vector-inspector 0.2.7__tar.gz → 0.3.1__tar.gz

{vector_inspector-0.2.7 → vector_inspector-0.3.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: vector-inspector
-Version: 0.2.7
+Version: 0.3.1
 Summary: A comprehensive desktop application for visualizing, querying, and managing vector database data
 Author-Email: Anthony Dawson <anthonypdawson+github@gmail.com>
 License: MIT

{vector_inspector-0.2.7 → vector_inspector-0.3.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "vector-inspector"
-version = "0.2.7"
+version = "0.3.1"
 description = "A comprehensive desktop application for visualizing, querying, and managing vector database data"
 authors = [
     { name = "Anthony Dawson", email = "anthonypdawson+github@gmail.com" },

vector_inspector-0.3.1/src/vector_inspector/config/__init__.py

@@ -0,0 +1,4 @@
+"""Data package for Vector Inspector.
+
+Contains static data files like the model registry.
+"""

vector_inspector-0.3.1/src/vector_inspector/config/known_embedding_models.json

@@ -0,0 +1,432 @@
+{
+    "models": [
+        {
+            "name": "all-MiniLM-L6-v2",
+            "type": "sentence-transformer",
+            "dimension": 384,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Fast, small-footprint text embeddings (good default for text search)"
+        },
+        {
+            "name": "openai/clip-vit-base-patch32",
+            "type": "clip",
+            "dimension": 512,
+            "modality": "multimodal",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Standard CLIP ViT-B/32 model; supports matching text ↔ images"
+        },
+        {
+            "name": "paraphrase-albert-small-v2",
+            "type": "sentence-transformer",
+            "dimension": 768,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Smaller paraphrase-specialized model"
+        },
+        {
+            "name": "all-mpnet-base-v2",
+            "type": "sentence-transformer",
+            "dimension": 768,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "High-quality text embeddings; recommended for semantic tasks"
+        },
+        {
+            "name": "all-roberta-large-v1",
+            "type": "sentence-transformer",
+            "dimension": 1024,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Large model — high quality, larger memory and compute"
+        },
+        {
+            "name": "gtr-t5-large",
+            "type": "sentence-transformer",
+            "dimension": 1536,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Very large embeddings useful for specialized high-recall tasks"
+        },
+        {
+            "name": "sentence-transformers/multi-qa-MiniLM-L6-cos-v1",
+            "type": "sentence-transformer",
+            "dimension": 384,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Optimized for semantic search and question-answering tasks"
+        },
+        {
+            "name": "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2",
+            "type": "sentence-transformer",
+            "dimension": 384,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Multilingual support for 50+ languages"
+        },
+        {
+            "name": "sentence-transformers/msmarco-distilbert-base-v4",
+            "type": "sentence-transformer",
+            "dimension": 768,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Trained on MS MARCO dataset, good for passage retrieval"
+        },
+        {
+            "name": "sentence-transformers/all-distilroberta-v1",
+            "type": "sentence-transformer",
+            "dimension": 768,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Distilled RoBERTa model, balance of speed and quality"
+        },
+        {
+            "name": "sentence-transformers/paraphrase-mpnet-base-v2",
+            "type": "sentence-transformer",
+            "dimension": 768,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "High-quality paraphrase detection and semantic similarity"
+        },
+        {
+            "name": "BAAI/bge-small-en-v1.5",
+            "type": "sentence-transformer",
+            "dimension": 384,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Beijing Academy of AI model, strong performance for size"
+        },
+        {
+            "name": "BAAI/bge-base-en-v1.5",
+            "type": "sentence-transformer",
+            "dimension": 768,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "High-quality English embeddings, MTEB benchmark leader"
+        },
+        {
+            "name": "BAAI/bge-large-en-v1.5",
+            "type": "sentence-transformer",
+            "dimension": 1024,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Large model with excellent retrieval performance"
+        },
+        {
+            "name": "thenlper/gte-small",
+            "type": "sentence-transformer",
+            "dimension": 384,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "General Text Embeddings (GTE) small variant"
+        },
+        {
+            "name": "thenlper/gte-base",
+            "type": "sentence-transformer",
+            "dimension": 768,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "General Text Embeddings (GTE) base model"
+        },
+        {
+            "name": "thenlper/gte-large",
+            "type": "sentence-transformer",
+            "dimension": 1024,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "General Text Embeddings (GTE) large variant"
+        },
+        {
+            "name": "intfloat/e5-small-v2",
+            "type": "sentence-transformer",
+            "dimension": 384,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "E5 family small model, prefix with 'query: ' or 'passage: '"
+        },
+        {
+            "name": "intfloat/e5-base-v2",
+            "type": "sentence-transformer",
+            "dimension": 768,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "E5 family base model, strong asymmetric retrieval"
+        },
+        {
+            "name": "intfloat/e5-large-v2",
+            "type": "sentence-transformer",
+            "dimension": 1024,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "E5 family large model, top MTEB performance"
+        },
+        {
+            "name": "intfloat/multilingual-e5-small",
+            "type": "sentence-transformer",
+            "dimension": 384,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Multilingual E5 model supporting 100+ languages"
+        },
+        {
+            "name": "intfloat/multilingual-e5-base",
+            "type": "sentence-transformer",
+            "dimension": 768,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Multilingual E5 base model, excellent cross-lingual retrieval"
+        },
+        {
+            "name": "intfloat/multilingual-e5-large",
+            "type": "sentence-transformer",
+            "dimension": 1024,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Multilingual E5 large model, best-in-class multilingual embeddings"
+        },
+        {
+            "name": "openai/clip-vit-large-patch14",
+            "type": "clip",
+            "dimension": 768,
+            "modality": "multimodal",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Larger CLIP ViT-L/14 model, better quality than base"
+        },
+        {
+            "name": "openai/clip-vit-large-patch14-336",
+            "type": "clip",
+            "dimension": 768,
+            "modality": "multimodal",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Higher resolution (336x336) variant of ViT-L/14"
+        },
+        {
+            "name": "laion/CLIP-ViT-B-32-laion2B-s34B-b79K",
+            "type": "clip",
+            "dimension": 512,
+            "modality": "multimodal",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "LAION's CLIP trained on 2B image-text pairs"
+        },
+        {
+            "name": "laion/CLIP-ViT-H-14-laion2B-s32B-b79K",
+            "type": "clip",
+            "dimension": 1024,
+            "modality": "multimodal",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "LAION's huge CLIP model, excellent quality"
+        },
+        {
+            "name": "text-embedding-ada-002",
+            "type": "openai",
+            "dimension": 1536,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "openai-api",
+            "description": "OpenAI's production embedding model (legacy). Requires API key."
+        },
+        {
+            "name": "text-embedding-3-small",
+            "type": "openai",
+            "dimension": 1536,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "openai-api",
+            "description": "OpenAI's newer small model, better than ada-002. Requires API key."
+        },
+        {
+            "name": "text-embedding-3-large",
+            "type": "openai",
+            "dimension": 3072,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "openai-api",
+            "description": "OpenAI's large embedding model, highest quality. Requires API key."
+        },
+        {
+            "name": "embed-english-v3.0",
+            "type": "cohere",
+            "dimension": 1024,
+            "modality": "text",
+            "normalization": "none",
+            "source": "cohere-api",
+            "description": "Cohere's English embedding model. Requires API key."
+        },
+        {
+            "name": "embed-english-light-v3.0",
+            "type": "cohere",
+            "dimension": 384,
+            "modality": "text",
+            "normalization": "none",
+            "source": "cohere-api",
+            "description": "Cohere's lightweight English model. Requires API key."
+        },
+        {
+            "name": "embed-multilingual-v3.0",
+            "type": "cohere",
+            "dimension": 1024,
+            "modality": "text",
+            "normalization": "none",
+            "source": "cohere-api",
+            "description": "Cohere's multilingual model supporting 100+ languages. Requires API key."
+        },
+        {
+            "name": "embed-multilingual-light-v3.0",
+            "type": "cohere",
+            "dimension": 384,
+            "modality": "text",
+            "normalization": "none",
+            "source": "cohere-api",
+            "description": "Cohere's lightweight multilingual model. Requires API key."
+        },
+        {
+            "name": "textembedding-gecko@003",
+            "type": "vertex-ai",
+            "dimension": 768,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "google-cloud",
+            "description": "Google's Gecko model for text embeddings. Requires Google Cloud credentials."
+        },
+        {
+            "name": "text-embedding-004",
+            "type": "vertex-ai",
+            "dimension": 768,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "google-cloud",
+            "description": "Google's latest text embedding model. Requires Google Cloud credentials."
+        },
+        {
+            "name": "text-multilingual-embedding-002",
+            "type": "vertex-ai",
+            "dimension": 768,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "google-cloud",
+            "description": "Google's multilingual embedding model. Requires Google Cloud credentials."
+        },
+        {
+            "name": "multimodalembedding@001",
+            "type": "vertex-ai",
+            "dimension": 1408,
+            "modality": "multimodal",
+            "normalization": "l2",
+            "source": "google-cloud",
+            "description": "Google's multimodal embedding model. Requires Google Cloud credentials."
+        },
+        {
+            "name": "voyage-large-2",
+            "type": "voyage",
+            "dimension": 1536,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "voyage-api",
+            "description": "Voyage AI's large model. Requires API key."
+        },
+        {
+            "name": "voyage-code-2",
+            "type": "voyage",
+            "dimension": 1536,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "voyage-api",
+            "description": "Voyage AI's code-optimized model. Requires API key."
+        },
+        {
+            "name": "voyage-2",
+            "type": "voyage",
+            "dimension": 1024,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "voyage-api",
+            "description": "Voyage AI's general-purpose model. Requires API key."
+        },
+        {
+            "name": "jinaai/jina-embeddings-v2-base-en",
+            "type": "sentence-transformer",
+            "dimension": 768,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Jina AI's 8k context length model, good for long documents"
+        },
+        {
+            "name": "jinaai/jina-embeddings-v2-small-en",
+            "type": "sentence-transformer",
+            "dimension": 512,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Jina AI's small model with 8k context length"
+        },
+        {
+            "name": "nomic-ai/nomic-embed-text-v1",
+            "type": "sentence-transformer",
+            "dimension": 768,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Nomic's open-source text embedding model with 8k context"
+        },
+        {
+            "name": "nomic-ai/nomic-embed-text-v1.5",
+            "type": "sentence-transformer",
+            "dimension": 768,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Nomic's improved model with better performance"
+        },
+        {
+            "name": "Alibaba-NLP/gte-Qwen2-7B-instruct",
+            "type": "sentence-transformer",
+            "dimension": 3584,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "Very large instruction-following embedding model, SOTA on many benchmarks"
+        },
+        {
+            "name": "nvidia/NV-Embed-v1",
+            "type": "sentence-transformer",
+            "dimension": 4096,
+            "modality": "text",
+            "normalization": "l2",
+            "source": "huggingface",
+            "description": "NVIDIA's embedding model, excellent for retrieval tasks"
+        }
+    ],
+    "metadata": {
+        "version": "1.0.0",
+        "last_updated": "2026-01-24",
+        "description": "Known embedding models registry for Vector Inspector"
+    }
+}

{vector_inspector-0.2.7 → vector_inspector-0.3.1}/src/vector_inspector/core/connections/__init__.py

@@ -3,5 +3,6 @@
 from .base_connection import VectorDBConnection
 from .chroma_connection import ChromaDBConnection
 from .qdrant_connection import QdrantConnection
+from .pinecone_connection import PineconeConnection
 
-__all__ = ["VectorDBConnection", "ChromaDBConnection", "QdrantConnection"]
+__all__ = ["VectorDBConnection", "ChromaDBConnection", "QdrantConnection", "PineconeConnection"]

{vector_inspector-0.2.7 → vector_inspector-0.3.1}/src/vector_inspector/core/connections/base_connection.py

@@ -229,5 +229,46 @@ class VectorDBConnection(ABC):
             {"name": "in", "server_side": True},
             {"name": "not in", "server_side": True},
             {"name": "contains", "server_side": False},
-            {"name": "not contains", "server_side": False},
         ]
+    
+    def get_embedding_model(self, collection_name: str, connection_id: Optional[str] = None) -> Optional[str]:
+        """
+        Get the embedding model used for a collection.
+        
+        Retrieves the model name from:
+        1. Collection-level metadata (if supported)
+        2. Vector metadata (_embedding_model field)
+        3. User settings (for collections we can't modify)
+        
+        Args:
+            collection_name: Name of collection
+            connection_id: Optional connection ID for settings lookup
+            
+        Returns:
+            Model name string (e.g., "sentence-transformers/all-MiniLM-L6-v2") or None
+        """
+        try:
+            # First try to get from collection-level metadata
+            info = self.get_collection_info(collection_name)
+            if info and info.get("embedding_model"):
+                return info["embedding_model"]
+            
+            # Fall back to checking a sample vector's metadata
+            data = self.get_all_items(collection_name, limit=1, offset=0)
+            if data and data.get("metadatas") and len(data["metadatas"]) > 0:
+                metadata = data["metadatas"][0]
+                if "_embedding_model" in metadata:
+                    return metadata["_embedding_model"]
+            
+            # Finally, check user settings (for collections we can't modify)
+            if connection_id:
+                from ...services.settings_service import SettingsService
+                settings = SettingsService()
+                model_info = settings.get_embedding_model(connection_id, collection_name)
+                if model_info:
+                    return model_info["model"]
+            
+            return None
+        except Exception as e:
+            print(f"Failed to get embedding model: {e}")
+            return None

{vector_inspector-0.2.7 → vector_inspector-0.3.1}/src/vector_inspector/core/connections/chroma_connection.py

@@ -212,27 +212,37 @@ class ChromaDBConnection(VectorDBConnection):
             # ChromaDB uses cosine distance by default (or can be configured)
             # Try to get metadata from collection if available
             distance_metric = "Cosine (default)"
+            embedding_model = None
             try:
                 # ChromaDB collections may have metadata about distance function
                 col_metadata = collection.metadata
-                if col_metadata and "hnsw:space" in col_metadata:
-                    space = col_metadata["hnsw:space"]
-                    if space == "l2":
-                        distance_metric = "Euclidean (L2)"
-                    elif space == "ip":
-                        distance_metric = "Inner Product"
-                    elif space == "cosine":
-                        distance_metric = "Cosine"
+                if col_metadata:
+                    if "hnsw:space" in col_metadata:
+                        space = col_metadata["hnsw:space"]
+                        if space == "l2":
+                            distance_metric = "Euclidean (L2)"
+                        elif space == "ip":
+                            distance_metric = "Inner Product"
+                        elif space == "cosine":
+                            distance_metric = "Cosine"
+                    # Get embedding model if stored
+                    if "embedding_model" in col_metadata:
+                        embedding_model = col_metadata["embedding_model"]
             except:
                 pass  # Use default if unable to determine
             
-            return {
+            result = {
                 "name": name,
                 "count": count,
                 "metadata_fields": metadata_fields,
                 "vector_dimension": vector_dimension,
                 "distance_metric": distance_metric,
             }
+            
+            if embedding_model:
+                result["embedding_model"] = embedding_model
+            
+            return result
         except Exception as e:
             print(f"Failed to get collection info: {e}")
             return None
@@ -453,8 +463,34 @@ class ChromaDBConnection(VectorDBConnection):
 
     # Implement base connection uniform APIs
     def create_collection(self, name: str, vector_size: int, distance: str = "Cosine") -> bool:
-        """Create a collection. Chroma doesn't require vector size at creation."""
-        return self.get_collection(name) is not None
+        """Create a collection. If it doesn't exist, attempt to create it using Chroma client APIs."""
+        if not self._client:
+            return False
+
+        try:
+            # Prefer get_or_create_collection if available
+            if hasattr(self._client, "get_or_create_collection"):
+                col = self._client.get_or_create_collection(name=name)
+                self._current_collection = col
+                return True
+
+            # Fallback to create_collection/create and then fetch
+            if hasattr(self._client, "create_collection"):
+                try:
+                    self._client.create_collection(name=name)
+                except Exception:
+                    # Some clients may raise if already exists; ignore
+                    pass
+                col = self._client.get_collection(name=name)
+                self._current_collection = col
+                return col is not None
+
+            # As a last resort, check if collection exists
+            col = self.get_collection(name)
+            return col is not None
+        except Exception as e:
+            print(f"Failed to create collection: {e}")
+            return False
 
     def get_items(self, name: str, ids: List[str]) -> Dict[str, Any]:
         """Retrieve items by IDs."""