Skip to content

Vectorstore agent module

Here we handle connections and management of various vectorstore services and semantic seach.

VectorDatabaseAgentMilvus

The VectorDatabaseAgentMilvus class manages vector databases in a connected host database. It manages an embedding collection _col_embeddings:langchain.vectorstores.Milvus, which is the main information on the embedded text fragments and the basis for similarity search, and a metadata collection _col_metadata:pymilvus.Collection, which stores the metadata of the embedded text fragments. A typical workflow includes the following operations:

  1. connect to a host using connect()
  2. get all documents in the active database using get_all_documents()
  3. save a number of fragments, usually from a specific document, using store_embeddings()
  4. do similarity search among all fragments of the currently active database using similarity_search()
  5. remove a document from the currently active database using remove_document()
Source code in biochatter/vectorstore_agent.py
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
class VectorDatabaseAgentMilvus:
    """
    The VectorDatabaseAgentMilvus class manages vector databases in a connected
    host database. It manages an embedding collection
    `_col_embeddings:langchain.vectorstores.Milvus`, which is the main
    information on the embedded text fragments and the basis for similarity
    search, and a metadata collection `_col_metadata:pymilvus.Collection`, which
    stores the metadata of the embedded text fragments. A typical workflow
    includes the following operations:

    1. connect to a host using `connect()`
    2. get all documents in the active database using `get_all_documents()`
    3. save a number of fragments, usually from a specific document, using
        `store_embeddings()`
    4. do similarity search among all fragments of the currently active database
        using `similarity_search()`
    5. remove a document from the currently active database using
        `remove_document()`
    """

    def __init__(
        self,
        embedding_func: OpenAIEmbeddings,
        connection_args: Optional[Dict] = None,
        embedding_collection_name: Optional[str] = None,
        metadata_collection_name: Optional[str] = None,
    ):
        """
        Args:
            embedding_func OpenAIEmbeddings: Function used to embed the text

            connection_args Optional dict: args to connect Vector Database

            embedding_collection_name Optional str: exposed for test

            metadata_collection_name Optional str: exposed for test
        """
        self._embedding_func = embedding_func
        self._col_embeddings: Optional[Milvus] = None
        self._col_metadata: Optional[Collection] = None
        self._connection_args = validate_connection_args(connection_args)
        self._embedding_name = (
            embedding_collection_name or DOCUMENT_EMBEDDINGS_COLLECTION_NAME
        )
        self._metadata_name = (
            metadata_collection_name or DOCUMENT_METADATA_COLLECTION_NAME
        )

    def connect(self) -> None:
        """
        Connect to a host and read two document collections (the default names
        are `DocumentEmbeddings` and `DocumentMetadata`) in the currently active
        database (default database name is `default`); if those document
        collections don't exist, create the two collections.
        """
        self._connect(**self._connection_args)
        self._init_host()

    def _connect(self, host: str, port: str, user: str, password: str) -> None:
        self.alias = self._create_connection_alias(host, port, user, password)

    def _init_host(self) -> None:
        """
        Initialize host. Will read/create document collection inside currently
        active database.
        """
        self._create_collections()

    def _create_connection_alias(
        self, host: str, port: str, user: str, password: str
    ) -> str:
        """
        Connect to host and create a connection alias for metadata collection
        using a random uuid.

        Args:
            host (str): host ip address
            port (str): host port

        Returns:
            str: connection alias
        """
        alias = uuid.uuid4().hex
        try:
            connections.connect(
                host=host, port=port, user=user, password=password, alias=alias
            )
            logger.debug(f"Created new connection using: {alias}")
            return alias
        except MilvusException as e:
            logger.error(f"Failed to create  new connection using: {alias}")
            raise e

    def _create_collections(self) -> None:
        """
        Create or load the embedding and metadata collections from the currently
        active database.
        """
        embedding_exists = utility.has_collection(
            self._embedding_name, using=self.alias
        )
        meta_exists = utility.has_collection(
            self._metadata_name,
            using=self.alias,
        )

        if embedding_exists:
            self._load_embeddings_collection()
        else:
            self._create_embeddings_collection()

        if meta_exists:
            self._load_metadata_collection()
        else:
            self._create_metadata_collection()

        self._create_metadata_collection_index()
        self._col_metadata.load()

    def _load_embeddings_collection(self) -> None:
        """
        Load embeddings collection from currently active database.
        """
        try:
            self._col_embeddings = Milvus(
                embedding_function=self._embedding_func,
                collection_name=self._embedding_name,
                connection_args=self._connection_args,
            )
        except MilvusException as e:
            logger.error(
                f"Failed to load embeddings collection {self._embedding_name}."
            )
            raise e

    def _create_embeddings_collection(self) -> None:
        """
        Create embedding collection.
        All fields: "meta_id", "vector"
        """
        try:
            self._col_embeddings = Milvus(
                embedding_function=self._embedding_func,
                collection_name=self._embedding_name,
                connection_args=self._connection_args,
            )
        except MilvusException as e:
            logger.error(
                f"Failed to create embeddings collection {self._embedding_name}"
            )
            raise e

    def _load_metadata_collection(self) -> None:
        """
        Load metadata collection from currently active database.
        """
        self._col_metadata = Collection(
            self._metadata_name,
            using=self.alias,
        )
        self._col_metadata.load()

    def _create_metadata_collection(self) -> None:
        """
        Create metadata collection.

        All fields: "id", "name", "author", "title", "format", "subject",
        "creator", "producer", "creationDate", "modDate", "source", "embedding",
        "isDeleted".

        As the vector database requires a vector field, we will create a fake
        vector "embedding". The field "isDeleted" is used to specify if the
        document is deleted.
        """
        doc_id = FieldSchema(
            name="id", dtype=DataType.INT64, is_primary=True, auto_id=True
        )
        doc_name = FieldSchema(
            name="name", dtype=DataType.VARCHAR, max_length=255
        )
        doc_author = FieldSchema(
            name="author", dtype=DataType.VARCHAR, max_length=255
        )
        doc_title = FieldSchema(
            name="title", dtype=DataType.VARCHAR, max_length=1000
        )
        doc_format = FieldSchema(
            name="format", dtype=DataType.VARCHAR, max_length=255
        )
        doc_subject = FieldSchema(
            name="subject", dtype=DataType.VARCHAR, max_length=255
        )
        doc_creator = FieldSchema(
            name="creator", dtype=DataType.VARCHAR, max_length=255
        )
        doc_producer = FieldSchema(
            name="producer", dtype=DataType.VARCHAR, max_length=255
        )
        doc_creationDate = FieldSchema(
            name="creationDate", dtype=DataType.VARCHAR, max_length=64
        )
        doc_modDate = FieldSchema(
            name="modDate", dtype=DataType.VARCHAR, max_length=64
        )
        doc_source = FieldSchema(
            name="source", dtype=DataType.VARCHAR, max_length=1000
        )
        embedding = FieldSchema(
            name="embedding",
            dtype=DataType.FLOAT_VECTOR,
            dim=METADATA_VECTOR_DIM,
        )
        isDeleted = FieldSchema(
            name="isDeleted",
            dtype=DataType.BOOL,
        )
        fields = [
            doc_id,
            doc_name,
            doc_author,
            doc_title,
            doc_format,
            doc_subject,
            doc_creator,
            doc_producer,
            doc_creationDate,
            doc_modDate,
            doc_source,
            embedding,
            isDeleted,
        ]
        schema = CollectionSchema(fields=fields)
        try:
            self._col_metadata = Collection(
                name=self._metadata_name, schema=schema, using=self.alias
            )
        except MilvusException as e:
            logger.error(f"Failed to create collection {self._metadata_name}")
            raise e

    def _create_metadata_collection_index(self) -> None:
        """
        Create index for metadata collection in currently active database.
        """
        if (
            not isinstance(self._col_metadata, Collection)
            or len(self._col_metadata.indexes) > 0
        ):
            return

        index_params = {
            "metric_type": "L2",
            "index_type": "HNSW",
            "params": {"M": 8, "efConstruction": 64},
        }

        try:
            self._col_metadata.create_index(
                field_name="embedding",
                index_params=index_params,
                using=self.alias,
            )
        except MilvusException as e:
            logger.error(
                "Failed to create index for meta collection "
                f"{self._metadata_name}."
            )
            raise e

    def _insert_data(self, documents: List[Document]) -> str:
        """
        Insert documents into the currently active database.

        Args:
            documents (List[Documents]): documents array, usually from
                DocumentReader.load_document, DocumentReader.document_from_pdf,
                DocumentReader.document_from_txt

        Returns:
            str: document id
        """
        if len(documents) == 0:
            return None
        metadata = [documents[0].metadata]
        aligned_metadata = align_metadata(metadata)
        try:
            result = self._col_metadata.insert(aligned_metadata)
            meta_id = str(result.primary_keys[0])
            self._col_metadata.flush()
        except MilvusException as e:
            logger.error(f"Failed to insert meta data")
            raise e
        aligned_docs = align_embeddings(documents, meta_id)
        try:
            # As we passed collection_name, documents will be added to existed collection
            self._col_embeddings = Milvus.from_documents(
                embedding=self._embedding_func,
                collection_name=self._embedding_name,
                connection_args=self._connection_args,
                documents=aligned_docs,
            )
        except MilvusException as e:
            logger.error(
                "Failed to insert data to embedding collection "
                f"{self._embedding_name}."
            )
            raise e
        return meta_id

    def store_embeddings(self, documents: List[Document]) -> str:
        """
        Store documents in the currently active database.

        Args:
            documents (List[Documents]): documents array, usually from
                DocumentReader.load_document, DocumentReader.document_from_pdf,
                DocumentReader.document_from_txt

        Returns:
            str: document id
        """
        if len(documents) == 0:
            return
        return self._insert_data(documents)

    def _build_embedding_search_expression(
        self, meta_ids: List[Dict]
    ) -> Optional[str]:
        """
        Build search expression for embedding collection. The generated
        expression follows the pattern: "meta_id in [{id1}, {id2}, ...]

        Args:
            meta_ids: the array of metadata id in metadata collection

        Returns:
            str: search expression or None
        """
        if len(meta_ids) == 0:
            return "meta_id in []"
        built_expr = """meta_id in ["""
        for item in meta_ids:
            id = f'"{item["id"]}",'
            built_expr += id
        built_expr = built_expr[:-1]
        built_expr += """]"""
        return built_expr

    def _join_embedding_and_metadata_results(
        self, result_embedding: List[Document], result_meta: List[Dict]
    ) -> List[Document]:
        """
        Join the search results of embedding collection and results of metadata.

        Args:
            result_embedding (List[Document]): search result of embedding
                collection

            result_meta (List[Dict]): search result of metadata collection

        Returns:
            List[Document]: combined results like
                [{page_content: str, metadata: {...}}]
        """

        def _find_metadata_by_id(
            metadata: List[Dict], id: str
        ) -> Optional[Dict]:
            for d in metadata:
                if str(d["id"]) == id:
                    return d
            return None

        joined_docs = []
        for res in result_embedding:
            found = _find_metadata_by_id(result_meta, res.metadata["meta_id"])
            if found is None:  # discard
                logger.error(
                    f"Failed to join meta_id {res.metadata['meta_id']}"
                )
                continue
            joined_docs.append(
                Document(page_content=res.page_content, metadata=found)
            )
        return joined_docs

    @staticmethod
    def _build_meta_col_query_expr_for_all_documents(
        doc_ids: Optional[List[str]] = None,
    ) -> str:
        """
        Build metadata collection query expression to obtain all documents.

        Args:
            doc_ids: the list of document ids (metadata ids), if thie argument is None,
                     that is, the query is to get all undeleted documents in metadata collection.
                     Otherwise, the query is to getr all undeleted documents form provided doc_ids

        Returns:
            query: str
        """
        expr = (
            f"id in {doc_ids} and isDeleted == false"
            if doc_ids is not None
            else "isDeleted == false"
        )
        return expr.replace('"', "").replace("'", "")

    def similarity_search(
        self, query: str, k: int = 3, doc_ids: List[str] = None
    ) -> List[Document]:
        """
        Perform similarity search insider the currently active database
        according to the input query.

        This method will:
        1. get all non-deleted meta_id and build to search expression for
            the currently active embedding collection
        2. do similarity search in the embedding collection
        3. combine metadata and embeddings

        Args:
            query (str): query string

            k (int): the number of results to return

            doc_ids(List[str] optional): the list of document ids, do similarity search across the
            specified documents

        Returns:
            List[Document]: search results
        """
        result_metadata = []
        expr = VectorDatabaseAgentMilvus._build_meta_col_query_expr_for_all_documents(
            doc_ids
        )
        result_metadata = self._col_metadata.query(
            expr=expr, output_fields=METADATA_FIELDS
        )
        expr = self._build_embedding_search_expression(result_metadata)
        result_embedding = self._col_embeddings.similarity_search(
            query=query, k=k, expr=expr
        )
        return self._join_embedding_and_metadata_results(
            result_embedding, result_metadata
        )

    def remove_document(
        self, doc_id: str, doc_ids: Optional[List[str]] = None
    ) -> bool:
        """
        Remove the document include meta data and its embeddings.

        Args:
            doc_id (str): the document to be deleted

            doc_ids(List[str] optional): the list of document ids, defines documents scope
            within which remove operation occurs.

        Returns:
            bool: True if the document is deleted, False otherwise
        """
        if not self._col_metadata:
            return False
        if doc_ids is not None and (
            len(doc_ids) == 0 or (len(doc_ids) > 0 and not doc_id in doc_ids)
        ):
            return False
        try:
            expr = f"id in [{doc_id}]"
            res = self._col_metadata.query(
                expr=expr, output_fields=METADATA_FIELDS
            )
            if len(res) == 0:
                return False
            del_res = self._col_metadata.delete(expr)
            self._col_metadata.flush()

            res = self._col_embeddings.col.query(f'meta_id in ["{doc_id}"]')
            if len(res) == 0:
                return True
            ids = [item["pk"] for item in res]
            embedding_expr = f"pk in {ids}"
            del_res = self._col_embeddings.col.delete(expr=embedding_expr)
            self._col_embeddings.col.flush()
            return True
        except MilvusException as e:
            logger.error(e)
            raise e

    def get_all_documents(
        self, doc_ids: Optional[List[str]] = None
    ) -> List[Dict]:
        """
        Get all non-deleted documents from the currently active database.

        Args:
            doc_ids(List[str] optional): the list of document ids, defines documents scope within
            which the operation of obaining all documents occurs

        Returns:
            List[Dict]: the metadata of all non-deleted documents in the form
                [{{id}, {author}, {source}, ...}]
        """
        try:
            expr = VectorDatabaseAgentMilvus._build_meta_col_query_expr_for_all_documents(
                doc_ids
            )
            result_metadata = self._col_metadata.query(
                expr=expr, output_fields=METADATA_FIELDS
            )
            return result_metadata
        except MilvusException as e:
            logger.error(e)
            raise e

__init__(embedding_func, connection_args=None, embedding_collection_name=None, metadata_collection_name=None)

Parameters:

Name Type Description Default
embedding_func OpenAIEmbeddings

Function used to embed the text

required
connection_args Optional dict

args to connect Vector Database

None
embedding_collection_name Optional str

exposed for test

None
metadata_collection_name Optional str

exposed for test

None
Source code in biochatter/vectorstore_agent.py
def __init__(
    self,
    embedding_func: OpenAIEmbeddings,
    connection_args: Optional[Dict] = None,
    embedding_collection_name: Optional[str] = None,
    metadata_collection_name: Optional[str] = None,
):
    """
    Args:
        embedding_func OpenAIEmbeddings: Function used to embed the text

        connection_args Optional dict: args to connect Vector Database

        embedding_collection_name Optional str: exposed for test

        metadata_collection_name Optional str: exposed for test
    """
    self._embedding_func = embedding_func
    self._col_embeddings: Optional[Milvus] = None
    self._col_metadata: Optional[Collection] = None
    self._connection_args = validate_connection_args(connection_args)
    self._embedding_name = (
        embedding_collection_name or DOCUMENT_EMBEDDINGS_COLLECTION_NAME
    )
    self._metadata_name = (
        metadata_collection_name or DOCUMENT_METADATA_COLLECTION_NAME
    )

connect()

Connect to a host and read two document collections (the default names are DocumentEmbeddings and DocumentMetadata) in the currently active database (default database name is default); if those document collections don't exist, create the two collections.

Source code in biochatter/vectorstore_agent.py
def connect(self) -> None:
    """
    Connect to a host and read two document collections (the default names
    are `DocumentEmbeddings` and `DocumentMetadata`) in the currently active
    database (default database name is `default`); if those document
    collections don't exist, create the two collections.
    """
    self._connect(**self._connection_args)
    self._init_host()

get_all_documents(doc_ids=None)

Get all non-deleted documents from the currently active database.

Parameters:

Name Type Description Default
doc_ids(List[str] optional

the list of document ids, defines documents scope within

required

Returns:

Type Description
List[Dict]

List[Dict]: the metadata of all non-deleted documents in the form [{{id}, {author}, {source}, ...}]

Source code in biochatter/vectorstore_agent.py
def get_all_documents(
    self, doc_ids: Optional[List[str]] = None
) -> List[Dict]:
    """
    Get all non-deleted documents from the currently active database.

    Args:
        doc_ids(List[str] optional): the list of document ids, defines documents scope within
        which the operation of obaining all documents occurs

    Returns:
        List[Dict]: the metadata of all non-deleted documents in the form
            [{{id}, {author}, {source}, ...}]
    """
    try:
        expr = VectorDatabaseAgentMilvus._build_meta_col_query_expr_for_all_documents(
            doc_ids
        )
        result_metadata = self._col_metadata.query(
            expr=expr, output_fields=METADATA_FIELDS
        )
        return result_metadata
    except MilvusException as e:
        logger.error(e)
        raise e

remove_document(doc_id, doc_ids=None)

Remove the document include meta data and its embeddings.

Parameters:

Name Type Description Default
doc_id str

the document to be deleted

required
doc_ids(List[str] optional

the list of document ids, defines documents scope

required

Returns:

Name Type Description
bool bool

True if the document is deleted, False otherwise

Source code in biochatter/vectorstore_agent.py
def remove_document(
    self, doc_id: str, doc_ids: Optional[List[str]] = None
) -> bool:
    """
    Remove the document include meta data and its embeddings.

    Args:
        doc_id (str): the document to be deleted

        doc_ids(List[str] optional): the list of document ids, defines documents scope
        within which remove operation occurs.

    Returns:
        bool: True if the document is deleted, False otherwise
    """
    if not self._col_metadata:
        return False
    if doc_ids is not None and (
        len(doc_ids) == 0 or (len(doc_ids) > 0 and not doc_id in doc_ids)
    ):
        return False
    try:
        expr = f"id in [{doc_id}]"
        res = self._col_metadata.query(
            expr=expr, output_fields=METADATA_FIELDS
        )
        if len(res) == 0:
            return False
        del_res = self._col_metadata.delete(expr)
        self._col_metadata.flush()

        res = self._col_embeddings.col.query(f'meta_id in ["{doc_id}"]')
        if len(res) == 0:
            return True
        ids = [item["pk"] for item in res]
        embedding_expr = f"pk in {ids}"
        del_res = self._col_embeddings.col.delete(expr=embedding_expr)
        self._col_embeddings.col.flush()
        return True
    except MilvusException as e:
        logger.error(e)
        raise e

Perform similarity search insider the currently active database according to the input query.

This method will: 1. get all non-deleted meta_id and build to search expression for the currently active embedding collection 2. do similarity search in the embedding collection 3. combine metadata and embeddings

Parameters:

Name Type Description Default
query str

query string

required
k int

the number of results to return

3
doc_ids(List[str] optional

the list of document ids, do similarity search across the

required

Returns:

Type Description
List[Document]

List[Document]: search results

Source code in biochatter/vectorstore_agent.py
def similarity_search(
    self, query: str, k: int = 3, doc_ids: List[str] = None
) -> List[Document]:
    """
    Perform similarity search insider the currently active database
    according to the input query.

    This method will:
    1. get all non-deleted meta_id and build to search expression for
        the currently active embedding collection
    2. do similarity search in the embedding collection
    3. combine metadata and embeddings

    Args:
        query (str): query string

        k (int): the number of results to return

        doc_ids(List[str] optional): the list of document ids, do similarity search across the
        specified documents

    Returns:
        List[Document]: search results
    """
    result_metadata = []
    expr = VectorDatabaseAgentMilvus._build_meta_col_query_expr_for_all_documents(
        doc_ids
    )
    result_metadata = self._col_metadata.query(
        expr=expr, output_fields=METADATA_FIELDS
    )
    expr = self._build_embedding_search_expression(result_metadata)
    result_embedding = self._col_embeddings.similarity_search(
        query=query, k=k, expr=expr
    )
    return self._join_embedding_and_metadata_results(
        result_embedding, result_metadata
    )

store_embeddings(documents)

Store documents in the currently active database.

Parameters:

Name Type Description Default
documents List[Documents]

documents array, usually from DocumentReader.load_document, DocumentReader.document_from_pdf, DocumentReader.document_from_txt

required

Returns:

Name Type Description
str str

document id

Source code in biochatter/vectorstore_agent.py
def store_embeddings(self, documents: List[Document]) -> str:
    """
    Store documents in the currently active database.

    Args:
        documents (List[Documents]): documents array, usually from
            DocumentReader.load_document, DocumentReader.document_from_pdf,
            DocumentReader.document_from_txt

    Returns:
        str: document id
    """
    if len(documents) == 0:
        return
    return self._insert_data(documents)

align_embeddings(docs, meta_id)

Ensure that the metadata id is present in each document.

Parameters:

Name Type Description Default
docs List[Document]

List of documents

required
meta_id int

Metadata id to assign to the documents

required

Returns:

Type Description
List[Document]

List[Document]: List of documents, with each document having a metadata id.

Source code in biochatter/vectorstore_agent.py
def align_embeddings(docs: List[Document], meta_id: int) -> List[Document]:
    """
    Ensure that the metadata id is present in each document.

    Args:
        docs (List[Document]): List of documents

        meta_id (int): Metadata id to assign to the documents

    Returns:
        List[Document]: List of documents, with each document having a metadata
            id.
    """
    ret = []
    for doc in docs:
        ret.append(
            Document(
                page_content=doc.page_content,
                metadata={"meta_id": meta_id},
            )
        )
    return ret

align_metadata(metadata, isDeleted=False)

Ensure that specific metadata fields are present; if not provided, fill with "unknown". Also, add a random vector to each metadata item to simulate an embedding.

Parameters:

Name Type Description Default
metadata List[Dict]

List of metadata items

required
isDeleted Optional[bool]

Whether the document is deleted. Defaults to False.

False

Returns:

Type Description
List[List]

List[List]: List of metadata items, with each item being a list of metadata fields.

Source code in biochatter/vectorstore_agent.py
def align_metadata(
    metadata: List[Dict], isDeleted: Optional[bool] = False
) -> List[List]:
    """

    Ensure that specific metadata fields are present; if not provided, fill with
    "unknown". Also, add a random vector to each metadata item to simulate an
    embedding.

    Args:
        metadata (List[Dict]): List of metadata items

        isDeleted (Optional[bool], optional): Whether the document is deleted.
            Defaults to False.

    Returns:
        List[List]: List of metadata items, with each item being a list of
            metadata fields.
    """
    ret = []
    fields = METADATA_FIELDS.copy()
    fields.pop(0)
    for ix, k in enumerate(fields):
        ret.append([item[k] if k in item else "unknown" for item in metadata])

    ret.append(
        [
            [random.random() for _ in range(METADATA_VECTOR_DIM)]
            for _ in range(len(metadata))
        ]
    )
    ret.append([isDeleted for _ in metadata])
    return ret