fix: qdrant deploy

demo/main.py

@@ -1,13 +1,11 @@
-import json
 import os
-from pathlib import Path
 import sys
+from pathlib import Path
 from typing import Any
 
 import gradio as gr
 from loguru import logger
 
-
 # Auto-configure for HF Spaces
 if os.getenv("SPACE_ID"):  # Detect HF Spaces environment
     os.environ.setdefault("QDRANT_URL", ":memory:")
@@ -16,11 +14,6 @@ if os.getenv("SPACE_ID"):  # Detect HF Spaces environment
 sys.path.insert(0, str(Path(__file__).parent.parent / "src"))
 
 from scientific_rag.application.rag.pipeline import RAGPipeline
-from scientific_rag.domain.documents import PaperChunk
-from scientific_rag.domain.queries import Query, QueryFilters
-from scientific_rag.domain.types import DataSource, SectionType
-from scientific_rag.settings import settings
-
 
 MAIN_HEADER = """
 <div style="text-align: center; margin-bottom: 40px;">
@@ -89,7 +82,9 @@ class RAGPipelineWrapper:
             raise ValueError("Please enter your API key.")
 
         if not use_bm25 and not use_dense:
-            raise ValueError("Please enable at least one retrieval method (BM25 or Dense).")
+            raise ValueError(
+                "Please enable at least one retrieval method (BM25 or Dense)."
+            )
 
         if top_k < 1 or top_k > 50:
             raise ValueError("Top-K must be between 1 and 20.")
@@ -103,7 +98,9 @@ class RAGPipelineWrapper:
             )
 
         if provider not in LLM_PROVIDERS:
-            raise ValueError(f"Invalid provider: {provider}. Must be one of {list(LLM_PROVIDERS.keys())}")
+            raise ValueError(
+                f"Invalid provider: {provider}. Must be one of {list(LLM_PROVIDERS.keys())}"
+            )
 
         if model not in LLM_PROVIDERS[provider]["models"]:
             raise ValueError(
@@ -146,7 +143,9 @@ class RAGPipelineWrapper:
 
         return answer, chunks_info
 
-    def _format_answer(self, response, provider: str, model: str, display_chunks: int) -> str:
+    def _format_answer(
+        self, response, provider: str, model: str, display_chunks: int
+    ) -> str:
         """Format RAG response as markdown."""
         lines = []
         lines.append(response.answer)
@@ -163,15 +162,25 @@ class RAGPipelineWrapper:
         metadata_badges.append(
             f'<span class="metadata-badge">📊 Display Chunks: {min(display_chunks, len(response.retrieved_chunks))}</span>'
         )
-        metadata_badges.append(f'<span class="metadata-badge">⏱️ Execution Time: {response.execution_time:.2f}s</span>')
-        metadata_badges.append(f'<span class="metadata-badge">🤖 Model: {provider} / {model}</span>')
+        metadata_badges.append(
+            f'<span class="metadata-badge">⏱️ Execution Time: {response.execution_time:.2f}s</span>'
+        )
+        metadata_badges.append(
+            f'<span class="metadata-badge">🤖 Model: {provider} / {model}</span>'
+        )
 
         if response.used_filters:
-            filters_str = ", ".join([f"{k}={v}" for k, v in response.used_filters.items() if v != "any"])
+            filters_str = ", ".join(
+                [f"{k}={v}" for k, v in response.used_filters.items() if v != "any"]
+            )
             if filters_str:
-                metadata_badges.append(f'<span class="metadata-badge">🔎 Filters: {filters_str}</span>')
+                metadata_badges.append(
+                    f'<span class="metadata-badge">🔎 Filters: {filters_str}</span>'
+                )
 
-        lines.append('<div class="metadata-container">' + " ".join(metadata_badges) + "</div>")
+        lines.append(
+            '<div class="metadata-container">' + " ".join(metadata_badges) + "</div>"
+        )
 
         return "\n".join(lines)
 
@@ -208,8 +217,16 @@ def process_query(
 ) -> tuple[str, str, gr.update, gr.update, gr.update]:
     try:
         if not rag_pipeline:
-            error_msg = "⚠️ **System Error**: RAG Pipeline not initialized. Please check logs."
-            return error_msg, "", gr.update(visible=False), gr.update(visible=True), gr.update(value="", visible=False)
+            error_msg = (
+                "⚠️ **System Error**: RAG Pipeline not initialized. Please check logs."
+            )
+            return (
+                error_msg,
+                "",
+                gr.update(visible=False),
+                gr.update(visible=True),
+                gr.update(value="", visible=False),
+            )
 
         answer, chunks = rag_pipeline.process_query(
             query=query,
@@ -439,9 +456,13 @@ Cross-encoder model to improve result relevance
             )
 
         with gr.Row():
-            loading_status = gr.Markdown(value="", visible=False, elem_classes="loading-indicator")
+            loading_status = gr.Markdown(
+                value="", visible=False, elem_classes="loading-indicator"
+            )
 
-        with gr.Group(visible=True, elem_classes="examples-section") as examples_section:
+        with gr.Group(
+            visible=True, elem_classes="examples-section"
+        ) as examples_section:
             gr.Markdown("## 📝 Example Questions")
 
             gr.HTML("""
@@ -488,7 +509,13 @@ Cross-encoder model to improve result relevance
                 ),
             ),
             inputs=[],
-            outputs=[answer_output, chunks_output, examples_section, answer_section, loading_status],
+            outputs=[
+                answer_output,
+                chunks_output,
+                examples_section,
+                answer_section,
+                loading_status,
+            ],
         ).then(
             fn=process_query,
             inputs=[
@@ -505,7 +532,13 @@ Cross-encoder model to improve result relevance
                 expansion_count,
                 display_chunks,
             ],
-            outputs=[answer_output, chunks_output, examples_section, answer_section, loading_status],
+            outputs=[
+                answer_output,
+                chunks_output,
+                examples_section,
+                answer_section,
+                loading_status,
+            ],
         )
 
         clear_btn.click(
@@ -518,7 +551,14 @@ Cross-encoder model to improve result relevance
                 gr.update(value="", visible=False),
             ),
             inputs=[],
-            outputs=[query, answer_output, chunks_output, examples_section, answer_section, loading_status],
+            outputs=[
+                query,
+                answer_output,
+                chunks_output,
+                examples_section,
+                answer_section,
+                loading_status,
+            ],
         )
 
     return demo

src/scientific_rag/infrastructure/qdrant.py

@@ -1,7 +1,5 @@
-from collections.abc import Sequence
 from typing import Any
 
-from fastembed import SparseTextEmbedding
 from loguru import logger
 from qdrant_client import QdrantClient as SyncQdrantClient
 from qdrant_client.models import (
@@ -10,7 +8,6 @@ from qdrant_client.models import (
     Filter,
     MatchValue,
     Modifier,
-    NamedSparseVector,
     PointStruct,
     SparseIndexParams,
     SparseVector,
@@ -30,14 +27,24 @@ class QdrantService:
         self.collection_name = settings.qdrant_collection_name
 
         logger.info(f"Initializing Qdrant client: {self.url}")
-        self.client = SyncQdrantClient(url=self.url, api_key=self.api_key, timeout=30)
 
-    def create_collection(self, vector_size: int = 384, distance: Distance = Distance.COSINE) -> None:
+        if self.url == ":memory:":
+            self.client = SyncQdrantClient(location=":memory:", timeout=30)
+        else:
+            self.client = SyncQdrantClient(
+                url=self.url, api_key=self.api_key, timeout=30
+            )
+
+    def create_collection(
+        self, vector_size: int = 384, distance: Distance = Distance.COSINE
+    ) -> None:
         if self.client.collection_exists(self.collection_name):
             logger.info(f"Collection '{self.collection_name}' already exists")
             return
 
-        logger.info(f"Creating collection '{self.collection_name}' with dense and sparse vectors")
+        logger.info(
+            f"Creating collection '{self.collection_name}' with dense and sparse vectors"
+        )
         self.client.create_collection(
             collection_name=self.collection_name,
             vectors_config={
@@ -61,7 +68,9 @@ class QdrantService:
             )
         logger.info(f"Collection '{self.collection_name}' created with indexes")
 
-    def upsert_chunks(self, chunks: list[PaperChunk], sparse_embeddings: list[Any] | None = None) -> int:
+    def upsert_chunks(
+        self, chunks: list[PaperChunk], sparse_embeddings: list[Any] | None = None
+    ) -> int:
         if not chunks:
             return 0
 
@@ -74,7 +83,9 @@ class QdrantService:
 
             if sparse_embeddings and i < len(sparse_embeddings):
                 sparse = sparse_embeddings[i]
-                vectors["bm25"] = SparseVector(indices=sparse.indices.tolist(), values=sparse.values.tolist())
+                vectors["bm25"] = SparseVector(
+                    indices=sparse.indices.tolist(), values=sparse.values.tolist()
+                )
 
             points.append(
                 PointStruct(
@@ -163,11 +174,17 @@ class QdrantService:
             return None
 
         must_conditions = []
-        target_list = filter_dict.get("must", []) if "must" in filter_dict else [filter_dict]
+        target_list = (
+            filter_dict.get("must", []) if "must" in filter_dict else [filter_dict]
+        )
 
         for item in target_list:
             if "key" in item and "match" in item:
-                must_conditions.append(FieldCondition(key=item["key"], match=MatchValue(value=item["match"]["value"])))
+                must_conditions.append(
+                    FieldCondition(
+                        key=item["key"], match=MatchValue(value=item["match"]["value"])
+                    )
+                )
 
         return Filter(must=must_conditions) if must_conditions else None