                qdrant_url or os.getenv("QDRANT_URL") or os.getenv("SIGIR_QDRANT_URL")  # legacy
             )
             if not qdrant_url:
-                raise ValueError(
-                    "QDRANT_URL is required (pass qdrant_url or set env var)."
-                )
+                raise ValueError("QDRANT_URL is required (pass qdrant_url or set env var).")
 
             qdrant_api_key = (
                 qdrant_api_key
@@ -97,7 +93,10 @@ class MultiVectorRetriever:
                     _ = client.get_collections()
                 except Exception as e:
                     msg = str(e)
-                    if "StatusCode.PERMISSION_DENIED" in msg or "http2 header with status: 403" in msg:
+                    if (
+                        "StatusCode.PERMISSION_DENIED" in msg
+                        or "http2 header with status: 403" in msg
+                    ):
                         client = _make_client(False)
                     else:
                         raise
@@ -158,9 +157,10 @@ class MultiVectorRetriever:
     ) -> List[Dict[str, Any]]:
         q = self.embedder.embed_query(query)
         if isinstance(q, torch.Tensor):
-            query_embedding = q.detach().cpu().numpy()
+            # .float() converts BFloat16 to Float32 (numpy doesn't support BFloat16)
+            query_embedding = q.detach().cpu().float().numpy()
         else:
-            query_embedding = np.asarray(q)
+            query_embedding = np.asarray(q, dtype=np.float32)
 
         return self.search_embedded(
             query_embedding=query_embedding,

visual_rag/retrieval/single_stage.py

@@ -129,5 +129,5 @@ class SingleStageRetriever:
         if isinstance(embedding, torch.Tensor):
             if embedding.dtype == torch.bfloat16:
                 return embedding.cpu().float().numpy()
-            return embedding.cpu().numpy()
+            return embedding.cpu().float().numpy()  # .float() for BFloat16 compatibility
         return np.array(embedding, dtype=np.float32)

visual_rag/retrieval/three_stage.py

@@ -51,7 +51,7 @@ class ThreeStageRetriever:
         if isinstance(embedding, torch.Tensor):
             if embedding.dtype == torch.bfloat16:
                 return embedding.cpu().float().numpy()
-            return embedding.cpu().numpy()
+            return embedding.cpu().float().numpy()  # .float() for BFloat16 compatibility
         return np.array(embedding, dtype=np.float32)
 
     def _infer_vector_is_multivector(self, vector_name: str) -> bool:

visual_rag/retrieval/two_stage.py

@@ -22,7 +22,6 @@ from typing import Any, Dict, List, Optional, Union
 
 import numpy as np
 import torch
-
 from qdrant_client.http import models as qdrant_models
 from qdrant_client.models import FieldCondition, Filter, MatchAny, MatchValue
 
@@ -419,7 +418,7 @@ class TwoStageRetriever:
         if isinstance(embedding, torch.Tensor):
             if embedding.dtype == torch.bfloat16:
                 return embedding.cpu().float().numpy()
-            return embedding.cpu().numpy()
+            return embedding.cpu().float().numpy()  # .float() for BFloat16 compatibility
         return np.array(embedding, dtype=np.float32)
 
     def build_filter(

visual_rag/visualization/saliency.py

@@ -55,14 +55,14 @@ def generate_saliency_map(
     if hasattr(query_embedding, "numpy"):
         query_np = query_embedding.numpy()
     elif hasattr(query_embedding, "cpu"):
-        query_np = query_embedding.cpu().numpy()
+        query_np = query_embedding.cpu().float().numpy()  # .float() for BFloat16
     else:
         query_np = np.array(query_embedding, dtype=np.float32)
 
     if hasattr(doc_embedding, "numpy"):
         doc_np = doc_embedding.numpy()
     elif hasattr(doc_embedding, "cpu"):
-        doc_np = doc_embedding.cpu().numpy()
+        doc_np = doc_embedding.cpu().float().numpy()  # .float() for BFloat16
     else:
         doc_np = np.array(doc_embedding, dtype=np.float32)
 

{visual_rag_toolkit-0.1.3.dist-info → visual_rag_toolkit-0.1.5.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: visual-rag-toolkit
-Version: 0.1.3
+Version: 0.1.5
 Summary: End-to-end visual document retrieval with ColPali, featuring two-stage pooling for scalable search
 Project-URL: Homepage, https://github.com/Ara-Yeroyan/visual-rag-toolkit
 Project-URL: Documentation, https://github.com/Ara-Yeroyan/visual-rag-toolkit#readme
@@ -88,9 +88,12 @@ Description-Content-Type: text/markdown
 [![PyPI](https://img.shields.io/pypi/v/visual-rag-toolkit)](https://pypi.org/project/visual-rag-toolkit/)
 [![Python](https://img.shields.io/pypi/pyversions/visual-rag-toolkit)](https://pypi.org/project/visual-rag-toolkit/)
 [![License](https://img.shields.io/pypi/l/visual-rag-toolkit)](LICENSE)
+[![Demo](https://img.shields.io/badge/Demo-Hugging%20Face-yellow)](https://huggingface.co/spaces/Yeroyan/visual-rag-toolkit)
 
 End-to-end visual document retrieval toolkit featuring **fast multi-stage retrieval** (prefetch with pooled vectors + exact MaxSim reranking).
 
+**[Try the Live Demo](https://huggingface.co/spaces/Yeroyan/visual-rag-toolkit)** - Upload PDFs, index to Qdrant, and query with visual retrieval.
+
 This repo contains:
 - a **Python package** (`visual_rag`)
 - a **Streamlit demo app** (`demo/`)
@@ -368,7 +371,7 @@ If you use this toolkit in your research, please cite:
 
 ```bibtex
 @software{visual_rag_toolkit,
-  title = {Visual RAG Toolkit: Scalable Visual Document Retrieval with Two-Stage Pooling},
+  title = {Visual RAG Toolkit: Scalable Visual Document Retrieval with 1D Convolutional Pooling},
   author = {Ara Yeroyan},
   year = {2026},
   url = {https://github.com/Ara-Yeroyan/visual-rag-toolkit}

{visual_rag_toolkit-0.1.3.dist-info → visual_rag_toolkit-0.1.5.dist-info}/RECORD

@@ -11,13 +11,12 @@ benchmarks/vidore_tatdqa_test/dataset_loader.py,sha256=gCCneGAKWQm0WlJHLvGjoMrAb
 benchmarks/vidore_tatdqa_test/metrics.py,sha256=cLdYbRt5VcxInO1cN79ve6ZLP3kaSxRkdzRX3IbPPMs,1112
 benchmarks/vidore_tatdqa_test/run_qdrant.py,sha256=_PikeqIYpWPim-KEQOwvT-aqwYoAWASjqJVisi8PfQg,28681
 benchmarks/vidore_tatdqa_test/sweep_eval.py,sha256=d_kbyNTJ1LoFfIVnsZyiRO1nKyMqmRB5jEweZL6kYd4,12688
-demo/__init__.py,sha256=jVzjsVKZl5ZZuFxawA8Pxj3yuIKL7llkao3rBpde-aQ,204
+demo/__init__.py,sha256=zxRkBWe4opMZXfJxNULuktXdZLuYsZFHeaijgxPe07w,204
 demo/app.py,sha256=nZbCz1mpRK-GZTgOHyz4m4AfgKFgsH-09JwXeL3d3ng,1405
 demo/commands.py,sha256=qxRE2x610yZvcjwEfSKiR9CyFonX-vRxFqQNJCUKfyA,13690
 demo/config.py,sha256=BNkV4NSEEMIV9e6Z-cxds2v247uVmTPCgL-M5ItPzMg,757
 demo/download_models.py,sha256=J10qQt2TpEshVOxvCX_ZSbV7YozIBqDATZnt8fUKFHs,2868
 demo/evaluation.py,sha256=4ixJGg50KAVNiZ_mr5FMVv-QKCrZRooJ80LbrjKXM1s,27467
-demo/example_metadata_mapping_sigir.json,sha256=UCgqZtr6Wnq_vS7zxPxpvuokk9gxOVgKydC7f1lauw8,824
 demo/indexing.py,sha256=qUVEB3QrIolS53Ggxurccbh-QyeLLbzcY5TLyVBVKME,10620
 demo/qdrant_utils.py,sha256=Xh-thLIrACrYkFCrqazYNH0p3vS8_yMCaTbvt4HAy98,7778
 demo/results.py,sha256=dprvxnyHwxJvkAQuh4deaCsiEG1wm0n9svPyxI37vJg,1050
@@ -26,14 +25,14 @@ demo/ui/__init__.py,sha256=EyBCvnXYfPbdyxJzyp9TjQBeJJUgmOY1yRHkUeC6JFQ,412
 demo/ui/benchmark.py,sha256=HiGCN4HrqeOC7L6t2kuzIiyWdcVE_cP2JTxoewrmPSo,14218
 demo/ui/header.py,sha256=J2hXr_nNyg1H9rmrd-EGx3WUl7lYo-Ca30ptgzBCfBs,806
 demo/ui/playground.py,sha256=yRlWWzJgsc596vALn5f0PHhmhtJCMmfv61nYakW75GQ,13672
-demo/ui/sidebar.py,sha256=DLVhEj-8xAJCXUwOhndNv8ZFT4K3u8iE6FVOoH-jRuA,7699
-demo/ui/upload.py,sha256=6iv4xDsacMtUF1FrquRBE_xNb92HevgxCMS0LBK4Ay0,20455
-visual_rag/__init__.py,sha256=4NksVCaN_p32ezMF1N-oxpPFKeOm8xRo70VC4OSa2a0,3911
+demo/ui/sidebar.py,sha256=k7CZ5V7SxwbPOL9nQTWUct5y9mPY-YnLIeBgha7aIEE,8379
+demo/ui/upload.py,sha256=NVmd1-xRj_dfgq14x_eqcN6MjDAZ3hbPBD0YRUz8OOY,20508
+visual_rag/__init__.py,sha256=mOcAcA2rkPVA9GQ63CKFb1g3KatGqILsjfWFj4kkpus,4384
 visual_rag/config.py,sha256=qqSQk2lM5MiRji-6xQNGS2gSiXA4NgyJnCbgGx7uGJQ,7395
-visual_rag/demo_runner.py,sha256=wi0Wz3gZ39l4aovMd6zURq_CKUSgma4kGjF6hpQHwGY,2793
+visual_rag/demo_runner.py,sha256=ahIKYhHQSLOF_kWbj0u_29SLX4qTdyxb3Smt_cMMfzI,2763
 visual_rag/qdrant_admin.py,sha256=NNczko2S5-K3qATNUxgYn51hNWgWb6boheL7vlCQGpM,7055
 visual_rag/cli/__init__.py,sha256=WgBRXm0VACfLltvVlLcSs3FTM1uQ7Uuw3CVD4-zWZwc,46
-visual_rag/cli/main.py,sha256=QmpnQ0lbC6Q9lwxaSCDh6paEEzI78IPY1jwc3_9y7VI,21083
+visual_rag/cli/main.py,sha256=lw2xS96FtdzFnftModGcqRyJYlxFWA0U3aE3C247mBk,21116
 visual_rag/embedding/__init__.py,sha256=7QIENmxwRnwnUzsYKRY3VQTyF3HJkRiL1D7Au9XHF0w,682
 visual_rag/embedding/pooling.py,sha256=x8uY4VHbxEnsJRM2JeOkzPHDiwOkbi5NK4XW21U1hAc,11401
 visual_rag/embedding/visual_embedder.py,sha256=he9JpVHmo_szOiXCwtJdrCseGmf2y5Gi0UEFjwazzVY,23198
@@ -41,18 +40,18 @@ visual_rag/indexing/__init__.py,sha256=rloBEBt3x8BQut1Tj1n8fuaQ3iXMS3pm64o8n-NlS
 visual_rag/indexing/cloudinary_uploader.py,sha256=e-G5du4D7z6mWWl2lahMidG-Wdc-baImFFILTojebpA,8826
 visual_rag/indexing/pdf_processor.py,sha256=V3RAKpwgIFicqUaXzaaljePxh_oP4UV5W0aiJyfv0BY,10247
 visual_rag/indexing/pipeline.py,sha256=1ScpVRlLCq2FWi3IPvlQcIfDCQQ2F64IlRd9ZZHiTaA,25037
-visual_rag/indexing/qdrant_indexer.py,sha256=Q0e8JCr9B1OxgOMW7BWeg7MlWiLPaBUmjoFof4gZFYo,19519
+visual_rag/indexing/qdrant_indexer.py,sha256=E9uDwOL3A1nFJhFhJcI8V3JR8FL7nDahEYiv886oG1s,19526
 visual_rag/preprocessing/__init__.py,sha256=rCzfBO0jaVKp6MpPRRused_4gasHfobAbG-139Y806E,121
 visual_rag/preprocessing/crop_empty.py,sha256=iHXITFkRlF40VPJ4k9d432RUAi_89BhAEvK4wOEn96Q,5211
 visual_rag/retrieval/__init__.py,sha256=J9pnbeB83Fqs9n4g3GcIp1VR9dnuyAlcsIDVsf0lSb8,601
-visual_rag/retrieval/multi_vector.py,sha256=ZZ_O4x7MZbhF--kRp8T4UJG5GuenfjJ91FKicklhK3Q,7006
-visual_rag/retrieval/single_stage.py,sha256=Ba06V-KRSFSZm0xzbjFR3EBEWaQkDo7U_pWNx25W8H0,4425
-visual_rag/retrieval/three_stage.py,sha256=YC0CVEohxTT5zhilcQHI7nYAk08E5jC3zkQ3-rNdLMw,5951
-visual_rag/retrieval/two_stage.py,sha256=JJ6rXv_3_3WLIjAcxOY7NuhSyuPzIMyHf3ooiGFTp9k,16776
+visual_rag/retrieval/multi_vector.py,sha256=oQowl-Yv4cYDj2jroX1nv_uZl8sb0d5VKMmwxL-GlBA,7119
+visual_rag/retrieval/single_stage.py,sha256=-hljRha6kWZAr1IQJVSii4pqiIqk0cHhW68bUyYK41o,4472
+visual_rag/retrieval/three_stage.py,sha256=UDjH92V6ILV_w4bOeN0vCF3SdXYHMml9oN2gl3kfwWk,5998
+visual_rag/retrieval/two_stage.py,sha256=OTTH-MYtwK45RHYMF1esEeciXMTU8h7AQzH68BZVrYE,16822
 visual_rag/visualization/__init__.py,sha256=SITKNvBEseDp7F3K6UzLPA-6OQFqYfY5azS5nlDdihQ,447
-visual_rag/visualization/saliency.py,sha256=F3Plc18Sf3tzWcyncuaruTmENm1IfW5j9NFGEQR93cY,11248
-visual_rag_toolkit-0.1.3.dist-info/METADATA,sha256=IQaXJV0GkBuRZG5JTkFA-Zv6pboaAeoWoQIeWzu7-Z4,13180
-visual_rag_toolkit-0.1.3.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
-visual_rag_toolkit-0.1.3.dist-info/entry_points.txt,sha256=6Tob1GPg_ILGELjYTPsAnNMZ1W0NS939nfI7xyW2DIY,102
-visual_rag_toolkit-0.1.3.dist-info/licenses/LICENSE,sha256=hEg_weKnHXJakQRR3sw2ygcZ101zCI00zMhBOPb3yfA,1069
-visual_rag_toolkit-0.1.3.dist-info/RECORD,,
+visual_rag/visualization/saliency.py,sha256=RuVcF1cPrmzdwfigqhP-e1aqzMr817F64aYucbwT79g,11314
+visual_rag_toolkit-0.1.5.dist-info/METADATA,sha256=66Z99JyOGp5QVb2eJR7U2zjF8Qd6nvRJanhezrbZ-3o,13460
+visual_rag_toolkit-0.1.5.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
+visual_rag_toolkit-0.1.5.dist-info/entry_points.txt,sha256=6Tob1GPg_ILGELjYTPsAnNMZ1W0NS939nfI7xyW2DIY,102
+visual_rag_toolkit-0.1.5.dist-info/licenses/LICENSE,sha256=hEg_weKnHXJakQRR3sw2ygcZ101zCI00zMhBOPb3yfA,1069
+visual_rag_toolkit-0.1.5.dist-info/RECORD,,

demo/example_metadata_mapping_sigir.json

@@ -1,37 +0,0 @@
-{
-  "filenames": {
-    "sigir2025-llms": {
-      "year": 2025,
-      "source": "Conference Paper",
-      "district": null,
-      "doc_type": "paper",
-      "project": "sigir-demo",
-      "tags": ["llms", "retrieval"]
-    },
-    "sigir2025-ginger": {
-      "year": 2025,
-      "source": "Conference Paper",
-      "district": null,
-      "doc_type": "paper",
-      "project": "sigir-demo",
-      "tags": ["ginger", "case-study"]
-    },
-    "2505.15859v1": {
-      "year": 2025,
-      "source": "arXiv",
-      "district": null,
-      "doc_type": "preprint",
-      "project": "sigir-demo",
-      "tags": ["arxiv", "ranking"]
-    },
-    "2507.04942v2": {
-      "year": 2025,
-      "source": "arXiv",
-      "district": null,
-      "doc_type": "preprint",
-      "project": "sigir-demo",
-      "tags": ["arxiv", "rag"]
-    }
-  }
-}
-