langroid 0.1.85__py3-none-any.whl → 0.1.219__py3-none-any.whl

langroid/vector_store/lancedb.py

@@ -0,0 +1,463 @@
+import logging
+from typing import Any, Dict, Generator, List, Optional, Sequence, Tuple, Type
+
+import lancedb
+import pandas as pd
+from dotenv import load_dotenv
+from lancedb.pydantic import LanceModel, Vector
+from lancedb.query import LanceVectorQueryBuilder
+from pydantic import BaseModel, ValidationError, create_model
+
+from langroid.embedding_models.base import (
+    EmbeddingModel,
+    EmbeddingModelsConfig,
+)
+from langroid.embedding_models.models import OpenAIEmbeddingsConfig
+from langroid.mytypes import Document, EmbeddingFunction
+from langroid.utils.configuration import settings
+from langroid.utils.pydantic_utils import (
+    dataframe_to_document_model,
+    dataframe_to_documents,
+    extend_document_class,
+    extra_metadata,
+    flatten_pydantic_instance,
+    flatten_pydantic_model,
+    nested_dict_from_flat,
+)
+from langroid.vector_store.base import VectorStore, VectorStoreConfig
+
+logger = logging.getLogger(__name__)
+
+
+class LanceDBConfig(VectorStoreConfig):
+    cloud: bool = False
+    collection_name: str | None = "temp"
+    storage_path: str = ".lancedb/data"
+    embedding: EmbeddingModelsConfig = OpenAIEmbeddingsConfig()
+    distance: str = "cosine"
+    # document_class is used to store in lancedb with right schema,
+    # and also to retrieve the right type of Documents when searching.
+    document_class: Type[Document] = Document
+    flatten: bool = False  # flatten Document class into LanceSchema ?
+
+
+class LanceDB(VectorStore):
+    def __init__(self, config: LanceDBConfig = LanceDBConfig()):
+        super().__init__(config)
+        self.config: LanceDBConfig = config
+        emb_model = EmbeddingModel.create(config.embedding)
+        self.embedding_fn: EmbeddingFunction = emb_model.embedding_fn()
+        self.embedding_dim = emb_model.embedding_dims
+        self.host = config.host
+        self.port = config.port
+        self.is_from_dataframe = False  # were docs ingested from a dataframe?
+        self.df_metadata_columns: List[str] = []  # metadata columns from dataframe
+        self._setup_schemas(config.document_class)
+
+        load_dotenv()
+        if self.config.cloud:
+            logger.warning(
+                "LanceDB Cloud is not available yet. Switching to local storage."
+            )
+            config.cloud = False
+        else:
+            try:
+                self.client = lancedb.connect(
+                    uri=config.storage_path,
+                )
+            except Exception as e:
+                new_storage_path = config.storage_path + ".new"
+                logger.warning(
+                    f"""
+                    Error connecting to local LanceDB at {config.storage_path}:
+                    {e}
+                    Switching to {new_storage_path}
+                    """
+                )
+                self.client = lancedb.connect(
+                    uri=new_storage_path,
+                )
+
+        # Note: Only create collection if a non-null collection name is provided.
+        # This is useful to delay creation of vecdb until we have a suitable
+        # collection name (e.g. we could get it from the url or folder path).
+        if config.collection_name is not None:
+            self.create_collection(
+                config.collection_name, replace=config.replace_collection
+            )
+
+    def _setup_schemas(self, doc_cls: Type[Document] | None) -> None:
+        doc_cls = doc_cls or self.config.document_class
+        self.unflattened_schema = self._create_lance_schema(doc_cls)
+        self.schema = (
+            self._create_flat_lance_schema(doc_cls)
+            if self.config.flatten
+            else self.unflattened_schema
+        )
+
+    def clear_empty_collections(self) -> int:
+        coll_names = self.list_collections()
+        n_deletes = 0
+        for name in coll_names:
+            nr = self.client.open_table(name).head(1).shape[0]
+            if nr == 0:
+                n_deletes += 1
+                self.client.drop_table(name)
+        return n_deletes
+
+    def clear_all_collections(self, really: bool = False, prefix: str = "") -> int:
+        """Clear all collections with the given prefix."""
+        if not really:
+            logger.warning("Not deleting all collections, set really=True to confirm")
+            return 0
+        coll_names = [
+            c for c in self.list_collections(empty=True) if c.startswith(prefix)
+        ]
+        if len(coll_names) == 0:
+            logger.warning(f"No collections found with prefix {prefix}")
+            return 0
+        n_empty_deletes = 0
+        n_non_empty_deletes = 0
+        for name in coll_names:
+            nr = self.client.open_table(name).head(1).shape[0]
+            n_empty_deletes += nr == 0
+            n_non_empty_deletes += nr > 0
+            self.client.drop_table(name)
+        logger.warning(
+            f"""
+            Deleted {n_empty_deletes} empty collections and 
+            {n_non_empty_deletes} non-empty collections.
+            """
+        )
+        return n_empty_deletes + n_non_empty_deletes
+
+    def list_collections(self, empty: bool = False) -> List[str]:
+        """
+        Returns:
+            List of collection names that have at least one vector.
+
+        Args:
+            empty (bool, optional): Whether to include empty collections.
+        """
+        colls = self.client.table_names(limit=None)
+        if len(colls) == 0:
+            return []
+        if empty:  # include empty tbls
+            return colls  # type: ignore
+        counts = [self.client.open_table(coll).head(1).shape[0] for coll in colls]
+        return [coll for coll, count in zip(colls, counts) if count > 0]
+
+    def _create_lance_schema(self, doc_cls: Type[Document]) -> Type[BaseModel]:
+        """
+        Create a subclass of LanceModel with fields:
+         - id (str)
+         - Vector field that has dims equal to
+            the embedding dimension of the embedding model, and a data field of type
+            DocClass.
+         - other fields from doc_cls
+
+        Args:
+            doc_cls (Type[Document]): A Pydantic model which should be a subclass of
+                Document, to be used as the type for the data field.
+
+        Returns:
+            Type[BaseModel]: A new Pydantic model subclassing from LanceModel.
+
+        Raises:
+            ValueError: If `n` is not a non-negative integer or if `DocClass` is not a
+                subclass of Document.
+        """
+        if not issubclass(doc_cls, Document):
+            raise ValueError("DocClass must be a subclass of Document")
+
+        n = self.embedding_dim
+
+        # Prepare fields for the new model
+        fields = {"id": (str, ...), "vector": (Vector(n), ...)}
+
+        sorted_fields = dict(
+            sorted(doc_cls.__fields__.items(), key=lambda item: item[0])
+        )
+        # Add both statically and dynamically defined fields from doc_cls
+        for field_name, field in sorted_fields.items():
+            fields[field_name] = (field.outer_type_, field.default)
+
+        # Create the new model with dynamic fields
+        NewModel = create_model(
+            "NewModel", __base__=LanceModel, **fields
+        )  # type: ignore
+        return NewModel  # type: ignore
+
+    def _create_flat_lance_schema(self, doc_cls: Type[Document]) -> Type[BaseModel]:
+        """
+        Flat version of the lance_schema, as nested Pydantic schemas are not yet
+        supported by LanceDB.
+        """
+        lance_model = self._create_lance_schema(doc_cls)
+        FlatModel = flatten_pydantic_model(lance_model, base_model=LanceModel)
+        return FlatModel
+
+    def create_collection(self, collection_name: str, replace: bool = False) -> None:
+        """
+        Create a collection with the given name, optionally replacing an existing
+            collection if `replace` is True.
+        Args:
+            collection_name (str): Name of the collection to create.
+            replace (bool): Whether to replace an existing collection
+                with the same name. Defaults to False.
+        """
+        self.config.collection_name = collection_name
+        collections = self.list_collections()
+        if collection_name in collections:
+            coll = self.client.open_table(collection_name)
+            if coll.head().shape[0] > 0:
+                logger.warning(f"Non-empty Collection {collection_name} already exists")
+                if not replace:
+                    logger.warning("Not replacing collection")
+                    return
+                else:
+                    logger.warning("Recreating fresh collection")
+        self.client.create_table(collection_name, schema=self.schema, mode="overwrite")
+        if settings.debug:
+            level = logger.getEffectiveLevel()
+            logger.setLevel(logging.INFO)
+            logger.setLevel(level)
+
+    def _maybe_set_doc_class_schema(self, doc: Document) -> None:
+        """
+        Set the config.document_class and self.schema based on doc if needed
+        Args:
+            doc: an instance of Document, to be added to a collection
+        """
+        extra_metadata_fields = extra_metadata(doc, self.config.document_class)
+        if len(extra_metadata_fields) > 0:
+            logger.warning(
+                f"""
+                    Added documents contain extra metadata fields:
+                    {extra_metadata_fields}
+                    which were not present in the original config.document_class.
+                    Trying to change document_class and corresponding schemas.
+                    Overriding LanceDBConfig.document_class with an auto-generated 
+                    Pydantic class that includes these extra fields.
+                    If this fails, or you see odd results, it is recommended that you 
+                    define a subclass of Document, with metadata of class derived from 
+                    DocMetaData, with extra fields defined via 
+                    `Field(..., description="...")` declarations,
+                    and set this document class as the value of the 
+                    LanceDBConfig.document_class attribute.
+                    """
+            )
+
+            doc_cls = extend_document_class(doc)
+            self.config.document_class = doc_cls
+            self._setup_schemas(doc_cls)
+
+    def add_documents(self, documents: Sequence[Document]) -> None:
+        super().maybe_add_ids(documents)
+        colls = self.list_collections(empty=True)
+        if len(documents) == 0:
+            return
+        embedding_vecs = self.embedding_fn([doc.content for doc in documents])
+        coll_name = self.config.collection_name
+        if coll_name is None:
+            raise ValueError("No collection name set, cannot ingest docs")
+        self._maybe_set_doc_class_schema(documents[0])
+        if (
+            coll_name not in colls
+            or self.client.open_table(coll_name).head(1).shape[0] == 0
+        ):
+            # collection either doesn't exist or is empty, so replace it,
+            self.create_collection(coll_name, replace=True)
+
+        ids = [str(d.id()) for d in documents]
+        # don't insert all at once, batch in chunks of b,
+        # else we get an API error
+        b = self.config.batch_size
+
+        def make_batches() -> Generator[List[BaseModel], None, None]:
+            for i in range(0, len(ids), b):
+                batch = [
+                    self.unflattened_schema(
+                        id=ids[i + j],
+                        vector=embedding_vecs[i + j],
+                        **doc.dict(),
+                    )
+                    for j, doc in enumerate(documents[i : i + b])
+                ]
+                if self.config.flatten:
+                    batch = [
+                        flatten_pydantic_instance(instance)  # type: ignore
+                        for instance in batch
+                    ]
+                yield batch
+
+        tbl = self.client.open_table(self.config.collection_name)
+        try:
+            tbl.add(make_batches())
+        except Exception as e:
+            logger.error(
+                f"""
+                Error adding documents to LanceDB: {e}
+                POSSIBLE REMEDY: Delete the LancdDB storage directory
+                {self.config.storage_path} and try again.
+                """
+            )
+
+    def add_dataframe(
+        self,
+        df: pd.DataFrame,
+        content: str = "content",
+        metadata: List[str] = [],
+    ) -> None:
+        """
+        Add a dataframe to the collection.
+        Args:
+            df (pd.DataFrame): A dataframe
+            content (str): The name of the column in the dataframe that contains the
+                text content to be embedded using the embedding model.
+            metadata (List[str]): A list of column names in the dataframe that contain
+                metadata to be stored in the database. Defaults to [].
+        """
+        self.is_from_dataframe = True
+        actual_metadata = metadata.copy()
+        self.df_metadata_columns = actual_metadata  # could be updated below
+        # get content column
+        content_values = df[content].values.tolist()
+        embedding_vecs = self.embedding_fn(content_values)
+
+        # add vector column
+        df["vector"] = embedding_vecs
+        if content != "content":
+            # rename content column to "content", leave existing column intact
+            df = df.rename(columns={content: "content"}, inplace=False)
+
+        if "id" not in df.columns:
+            docs = dataframe_to_documents(df, content="content", metadata=metadata)
+            ids = [str(d.id()) for d in docs]
+            df["id"] = ids
+
+        if "id" not in actual_metadata:
+            actual_metadata += ["id"]
+
+        colls = self.list_collections(empty=True)
+        coll_name = self.config.collection_name
+        if (
+            coll_name not in colls
+            or self.client.open_table(coll_name).head(1).shape[0] == 0
+        ):
+            # collection either doesn't exist or is empty, so replace it
+            # and set new schema from df
+            self.client.create_table(
+                self.config.collection_name,
+                data=df,
+                mode="overwrite",
+            )
+            doc_cls = dataframe_to_document_model(
+                df,
+                content=content,
+                metadata=actual_metadata,
+                exclude=["vector"],
+            )
+            self.config.document_class = doc_cls  # type: ignore
+            self._setup_schemas(doc_cls)  # type: ignore
+        else:
+            # collection exists and is not empty, so append to it
+            tbl = self.client.open_table(self.config.collection_name)
+            tbl.add(df)
+
+    def delete_collection(self, collection_name: str) -> None:
+        self.client.drop_table(collection_name, ignore_missing=True)
+
+    def _lance_result_to_docs(self, result: LanceVectorQueryBuilder) -> List[Document]:
+        if self.is_from_dataframe:
+            df = result.to_pandas()
+            return dataframe_to_documents(
+                df,
+                content="content",
+                metadata=self.df_metadata_columns,
+                doc_cls=self.config.document_class,
+            )
+        else:
+            records = result.to_arrow().to_pylist()
+            return self._records_to_docs(records)
+
+    def _records_to_docs(self, records: List[Dict[str, Any]]) -> List[Document]:
+        if self.config.flatten:
+            docs = [
+                self.unflattened_schema(**nested_dict_from_flat(rec)) for rec in records
+            ]
+        else:
+            try:
+                docs = [self.schema(**rec) for rec in records]
+            except ValidationError as e:
+                raise ValueError(
+                    f"""
+                Error validating LanceDB result: {e}
+                HINT: This could happen when you're re-using an 
+                existing LanceDB store with a different schema.
+                Try deleting your local lancedb storage at `{self.config.storage_path}`
+                re-ingesting your documents and/or replacing the collections.
+                """
+                )
+
+        doc_cls = self.config.document_class
+        doc_cls_field_names = doc_cls.__fields__.keys()
+        return [
+            doc_cls(
+                **{
+                    field_name: getattr(doc, field_name)
+                    for field_name in doc_cls_field_names
+                }
+            )
+            for doc in docs
+        ]
+
+    def get_all_documents(self, where: str = "") -> List[Document]:
+        if self.config.collection_name is None:
+            raise ValueError("No collection name set, cannot retrieve docs")
+        tbl = self.client.open_table(self.config.collection_name)
+        pre_result = tbl.search(None).where(where or None).limit(None)
+        return self._lance_result_to_docs(pre_result)
+
+    def get_documents_by_ids(self, ids: List[str]) -> List[Document]:
+        if self.config.collection_name is None:
+            raise ValueError("No collection name set, cannot retrieve docs")
+        _ids = [str(id) for id in ids]
+        tbl = self.client.open_table(self.config.collection_name)
+        docs = []
+        for _id in _ids:
+            results = self._lance_result_to_docs(tbl.search().where(f"id == '{_id}'"))
+            if len(results) > 0:
+                docs.append(results[0])
+        return docs
+
+    def similar_texts_with_scores(
+        self,
+        text: str,
+        k: int = 1,
+        where: Optional[str] = None,
+    ) -> List[Tuple[Document, float]]:
+        embedding = self.embedding_fn([text])[0]
+        tbl = self.client.open_table(self.config.collection_name)
+        result = (
+            tbl.search(embedding)
+            .metric(self.config.distance)
+            .where(where, prefilter=True)
+            .limit(k)
+        )
+        docs = self._lance_result_to_docs(result)
+        # note _distance is 1 - cosine
+        if self.is_from_dataframe:
+            scores = [
+                1 - rec["_distance"] for rec in result.to_pandas().to_dict("records")
+            ]
+        else:
+            scores = [1 - rec["_distance"] for rec in result.to_arrow().to_pylist()]
+        if len(docs) == 0:
+            logger.warning(f"No matches found for {text}")
+            return []
+        if settings.debug:
+            logger.info(f"Found {len(docs)} matches, max score: {max(scores)}")
+        doc_score_pairs = list(zip(docs, scores))
+        self.show_if_debug(doc_score_pairs)
+        return doc_score_pairs

langroid/vector_store/meilisearch.py

@@ -32,7 +32,7 @@ class MeiliSearchConfig(VectorStoreConfig):
 
 
 class MeiliSearch(VectorStore):
-    def __init__(self, config: MeiliSearchConfig):
+    def __init__(self, config: MeiliSearchConfig = MeiliSearchConfig()):
         super().__init__(config)
         self.config: MeiliSearchConfig = config
         self.host = config.host
@@ -165,12 +165,13 @@ class MeiliSearch(VectorStore):
         async with self.client() as client:
             index = client.index(collection_name)
             await index.add_documents_in_batches(
-                documents=documents,  # type: ignore
+                documents=documents,
                 batch_size=self.config.batch_size,
                 primary_key=self.config.primary_key,
             )
 
     def add_documents(self, documents: Sequence[Document]) -> None:
+        super().maybe_add_ids(documents)
         if len(documents) == 0:
             return
         colls = self._list_all_collections()
@@ -197,18 +198,19 @@ class MeiliSearch(VectorStore):
         except ValueError:
             return id
 
-    async def _async_get_documents(self) -> DocumentsInfo:
+    async def _async_get_documents(self, where: str = "") -> DocumentsInfo:
         if self.config.collection_name is None:
             raise ValueError("No collection name set, cannot retrieve docs")
+        filter = [] if where is None else where
         async with self.client() as client:
             index = client.index(self.config.collection_name)
-            documents = await index.get_documents(limit=10_000)
+            documents = await index.get_documents(limit=10_000, filter=filter)
         return documents
 
-    def get_all_documents(self) -> List[Document]:
+    def get_all_documents(self, where: str = "") -> List[Document]:
         if self.config.collection_name is None:
             raise ValueError("No collection name set, cannot retrieve docs")
-        docs = asyncio.run(self._async_get_documents())
+        docs = asyncio.run(self._async_get_documents(where))
         if docs is None:
             return []
         doc_results = docs.results
@@ -226,7 +228,7 @@ class MeiliSearch(VectorStore):
         async with self.client() as client:
             index = client.index(self.config.collection_name)
             documents = await asyncio.gather(*[index.get_document(id) for id in ids])
-        return documents  # type: ignore
+        return documents
 
     def get_documents_by_ids(self, ids: List[str]) -> List[Document]:
         if self.config.collection_name is None:
@@ -263,6 +265,7 @@ class MeiliSearch(VectorStore):
         text: str,
         k: int = 20,
         where: Optional[str] = None,
+        neighbors: int = 0,  # ignored
     ) -> List[Tuple[Document, float]]:
         filter = [] if where is None else where
         if self.config.collection_name is None: