agno 1.8.2__py3-none-any.whl → 2.0.0__py3-none-any.whl

agno/vectordb/chroma/chromadb.py

@@ -1,21 +1,21 @@
 import asyncio
 from hashlib import md5
-from typing import Any, Dict, List, Optional
+from typing import Any, Dict, List, Mapping, Optional, Union, cast
 
 try:
     from chromadb import Client as ChromaDbClient
     from chromadb import PersistentClient as PersistentChromaDbClient
     from chromadb.api.client import ClientAPI
     from chromadb.api.models.Collection import Collection
-    from chromadb.api.types import GetResult, QueryResult
+    from chromadb.api.types import QueryResult
 
 except ImportError:
     raise ImportError("The `chromadb` package is not installed. Please install it via `pip install chromadb`.")
 
-from agno.document import Document
-from agno.embedder import Embedder
-from agno.reranker.base import Reranker
-from agno.utils.log import log_debug, log_info, logger
+from agno.knowledge.document import Document
+from agno.knowledge.embedder import Embedder
+from agno.knowledge.reranker.base import Reranker
+from agno.utils.log import log_debug, log_error, log_info, logger
 from agno.vectordb.base import VectorDb
 from agno.vectordb.distance import Distance
 
@@ -36,7 +36,7 @@ class ChromaDb(VectorDb):
 
         # Embedder for embedding the document contents
         if embedder is None:
-            from agno.embedder.openai import OpenAIEmbedder
+            from agno.knowledge.embedder.openai import OpenAIEmbedder
 
             embedder = OpenAIEmbedder()
             log_info("Embedder not provided, using OpenAIEmbedder as default.")
@@ -91,60 +91,36 @@ class ChromaDb(VectorDb):
         """Create the collection asynchronously by running in a thread."""
         await asyncio.to_thread(self.create)
 
-    def doc_exists(self, document: Document) -> bool:
-        """Check if a document exists in the collection.
+    def name_exists(self, name: str) -> bool:
+        """Check if a document with a given name exists in the collection.
         Args:
-            document (Document): Document to check.
+            name (str): Name of the document to check.
         Returns:
-            bool: True if document exists, False otherwise.
-        """
+            bool: True if document exists, False otherwise."""
         if not self.client:
             logger.warning("Client not initialized")
             return False
 
         try:
             collection: Collection = self.client.get_collection(name=self.collection_name)
-            collection_data: GetResult = collection.get(include=["documents"])  # type: ignore
-            existing_documents = collection_data.get("documents", [])
-            cleaned_content = document.content.replace("\x00", "\ufffd")
-            if cleaned_content in existing_documents:  # type: ignore
-                return True
+            result = collection.get(where=cast(Any, {"name": {"$eq": name}}), limit=1)
+            return len(result.get("ids", [])) > 0
         except Exception as e:
-            logger.error(f"Document does not exist: {e}")
-        return False
-
-    async def async_doc_exists(self, document: Document) -> bool:
-        """Check if a document exists asynchronously."""
-        return await asyncio.to_thread(self.doc_exists, document)
-
-    def name_exists(self, name: str) -> bool:
-        """Check if a document with a given name exists in the collection.
-        Args:
-            name (str): Name of the document to check.
-        Returns:
-            bool: True if document exists, False otherwise."""
-        if self.client:
-            try:
-                collections: Collection = self.client.get_collection(name=self.collection_name)
-                for collection in collections:  # type: ignore
-                    if name in collection:
-                        return True
-            except Exception as e:
-                logger.error(f"Document with given name does not exist: {e}")
+            logger.error(f"Error checking name existence: {e}")
         return False
 
     async def async_name_exists(self, name: str) -> bool:
         """Check if a document with given name exists asynchronously."""
         return await asyncio.to_thread(self.name_exists, name)
 
-    def insert(self, documents: List[Document], filters: Optional[Dict[str, Any]] = None) -> None:
+    def insert(self, content_hash: str, documents: List[Document], filters: Optional[Dict[str, Any]] = None) -> None:
         """Insert documents into the collection.
 
         Args:
             documents (List[Document]): List of documents to insert
             filters (Optional[Dict[str, Any]]): Filters to merge with document metadata
         """
-        log_debug(f"Inserting {len(documents)} documents")
+        log_info(f"Inserting {len(documents)} documents")
         ids: List = []
         docs: List = []
         docs_embeddings: List = []
@@ -163,6 +139,14 @@ class ChromaDb(VectorDb):
             if filters:
                 metadata.update(filters)
 
+            # Add name, content_id to metadata
+            if document.name is not None:
+                metadata["name"] = document.name
+            if document.content_id is not None:
+                metadata["content_id"] = document.content_id
+
+            metadata["content_hash"] = content_hash
+
             docs_embeddings.append(document.embedding)
             docs.append(cleaned_content)
             ids.append(doc_id)
@@ -176,22 +160,82 @@ class ChromaDb(VectorDb):
                 self._collection.add(ids=ids, embeddings=docs_embeddings, documents=docs, metadatas=docs_metadata)
                 log_debug(f"Committed {len(docs)} documents")
 
-    async def async_insert(self, documents: List[Document], filters: Optional[Dict[str, Any]] = None) -> None:
+    async def async_insert(
+        self, content_hash: str, documents: List[Document], filters: Optional[Dict[str, Any]] = None
+    ) -> None:
         """Insert documents asynchronously by running in a thread."""
-        await asyncio.to_thread(self.insert, documents, filters)
+        log_info(f"Async Inserting {len(documents)} documents")
+        ids: List = []
+        docs: List = []
+        docs_embeddings: List = []
+        docs_metadata: List = []
+
+        if not self._collection:
+            self._collection = self.client.get_collection(name=self.collection_name)
+
+        try:
+            embed_tasks = [document.async_embed(embedder=self.embedder) for document in documents]
+            await asyncio.gather(*embed_tasks, return_exceptions=True)
+        except Exception as e:
+            log_error(f"Error processing document: {e}")
+
+        for document in documents:
+            cleaned_content = document.content.replace("\x00", "\ufffd")
+            doc_id = md5(cleaned_content.encode()).hexdigest()
+
+            # Handle metadata and filters
+            metadata = document.meta_data or {}
+            if filters:
+                metadata.update(filters)
+
+            # Add name, content_id to metadata
+            if document.name is not None:
+                metadata["name"] = document.name
+            if document.content_id is not None:
+                metadata["content_id"] = document.content_id
+
+            metadata["content_hash"] = content_hash
+
+            docs_embeddings.append(document.embedding)
+            docs.append(cleaned_content)
+            ids.append(doc_id)
+            docs_metadata.append(metadata)
+            log_debug(f"Prepared document: {document.id} | {document.name} | {metadata}")
+
+        if self._collection is None:
+            logger.warning("Collection does not exist")
+        else:
+            if len(docs) > 0:
+                self._collection.add(ids=ids, embeddings=docs_embeddings, documents=docs, metadatas=docs_metadata)
+                log_debug(f"Committed {len(docs)} documents")
 
     def upsert_available(self) -> bool:
         """Check if upsert is available in ChromaDB."""
         return True
 
-    def upsert(self, documents: List[Document], filters: Optional[Dict[str, Any]] = None) -> None:
+    def upsert(self, content_hash: str, documents: List[Document], filters: Optional[Dict[str, Any]] = None) -> None:
         """Upsert documents into the collection.
 
         Args:
             documents (List[Document]): List of documents to upsert
             filters (Optional[Dict[str, Any]]): Filters to apply while upserting
         """
-        log_debug(f"Upserting {len(documents)} documents")
+        try:
+            if self.content_hash_exists(content_hash):
+                self._delete_by_content_hash(content_hash)
+            self._upsert(content_hash, documents, filters)
+        except Exception as e:
+            logger.error(f"Error upserting documents by content hash: {e}")
+            raise
+
+    def _upsert(self, content_hash: str, documents: List[Document], filters: Optional[Dict[str, Any]] = None) -> None:
+        """Upsert documents into the collection.
+
+        Args:
+            documents (List[Document]): List of documents to upsert
+            filters (Optional[Dict[str, Any]]): Filters to apply while upserting
+        """
+        log_info(f"Upserting {len(documents)} documents")
         ids: List = []
         docs: List = []
         docs_embeddings: List = []
@@ -204,11 +248,25 @@ class ChromaDb(VectorDb):
             document.embed(embedder=self.embedder)
             cleaned_content = document.content.replace("\x00", "\ufffd")
             doc_id = md5(cleaned_content.encode()).hexdigest()
+
+            # Handle metadata and filters
+            metadata = document.meta_data or {}
+            if filters:
+                metadata.update(filters)
+
+            # Add name, content_id to metadata
+            if document.name is not None:
+                metadata["name"] = document.name
+            if document.content_id is not None:
+                metadata["content_id"] = document.content_id
+
+            metadata["content_hash"] = content_hash
+
             docs_embeddings.append(document.embedding)
             docs.append(cleaned_content)
             ids.append(doc_id)
-            docs_metadata.append(document.meta_data)
-            log_debug(f"Upserted document: {document.id} | {document.name} | {document.meta_data}")
+            docs_metadata.append(metadata)
+            log_debug(f"Upserted document: {document.id} | {document.name} | {metadata}")
 
         if self._collection is None:
             logger.warning("Collection does not exist")
@@ -217,9 +275,68 @@ class ChromaDb(VectorDb):
                 self._collection.upsert(ids=ids, embeddings=docs_embeddings, documents=docs, metadatas=docs_metadata)
                 log_debug(f"Committed {len(docs)} documents")
 
-    async def async_upsert(self, documents: List[Document], filters: Optional[Dict[str, Any]] = None) -> None:
+    async def _async_upsert(
+        self, content_hash: str, documents: List[Document], filters: Optional[Dict[str, Any]] = None
+    ) -> None:
+        """Upsert documents into the collection.
+
+        Args:
+            documents (List[Document]): List of documents to upsert
+            filters (Optional[Dict[str, Any]]): Filters to apply while upserting
+        """
+        log_info(f"Async Upserting {len(documents)} documents")
+        ids: List = []
+        docs: List = []
+        docs_embeddings: List = []
+        docs_metadata: List = []
+
+        if not self._collection:
+            self._collection = self.client.get_collection(name=self.collection_name)
+
+        embed_tasks = [document.async_embed(embedder=self.embedder) for document in documents]
+        await asyncio.gather(*embed_tasks, return_exceptions=True)
+
+        for document in documents:
+            cleaned_content = document.content.replace("\x00", "\ufffd")
+            doc_id = md5(cleaned_content.encode()).hexdigest()
+
+            # Handle metadata and filters
+            metadata = document.meta_data or {}
+            if filters:
+                metadata.update(filters)
+
+            # Add name, content_id to metadata
+            if document.name is not None:
+                metadata["name"] = document.name
+            if document.content_id is not None:
+                metadata["content_id"] = document.content_id
+
+            metadata["content_hash"] = content_hash
+
+            docs_embeddings.append(document.embedding)
+            docs.append(cleaned_content)
+            ids.append(doc_id)
+            docs_metadata.append(metadata)
+            log_debug(f"Upserted document: {document.id} | {document.name} | {metadata}")
+
+        if self._collection is None:
+            logger.warning("Collection does not exist")
+        else:
+            if len(docs) > 0:
+                self._collection.upsert(ids=ids, embeddings=docs_embeddings, documents=docs, metadatas=docs_metadata)
+                log_debug(f"Committed {len(docs)} documents")
+
+    async def async_upsert(
+        self, content_hash: str, documents: List[Document], filters: Optional[Dict[str, Any]] = None
+    ) -> None:
         """Upsert documents asynchronously by running in a thread."""
-        await asyncio.to_thread(self.upsert, documents, filters)
+        try:
+            if self.content_hash_exists(content_hash):
+                self._delete_by_content_hash(content_hash)
+            await self._async_upsert(content_hash, documents, filters)
+        except Exception as e:
+            logger.error(f"Error upserting documents by content hash: {e}")
+            raise
 
     def search(self, query: str, limit: int = 5, filters: Optional[Dict[str, Any]] = None) -> List[Document]:
         """Search the collection for a query.
@@ -257,24 +374,62 @@ class ChromaDb(VectorDb):
         # Build search results
         search_results: List[Document] = []
 
-        ids = result.get("ids", [[]])[0]
-        metadata = result.get("metadatas", [{}])[0]
-        documents = result.get("documents", [[]])[0]
-        embeddings = result.get("embeddings")[0]
-        embeddings = [e.tolist() if hasattr(e, "tolist") else e for e in embeddings]
-        distances = result.get("distances", [[]])[0]
+        ids_list = result.get("ids", [[]])
+        metadata_list = result.get("metadatas", [[{}]])
+        documents_list = result.get("documents", [[]])
+        embeddings_list = result.get("embeddings")
+        distances_list = result.get("distances", [[]])
+
+        if not ids_list or not metadata_list or not documents_list or embeddings_list is None or not distances_list:
+            return search_results
+
+        ids = ids_list[0]
+        metadata = [dict(m) if m else {} for m in metadata_list[0]]  # Convert to mutable dicts
+        documents = documents_list[0]
+        embeddings_raw = embeddings_list[0] if embeddings_list else []
+        embeddings = []
+        for e in embeddings_raw:
+            if hasattr(e, "tolist") and callable(getattr(e, "tolist", None)):
+                try:
+                    embeddings.append(list(cast(Any, e).tolist()))
+                except (AttributeError, TypeError):
+                    embeddings.append(list(e) if isinstance(e, (list, tuple)) else [])
+            elif isinstance(e, (list, tuple)):
+                embeddings.append([float(x) for x in e if isinstance(x, (int, float))])
+            elif isinstance(e, (int, float)):
+                embeddings.append([float(e)])
+            else:
+                embeddings.append([])
+        distances = distances_list[0]
 
         for idx, distance in enumerate(distances):
-            metadata[idx]["distances"] = distance
+            if idx < len(metadata):
+                metadata[idx]["distances"] = distance
 
         try:
-            for idx, (id_, metadata, document) in enumerate(zip(ids, metadata, documents)):
+            for idx, (id_, doc_metadata, document) in enumerate(zip(ids, metadata, documents)):
+                # Extract the fields we added to metadata
+                name_val = doc_metadata.pop("name", None)
+                content_id_val = doc_metadata.pop("content_id", None)
+
+                # Convert types to match Document constructor expectations
+                name = str(name_val) if name_val is not None and not isinstance(name_val, str) else name_val
+                content_id = (
+                    str(content_id_val)
+                    if content_id_val is not None and not isinstance(content_id_val, str)
+                    else content_id_val
+                )
+                content = str(document) if document is not None else ""
+                embedding = embeddings[idx] if idx < len(embeddings) else None
+
                 search_results.append(
                     Document(
                         id=id_,
-                        meta_data=metadata,
-                        content=document,
-                        embedding=embeddings[idx],
+                        name=name,
+                        meta_data=doc_metadata,
+                        content=content,
+                        embedding=embedding,
+                        content_id=content_id,
                     )
                 )
         except Exception as e:
@@ -360,3 +515,271 @@ class ChromaDb(VectorDb):
         except Exception as e:
             logger.error(f"Error clearing collection: {e}")
             return False
+
+    def delete_by_id(self, id: str) -> bool:
+        """Delete document by ID."""
+        if not self.client:
+            logger.error("Client not initialized")
+            return False
+
+        try:
+            collection: Collection = self.client.get_collection(name=self.collection_name)
+
+            # Check if document exists
+            if not self.id_exists(id):
+                log_info(f"Document with ID '{id}' not found")
+                return False
+
+            # Delete the document
+            collection.delete(ids=[id])
+            log_info(f"Deleted document with ID '{id}'")
+            return True
+        except Exception as e:
+            logger.error(f"Error deleting document by ID '{id}': {e}")
+            return False
+
+    def delete_by_name(self, name: str) -> bool:
+        """Delete documents by name."""
+        if not self.client:
+            logger.error("Client not initialized")
+            return False
+
+        try:
+            collection: Collection = self.client.get_collection(name=self.collection_name)
+
+            # Find all documents with the given name
+            result = collection.get(where=cast(Any, {"name": {"$eq": name}}))
+            ids_to_delete = result.get("ids", [])
+
+            if not ids_to_delete:
+                log_info(f"No documents found with name '{name}'")
+                return False
+
+            # Delete all matching documents
+            collection.delete(ids=ids_to_delete)
+            log_info(f"Deleted {len(ids_to_delete)} documents with name '{name}'")
+            return True
+        except Exception as e:
+            logger.error(f"Error deleting documents by name '{name}': {e}")
+            return False
+
+    def delete_by_metadata(self, metadata: Dict[str, Any]) -> bool:
+        """Delete documents by metadata."""
+        if not self.client:
+            logger.error("Client not initialized")
+            return False
+
+        try:
+            collection: Collection = self.client.get_collection(name=self.collection_name)
+
+            # Build where clause for metadata filtering
+            where_clause = {}
+            for key, value in metadata.items():
+                where_clause[key] = {"$eq": value}
+
+            # Find all documents with the matching metadata
+            result = collection.get(where=cast(Any, where_clause))
+            ids_to_delete = result.get("ids", [])
+
+            if not ids_to_delete:
+                log_info(f"No documents found with metadata '{metadata}'")
+                return False
+
+            # Delete all matching documents
+            collection.delete(ids=ids_to_delete)
+            log_info(f"Deleted {len(ids_to_delete)} documents with metadata '{metadata}'")
+            return True
+        except Exception as e:
+            logger.error(f"Error deleting documents by metadata '{metadata}': {e}")
+            return False
+
+    def delete_by_content_id(self, content_id: str) -> bool:
+        """Delete documents by content ID."""
+        if not self.client:
+            logger.error("Client not initialized")
+            return False
+
+        try:
+            collection: Collection = self.client.get_collection(name=self.collection_name)
+
+            # Find all documents with the given content_id
+            result = collection.get(where=cast(Any, {"content_id": {"$eq": content_id}}))
+            ids_to_delete = result.get("ids", [])
+
+            if not ids_to_delete:
+                log_info(f"No documents found with content_id '{content_id}'")
+                return False
+
+            # Delete all matching documents
+            collection.delete(ids=ids_to_delete)
+            log_info(f"Deleted {len(ids_to_delete)} documents with content_id '{content_id}'")
+            return True
+        except Exception as e:
+            logger.error(f"Error deleting documents by content_id '{content_id}': {e}")
+            return False
+
+    def _delete_by_content_hash(self, content_hash: str) -> bool:
+        """Delete documents by content hash."""
+        if not self.client:
+            logger.error("Client not initialized")
+            return False
+
+        try:
+            collection: Collection = self.client.get_collection(name=self.collection_name)
+
+            # Find all documents with the given content_hash
+            result = collection.get(where=cast(Any, {"content_hash": {"$eq": content_hash}}))
+            ids_to_delete = result.get("ids", [])
+
+            if not ids_to_delete:
+                log_info(f"No documents found with content_hash '{content_hash}'")
+                return False
+
+            # Delete all matching documents
+            collection.delete(ids=ids_to_delete)
+            log_info(f"Deleted {len(ids_to_delete)} documents with content_hash '{content_hash}'")
+            return True
+        except Exception as e:
+            logger.error(f"Error deleting documents by content_hash '{content_hash}': {e}")
+            return False
+
+    def id_exists(self, id: str) -> bool:
+        """Check if a document with the given ID exists in the collection.
+
+        Args:
+            id (str): The document ID to check.
+
+        Returns:
+            bool: True if the document exists, False otherwise.
+        """
+        if not self.client:
+            logger.error("Client not initialized")
+            return False
+
+        try:
+            collection: Collection = self.client.get_collection(name=self.collection_name)
+            print("COLLECTION_----------", collection)
+            # Try to get the document by ID
+            result = collection.get(ids=[id])
+            found_ids = result.get("ids", [])
+
+            # Return True if the document was found
+            return len(found_ids) > 0
+        except Exception as e:
+            logger.error(f"Error checking if ID '{id}' exists: {e}")
+            return False
+
+    def content_hash_exists(self, content_hash: str) -> bool:
+        """Check if documents with the given content hash exist."""
+        if not self.client:
+            logger.error("Client not initialized")
+            return False
+
+        try:
+            collection: Collection = self.client.get_collection(name=self.collection_name)
+
+            # Try to query for documents with the given content_hash
+            try:
+                result = collection.get(where=cast(Any, {"content_hash": {"$eq": content_hash}}))
+                # Safely extract ids from result
+                if hasattr(result, "get") and callable(result.get):
+                    found_ids = result.get("ids", [])
+                elif hasattr(result, "__getitem__") and "ids" in result:
+                    found_ids = result["ids"]
+                else:
+                    found_ids = []
+
+                # Return True if any documents were found
+                if isinstance(found_ids, (list, tuple)):
+                    return len(found_ids) > 0
+                elif isinstance(found_ids, int):
+                    # Some ChromaDB versions might return a count instead of a list
+                    return found_ids > 0
+                else:
+                    return False
+
+            except TypeError as te:
+                if "object of type 'int' has no len()" in str(te):
+                    # Known issue with ChromaDB 0.5.0 - internal bug
+                    # As a workaround, assume content doesn't exist to allow processing to continue
+                    logger.warning(
+                        f"ChromaDB internal error (version 0.5.0 bug): {te}. Assuming content_hash '{content_hash}' does not exist."
+                    )
+                    return False
+                else:
+                    raise te
+
+        except Exception as e:
+            logger.error(f"Error checking if content_hash '{content_hash}' exists: {e}")
+            return False
+
+    def update_metadata(self, content_id: str, metadata: Dict[str, Any]) -> None:
+        """
+        Update the metadata for documents with the given content_id.
+
+        Args:
+            content_id (str): The content ID to update
+            metadata (Dict[str, Any]): The metadata to update
+        """
+        try:
+            if not self.client:
+                logger.error("Client not initialized")
+                return
+
+            collection: Collection = self.client.get_collection(name=self.collection_name)
+
+            # Find documents with the given content_id
+            try:
+                result = collection.get(where=cast(Any, {"content_id": {"$eq": content_id}}))
+
+                # Extract IDs and current metadata
+                if hasattr(result, "get") and callable(result.get):
+                    ids = result.get("ids", [])
+                    current_metadatas = result.get("metadatas", [])
+                elif hasattr(result, "__getitem__"):
+                    ids = result.get("ids", []) if "ids" in result else []
+                    current_metadatas = result.get("metadatas", []) if "metadatas" in result else []
+                else:
+                    ids = []
+                    current_metadatas = []
+
+                if not ids:
+                    logger.debug(f"No documents found with content_id: {content_id}")
+                    return
+
+                # Merge metadata for each document
+                updated_metadatas = []
+                for i, current_meta in enumerate(current_metadatas or []):
+                    if current_meta is None:
+                        meta_dict: Dict[str, Any] = {}
+                    else:
+                        meta_dict = dict(current_meta)  # Convert Mapping to dict
+                    updated_meta: Dict[str, Any] = meta_dict.copy()
+                    updated_meta.update(metadata)
+
+                    if "filters" not in updated_meta:
+                        updated_meta["filters"] = {}
+                    if isinstance(updated_meta["filters"], dict):
+                        updated_meta["filters"].update(metadata)
+                    else:
+                        updated_meta["filters"] = metadata
+                    updated_metadatas.append(updated_meta)
+
+                # Update the documents
+                # Convert to the expected type for ChromaDB
+                chroma_metadatas = cast(List[Mapping[str, Union[str, int, float, bool, None]]], updated_metadatas)
+                collection.update(ids=ids, metadatas=chroma_metadatas)
+                logger.debug(f"Updated metadata for {len(ids)} documents with content_id: {content_id}")
+
+            except TypeError as te:
+                if "object of type 'int' has no len()" in str(te):
+                    logger.warning(
+                        f"ChromaDB internal error (version 0.5.0 bug): {te}. Cannot update metadata for content_id '{content_id}'."
+                    )
+                    return
+                else:
+                    raise te
+
+        except Exception as e:
+            logger.error(f"Error updating metadata for content_id '{content_id}': {e}")
+            raise