Skip to content

Vectorstore Agent Reference

Here we handle the application of vectorstore services to retrieval-augmented generation tasks by embedding documents and connections/management of vectorstore services and semantic search.

Vectorstore Implementation

DocumentEmbedder

Source code in biochatter/vectorstore.py 
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
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
class DocumentEmbedder:
    def __init__(
        self,
        used: bool = False,
        online: bool = False,
        chunk_size: int = 1000,
        chunk_overlap: int = 0,
        split_by_characters: bool = True,
        separators: Optional[list] = None,
        n_results: int = 3,
        model: Optional[str] = "text-embedding-ada-002",
        vector_db_vendor: Optional[str] = None,
        connection_args: Optional[dict] = None,
        embedding_collection_name: Optional[str] = None,
        metadata_collection_name: Optional[str] = None,
        api_key: Optional[str] = None,
        is_azure: Optional[bool] = False,
        azure_deployment: Optional[str] = None,
        azure_endpoint: Optional[str] = None,
        base_url: Optional[str] = None,
        embeddings: Optional[
            OpenAIEmbeddings | XinferenceEmbeddings | OllamaEmbeddings
        ] = None,
        documentids_workspace: Optional[list[str]] = None,
    ) -> None:
        """
        Class that handles the retrieval-augmented generation (RAG) functionality
        of BioChatter. It splits text into chunks, embeds them, and stores them in
        a vector database. It can then be used to do similarity search on the
        database.

        Args:

            used (bool, optional): whether RAG has been used (frontend setting).
                Defaults to False.

            online (bool, optional): whether we are running the frontend online.
                Defaults to False.

            chunk_size (int, optional): size of chunks to split text into.
                Defaults to 1000.

            chunk_overlap (int, optional): overlap between chunks. Defaults to 0.

            split_by_characters (bool, optional): whether to split by characters
                or tokens. Defaults to True.

            separators (Optional[list], optional): list of separators to use when
                splitting by characters. Defaults to [" ", ",", "\n"].

            n_results (int, optional): number of results to return from
                similarity search. Defaults to 3.

            model (Optional[str], optional): name of model to use for embeddings.
                Defaults to 'text-embedding-ada-002'.

            vector_db_vendor (Optional[str], optional): name of vector database
                to use. Defaults to Milvus.

            connection_args (Optional[dict], optional): arguments to pass to
                vector database connection. Defaults to None.

            api_key (Optional[str], optional): OpenAI API key. Defaults to None.

            base_url (Optional[str], optional): base url of OpenAI API.

            embeddings (Optional[OpenAIEmbeddings | XinferenceEmbeddings],
                optional): Embeddings object to use. Defaults to OpenAI.

            documentids_workspace (Optional[List[str]], optional): a list of document IDs
                that defines the scope within which rag operations (remove, similarity search,
                and get all) occur. Defaults to None, which means the operations will be
                performed across all documents in the database.

            is_azure (Optional[bool], optional): if we are using Azure
            azure_deployment (Optional[str], optional): Azure embeddings model deployment,
                should work with azure_endpoint when is_azure is True
            azure_endpoint (Optional[str], optional): Azure endpoint, should work with
                azure_deployment when is_azure is True

        """
        self.used = used
        self.online = online
        self.chunk_size = chunk_size
        self.chunk_overlap = chunk_overlap
        self.separators = separators or [" ", ",", "\n"]
        self.n_results = n_results
        self.split_by_characters = split_by_characters
        self.model_name = model

        # TODO API Key handling to central config?
        if base_url:
            openai.api_base = base_url

        if embeddings:
            self.embeddings = embeddings
        else:
            if not self.online:
                self.embeddings = (
                    OpenAIEmbeddings(openai_api_key=api_key, model=model)
                    if not is_azure
                    else AzureOpenAIEmbeddings(
                        api_key=api_key,
                        azure_deployment=azure_deployment,
                        azure_endpoint=azure_endpoint,
                        model=model,
                    )
                )
            else:
                self.embeddings = None

        # connection arguments
        self.connection_args = connection_args or {
            "host": "127.0.0.1",
            "port": "19530",
        }
        self.embedding_collection_name = embedding_collection_name
        self.metadata_collection_name = metadata_collection_name
        self.documentids_workspace = documentids_workspace

        # TODO: vector db selection
        self.vector_db_vendor = vector_db_vendor or "milvus"
        # instantiate VectorDatabaseHost
        self.database_host = None
        self._init_database_host()

    def _set_embeddings(self, embeddings):
        print("setting embedder")
        self.embeddings = embeddings

    def _init_database_host(self):
        if self.vector_db_vendor == "milvus":
            self.database_host = VectorDatabaseAgentMilvus(
                embedding_func=self.embeddings,
                connection_args=self.connection_args,
                embedding_collection_name=self.embedding_collection_name,
                metadata_collection_name=self.metadata_collection_name,
            )
        else:
            raise NotImplementedError(self.vector_db_vendor)

    def set_chunk_siue(self, chunk_size: int) -> None:
        self.chunk_size = chunk_size

    def set_chunk_overlap(self, chunk_overlap: int) -> None:
        self.chunk_overlap = chunk_overlap

    def set_separators(self, separators: list) -> None:
        self.separators = separators

    def _characters_splitter(self) -> RecursiveCharacterTextSplitter:
        return RecursiveCharacterTextSplitter(
            chunk_size=self.chunk_size,
            chunk_overlap=self.chunk_overlap,
            separators=self.separators,
        )

    def _tokens_splitter(self) -> RecursiveCharacterTextSplitter:
        DEFAULT_OPENAI_MODEL = "gpt-3.5-turbo"
        HUGGINGFACE_MODELS = ["bigscience/bloom"]
        if self.model_name and self.model_name in HUGGINGFACE_MODELS:
            tokenizer = GPT2TokenizerFast.from_pretrained("gpt2")
            return RecursiveCharacterTextSplitter.from_huggingface_tokenizer(
                tokenizer,
                chunk_size=self.chunk_size,
                chunk_overlap=self.chunk_overlap,
                separators=self.separators,
            )
        else:
            return RecursiveCharacterTextSplitter.from_tiktoken_encoder(
                encoding_name="",
                model_name=(
                    DEFAULT_OPENAI_MODEL
                    if not self.model_name
                    else self.model_name
                ),
                chunk_size=self.chunk_size,
                chunk_overlap=self.chunk_overlap,
                separators=self.separators,
            )

    def _text_splitter(self) -> RecursiveCharacterTextSplitter:
        return (
            self._characters_splitter()
            if self.split_by_characters
            else self._tokens_splitter()
        )

    def save_document(self, doc: list[Document]) -> str:
        """
        This function saves document to the vector database
        Args:
            doc List[Document]: document content, read with DocumentReader load_document(),
                or document_from_pdf(), document_from_txt()
        Returns:
            str: document id, which can be used to remove an uploaded document with remove_document()
        """
        splitted = self._split_document(doc)
        return self._store_embeddings(splitted)

    def _split_document(self, document: list[Document]) -> list[Document]:
        text_splitter = self._text_splitter()
        return text_splitter.split_documents(document)

    def _store_embeddings(self, doc: list[Document]) -> str:
        return self.database_host.store_embeddings(documents=doc)

    def connect(self) -> None:
        self.database_host.connect()

    def get_all_documents(self) -> list[dict]:
        return self.database_host.get_all_documents(
            doc_ids=self.documentids_workspace
        )

    def remove_document(self, doc_id: str) -> None:
        return self.database_host.remove_document(
            doc_id, self.documentids_workspace
        )

__init__(used=False, online=False, chunk_size=1000, chunk_overlap=0, split_by_characters=True, separators=None, n_results=3, model='text-embedding-ada-002', vector_db_vendor=None, connection_args=None, embedding_collection_name=None, metadata_collection_name=None, api_key=None, is_azure=False, azure_deployment=None, azure_endpoint=None, base_url=None, embeddings=None, documentids_workspace=None) 

    Class that handles the retrieval-augmented generation (RAG) functionality
    of BioChatter. It splits text into chunks, embeds them, and stores them in
    a vector database. It can then be used to do similarity search on the
    database.

    Args:

        used (bool, optional): whether RAG has been used (frontend setting).
            Defaults to False.

        online (bool, optional): whether we are running the frontend online.
            Defaults to False.

        chunk_size (int, optional): size of chunks to split text into.
            Defaults to 1000.

        chunk_overlap (int, optional): overlap between chunks. Defaults to 0.

        split_by_characters (bool, optional): whether to split by characters
            or tokens. Defaults to True.

        separators (Optional[list], optional): list of separators to use when
            splitting by characters. Defaults to [" ", ",", "

"].

        n_results (int, optional): number of results to return from
            similarity search. Defaults to 3.

        model (Optional[str], optional): name of model to use for embeddings.
            Defaults to 'text-embedding-ada-002'.

        vector_db_vendor (Optional[str], optional): name of vector database
            to use. Defaults to Milvus.

        connection_args (Optional[dict], optional): arguments to pass to
            vector database connection. Defaults to None.

        api_key (Optional[str], optional): OpenAI API key. Defaults to None.

        base_url (Optional[str], optional): base url of OpenAI API.

        embeddings (Optional[OpenAIEmbeddings | XinferenceEmbeddings],
            optional): Embeddings object to use. Defaults to OpenAI.

        documentids_workspace (Optional[List[str]], optional): a list of document IDs
            that defines the scope within which rag operations (remove, similarity search,
            and get all) occur. Defaults to None, which means the operations will be
            performed across all documents in the database.

        is_azure (Optional[bool], optional): if we are using Azure
        azure_deployment (Optional[str], optional): Azure embeddings model deployment,
            should work with azure_endpoint when is_azure is True
        azure_endpoint (Optional[str], optional): Azure endpoint, should work with
            azure_deployment when is_azure is True
Source code in biochatter/vectorstore.py 
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
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
def __init__(
    self,
    used: bool = False,
    online: bool = False,
    chunk_size: int = 1000,
    chunk_overlap: int = 0,
    split_by_characters: bool = True,
    separators: Optional[list] = None,
    n_results: int = 3,
    model: Optional[str] = "text-embedding-ada-002",
    vector_db_vendor: Optional[str] = None,
    connection_args: Optional[dict] = None,
    embedding_collection_name: Optional[str] = None,
    metadata_collection_name: Optional[str] = None,
    api_key: Optional[str] = None,
    is_azure: Optional[bool] = False,
    azure_deployment: Optional[str] = None,
    azure_endpoint: Optional[str] = None,
    base_url: Optional[str] = None,
    embeddings: Optional[
        OpenAIEmbeddings | XinferenceEmbeddings | OllamaEmbeddings
    ] = None,
    documentids_workspace: Optional[list[str]] = None,
) -> None:
    """
    Class that handles the retrieval-augmented generation (RAG) functionality
    of BioChatter. It splits text into chunks, embeds them, and stores them in
    a vector database. It can then be used to do similarity search on the
    database.

    Args:

        used (bool, optional): whether RAG has been used (frontend setting).
            Defaults to False.

        online (bool, optional): whether we are running the frontend online.
            Defaults to False.

        chunk_size (int, optional): size of chunks to split text into.
            Defaults to 1000.

        chunk_overlap (int, optional): overlap between chunks. Defaults to 0.

        split_by_characters (bool, optional): whether to split by characters
            or tokens. Defaults to True.

        separators (Optional[list], optional): list of separators to use when
            splitting by characters. Defaults to [" ", ",", "\n"].

        n_results (int, optional): number of results to return from
            similarity search. Defaults to 3.

        model (Optional[str], optional): name of model to use for embeddings.
            Defaults to 'text-embedding-ada-002'.

        vector_db_vendor (Optional[str], optional): name of vector database
            to use. Defaults to Milvus.

        connection_args (Optional[dict], optional): arguments to pass to
            vector database connection. Defaults to None.

        api_key (Optional[str], optional): OpenAI API key. Defaults to None.

        base_url (Optional[str], optional): base url of OpenAI API.

        embeddings (Optional[OpenAIEmbeddings | XinferenceEmbeddings],
            optional): Embeddings object to use. Defaults to OpenAI.

        documentids_workspace (Optional[List[str]], optional): a list of document IDs
            that defines the scope within which rag operations (remove, similarity search,
            and get all) occur. Defaults to None, which means the operations will be
            performed across all documents in the database.

        is_azure (Optional[bool], optional): if we are using Azure
        azure_deployment (Optional[str], optional): Azure embeddings model deployment,
            should work with azure_endpoint when is_azure is True
        azure_endpoint (Optional[str], optional): Azure endpoint, should work with
            azure_deployment when is_azure is True

    """
    self.used = used
    self.online = online
    self.chunk_size = chunk_size
    self.chunk_overlap = chunk_overlap
    self.separators = separators or [" ", ",", "\n"]
    self.n_results = n_results
    self.split_by_characters = split_by_characters
    self.model_name = model

    # TODO API Key handling to central config?
    if base_url:
        openai.api_base = base_url

    if embeddings:
        self.embeddings = embeddings
    else:
        if not self.online:
            self.embeddings = (
                OpenAIEmbeddings(openai_api_key=api_key, model=model)
                if not is_azure
                else AzureOpenAIEmbeddings(
                    api_key=api_key,
                    azure_deployment=azure_deployment,
                    azure_endpoint=azure_endpoint,
                    model=model,
                )
            )
        else:
            self.embeddings = None

    # connection arguments
    self.connection_args = connection_args or {
        "host": "127.0.0.1",
        "port": "19530",
    }
    self.embedding_collection_name = embedding_collection_name
    self.metadata_collection_name = metadata_collection_name
    self.documentids_workspace = documentids_workspace

    # TODO: vector db selection
    self.vector_db_vendor = vector_db_vendor or "milvus"
    # instantiate VectorDatabaseHost
    self.database_host = None
    self._init_database_host()

save_document(doc)

This function saves document to the vector database Args: doc List[Document]: document content, read with DocumentReader load_document(), or document_from_pdf(), document_from_txt() Returns: str: document id, which can be used to remove an uploaded document with remove_document()

Source code in biochatter/vectorstore.py 
208
209
210
211
212
213
214
215
216
217
218
def save_document(self, doc: list[Document]) -> str:
    """
    This function saves document to the vector database
    Args:
        doc List[Document]: document content, read with DocumentReader load_document(),
            or document_from_pdf(), document_from_txt()
    Returns:
        str: document id, which can be used to remove an uploaded document with remove_document()
    """
    splitted = self._split_document(doc)
    return self._store_embeddings(splitted)

DocumentReader

Source code in biochatter/vectorstore.py 
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
class DocumentReader:
    def load_document(self, path: str) -> list[Document]:
        """
        Loads a document from a path; accepts txt and pdf files. Txt files are
        loaded as-is, pdf files are converted to text using fitz.

        Args:
            path (str): path to document

        Returns:
            List[Document]: list of documents
        """
        if path.endswith(".txt"):
            loader = TextLoader(path)
            return loader.load()

        elif path.endswith(".pdf"):
            doc = fitz.open(path)
            text = ""
            for page in doc:
                text += page.get_text()

            meta = {k: v for k, v in doc.metadata.items() if v}
            meta.update({"source": path})

            return [
                Document(
                    page_content=text,
                    metadata=meta,
                )
            ]

    def document_from_pdf(self, pdf: bytes) -> list[Document]:
        """
        Receive a byte representation of a pdf file and return a list of Documents
        with metadata.

        Args:
            pdf (bytes): byte representation of pdf file

        Returns:
            List[Document]: list of documents
        """
        doc = fitz.open(stream=pdf, filetype="pdf")
        text = ""
        for page in doc:
            text += page.get_text()

        meta = {k: v for k, v in doc.metadata.items() if v}
        meta.update({"source": "pdf"})

        return [
            Document(
                page_content=text,
                metadata=meta,
            )
        ]

    def document_from_txt(self, txt: bytes) -> list[Document]:
        """
        Receive a byte representation of a txt file and return a list of Documents
        with metadata.

        Args:
            txt (bytes): byte representation of txt file

        Returns:
            List[Document]: list of documents
        """
        meta = {"source": "txt"}
        return [
            Document(
                page_content=txt,
                metadata=meta,
            )
        ]

document_from_pdf(pdf)

Receive a byte representation of a pdf file and return a list of Documents with metadata.

Parameters:

Name Type Description Default
pdf bytes

byte representation of pdf file

 required

Returns:

Type Description
list[Document]

List[Document]: list of documents
Source code in biochatter/vectorstore.py 
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
def document_from_pdf(self, pdf: bytes) -> list[Document]:
    """
    Receive a byte representation of a pdf file and return a list of Documents
    with metadata.

    Args:
        pdf (bytes): byte representation of pdf file

    Returns:
        List[Document]: list of documents
    """
    doc = fitz.open(stream=pdf, filetype="pdf")
    text = ""
    for page in doc:
        text += page.get_text()

    meta = {k: v for k, v in doc.metadata.items() if v}
    meta.update({"source": "pdf"})

    return [
        Document(
            page_content=text,
            metadata=meta,
        )
    ]

document_from_txt(txt)

Receive a byte representation of a txt file and return a list of Documents with metadata.

Parameters:

Name Type Description Default
txt bytes

byte representation of txt file

 required

Returns:

Type Description
list[Document]

List[Document]: list of documents
Source code in biochatter/vectorstore.py 
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
def document_from_txt(self, txt: bytes) -> list[Document]:
    """
    Receive a byte representation of a txt file and return a list of Documents
    with metadata.

    Args:
        txt (bytes): byte representation of txt file

    Returns:
        List[Document]: list of documents
    """
    meta = {"source": "txt"}
    return [
        Document(
            page_content=txt,
            metadata=meta,
        )
    ]

load_document(path)

Loads a document from a path; accepts txt and pdf files. Txt files are loaded as-is, pdf files are converted to text using fitz.

Parameters:

Name Type Description Default
path str

path to document

 required

Returns:

Type Description
list[Document]

List[Document]: list of documents
Source code in biochatter/vectorstore.py 
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
def load_document(self, path: str) -> list[Document]:
    """
    Loads a document from a path; accepts txt and pdf files. Txt files are
    loaded as-is, pdf files are converted to text using fitz.

    Args:
        path (str): path to document

    Returns:
        List[Document]: list of documents
    """
    if path.endswith(".txt"):
        loader = TextLoader(path)
        return loader.load()

    elif path.endswith(".pdf"):
        doc = fitz.open(path)
        text = ""
        for page in doc:
            text += page.get_text()

        meta = {k: v for k, v in doc.metadata.items() if v}
        meta.update({"source": path})

        return [
            Document(
                page_content=text,
                metadata=meta,
            )
        ]

OllamaDocumentEmbedder

Bases: DocumentEmbedder

Source code in biochatter/vectorstore.py 
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
class OllamaDocumentEmbedder(DocumentEmbedder):
    def __init__(
        self,
        used: bool = False,
        chunk_size: int = 1000,
        chunk_overlap: int = 0,
        split_by_characters: bool = True,
        separators: Optional[list] = None,
        n_results: int = 3,
        model: Optional[str] = "nomic-embed-text",
        vector_db_vendor: Optional[str] = None,
        connection_args: Optional[dict] = None,
        embedding_collection_name: Optional[str] = None,
        metadata_collection_name: Optional[str] = None,
        api_key: Optional[str] = "none",
        base_url: Optional[str] = None,
        documentids_workspace: Optional[list[str]] = None,
    ):
        """
        Extension of the DocumentEmbedder class that uses Ollama for
        embeddings.

        Args:

            used (bool, optional): whether RAG has been used (frontend setting).

            chunk_size (int, optional): size of chunks to split text into.

            chunk_overlap (int, optional): overlap between chunks.

            split_by_characters (bool, optional): whether to split by characters
            or tokens.

            separators (Optional[list], optional): list of separators to use when
            splitting by characters.

            n_results (int, optional): number of results to return from
            similarity search.

            model (Optional[str], optional): name of model to use for embeddings.
            Can be "auto" to use the first available model.

            vector_db_vendor (Optional[str], optional): name of vector database
            to use.

            connection_args (Optional[dict], optional): arguments to pass to
            vector database connection.

            embedding_collection_name (Optional[str], optional): name of
            collection to store embeddings in.

            metadata_collection_name (Optional[str], optional): name of
            collection to store metadata in.

            api_key (Optional[str], optional): Xinference API key.

            base_url (Optional[str], optional): base url of Xinference API.

            documentids_workspace (Optional[List[str]], optional): a list of document IDs
            that defines the scope within which rag operations (remove, similarity search,
            and get all) occur. Defaults to None, which means the operations will be
            performed across all documents in the database.

        """
        from langchain_community.embeddings import OllamaEmbeddings

        self.model_name = model

        super().__init__(
            used=used,
            online=True,
            chunk_size=chunk_size,
            chunk_overlap=chunk_overlap,
            split_by_characters=split_by_characters,
            separators=separators,
            n_results=n_results,
            model=model,
            vector_db_vendor=vector_db_vendor,
            connection_args=connection_args,
            embedding_collection_name=embedding_collection_name,
            metadata_collection_name=metadata_collection_name,
            api_key=api_key,
            base_url=base_url,
            embeddings=OllamaEmbeddings(
                base_url=base_url, model=self.model_name
            ),
            documentids_workspace=documentids_workspace,
        )

__init__(used=False, chunk_size=1000, chunk_overlap=0, split_by_characters=True, separators=None, n_results=3, model='nomic-embed-text', vector_db_vendor=None, connection_args=None, embedding_collection_name=None, metadata_collection_name=None, api_key='none', base_url=None, documentids_workspace=None)

Extension of the DocumentEmbedder class that uses Ollama for embeddings.

Args:

used (bool, optional): whether RAG has been used (frontend setting).

chunk_size (int, optional): size of chunks to split text into.

chunk_overlap (int, optional): overlap between chunks.

split_by_characters (bool, optional): whether to split by characters
or tokens.

separators (Optional[list], optional): list of separators to use when
splitting by characters.

n_results (int, optional): number of results to return from
similarity search.

model (Optional[str], optional): name of model to use for embeddings.
Can be "auto" to use the first available model.

vector_db_vendor (Optional[str], optional): name of vector database
to use.

connection_args (Optional[dict], optional): arguments to pass to
vector database connection.

embedding_collection_name (Optional[str], optional): name of
collection to store embeddings in.

metadata_collection_name (Optional[str], optional): name of
collection to store metadata in.

api_key (Optional[str], optional): Xinference API key.

base_url (Optional[str], optional): base url of Xinference API.

documentids_workspace (Optional[List[str]], optional): a list of document IDs
that defines the scope within which rag operations (remove, similarity search,
and get all) occur. Defaults to None, which means the operations will be
performed across all documents in the database.
Source code in biochatter/vectorstore.py 
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
def __init__(
    self,
    used: bool = False,
    chunk_size: int = 1000,
    chunk_overlap: int = 0,
    split_by_characters: bool = True,
    separators: Optional[list] = None,
    n_results: int = 3,
    model: Optional[str] = "nomic-embed-text",
    vector_db_vendor: Optional[str] = None,
    connection_args: Optional[dict] = None,
    embedding_collection_name: Optional[str] = None,
    metadata_collection_name: Optional[str] = None,
    api_key: Optional[str] = "none",
    base_url: Optional[str] = None,
    documentids_workspace: Optional[list[str]] = None,
):
    """
    Extension of the DocumentEmbedder class that uses Ollama for
    embeddings.

    Args:

        used (bool, optional): whether RAG has been used (frontend setting).

        chunk_size (int, optional): size of chunks to split text into.

        chunk_overlap (int, optional): overlap between chunks.

        split_by_characters (bool, optional): whether to split by characters
        or tokens.

        separators (Optional[list], optional): list of separators to use when
        splitting by characters.

        n_results (int, optional): number of results to return from
        similarity search.

        model (Optional[str], optional): name of model to use for embeddings.
        Can be "auto" to use the first available model.

        vector_db_vendor (Optional[str], optional): name of vector database
        to use.

        connection_args (Optional[dict], optional): arguments to pass to
        vector database connection.

        embedding_collection_name (Optional[str], optional): name of
        collection to store embeddings in.

        metadata_collection_name (Optional[str], optional): name of
        collection to store metadata in.

        api_key (Optional[str], optional): Xinference API key.

        base_url (Optional[str], optional): base url of Xinference API.

        documentids_workspace (Optional[List[str]], optional): a list of document IDs
        that defines the scope within which rag operations (remove, similarity search,
        and get all) occur. Defaults to None, which means the operations will be
        performed across all documents in the database.

    """
    from langchain_community.embeddings import OllamaEmbeddings

    self.model_name = model

    super().__init__(
        used=used,
        online=True,
        chunk_size=chunk_size,
        chunk_overlap=chunk_overlap,
        split_by_characters=split_by_characters,
        separators=separators,
        n_results=n_results,
        model=model,
        vector_db_vendor=vector_db_vendor,
        connection_args=connection_args,
        embedding_collection_name=embedding_collection_name,
        metadata_collection_name=metadata_collection_name,
        api_key=api_key,
        base_url=base_url,
        embeddings=OllamaEmbeddings(
            base_url=base_url, model=self.model_name
        ),
        documentids_workspace=documentids_workspace,
    )

XinferenceDocumentEmbedder

Bases: DocumentEmbedder

Source code in biochatter/vectorstore.py 
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
class XinferenceDocumentEmbedder(DocumentEmbedder):
    def __init__(
        self,
        used: bool = False,
        chunk_size: int = 1000,
        chunk_overlap: int = 0,
        split_by_characters: bool = True,
        separators: Optional[list] = None,
        n_results: int = 3,
        model: Optional[str] = "auto",
        vector_db_vendor: Optional[str] = None,
        connection_args: Optional[dict] = None,
        embedding_collection_name: Optional[str] = None,
        metadata_collection_name: Optional[str] = None,
        api_key: Optional[str] = "none",
        base_url: Optional[str] = None,
        documentids_workspace: Optional[list[str]] = None,
    ):
        """
        Extension of the DocumentEmbedder class that uses Xinference for
        embeddings.

        Args:

            used (bool, optional): whether RAG has been used (frontend setting).

            chunk_size (int, optional): size of chunks to split text into.

            chunk_overlap (int, optional): overlap between chunks.

            split_by_characters (bool, optional): whether to split by characters
            or tokens.

            separators (Optional[list], optional): list of separators to use when
            splitting by characters.

            n_results (int, optional): number of results to return from
            similarity search.

            model (Optional[str], optional): name of model to use for embeddings.
            Can be "auto" to use the first available model.

            vector_db_vendor (Optional[str], optional): name of vector database
            to use.

            connection_args (Optional[dict], optional): arguments to pass to
            vector database connection.

            embedding_collection_name (Optional[str], optional): name of
            collection to store embeddings in.

            metadata_collection_name (Optional[str], optional): name of
            collection to store metadata in.

            api_key (Optional[str], optional): Xinference API key.

            base_url (Optional[str], optional): base url of Xinference API.

            documentids_workspace (Optional[List[str]], optional): a list of document IDs
            that defines the scope within which rag operations (remove, similarity search,
            and get all) occur. Defaults to None, which means the operations will be
            performed across all documents in the database.

        """
        from xinference.client import Client

        self.model_name = model
        self.client = Client(base_url=base_url)
        self.models = {}
        self.load_models()

        if self.model_name is None or self.model_name == "auto":
            self.model_name = self.list_models_by_type("embedding")[0]
        self.model_uid = self.models[self.model_name]["id"]

        super().__init__(
            used=used,
            online=True,
            chunk_size=chunk_size,
            chunk_overlap=chunk_overlap,
            split_by_characters=split_by_characters,
            separators=separators,
            n_results=n_results,
            model=model,
            vector_db_vendor=vector_db_vendor,
            connection_args=connection_args,
            embedding_collection_name=embedding_collection_name,
            metadata_collection_name=metadata_collection_name,
            api_key=api_key,
            base_url=base_url,
            embeddings=XinferenceEmbeddings(
                server_url=base_url, model_uid=self.model_uid
            ),
            documentids_workspace=documentids_workspace,
        )

    def load_models(self) -> None:
        """
        Get all models that are currently available on the Xinference server and
        write them to `self.models`.
        """
        for id, model in self.client.list_models().items():
            model["id"] = id
            self.models[model["model_name"]] = model

    def list_models_by_type(self, type: str) -> list[str]:
        """
        Return all models of a certain type that are currently available on the
        Xinference server.

        Args:
            type (str): type of model to list (e.g. "embedding", "chat")

        Returns:
            List[str]: list of model names
        """
        names = []
        for name, model in self.models.items():
            if "model_ability" in model:
                if type in model["model_ability"]:
                    names.append(name)
            elif model["model_type"] == type:
                names.append(name)
        return names

__init__(used=False, chunk_size=1000, chunk_overlap=0, split_by_characters=True, separators=None, n_results=3, model='auto', vector_db_vendor=None, connection_args=None, embedding_collection_name=None, metadata_collection_name=None, api_key='none', base_url=None, documentids_workspace=None)

Extension of the DocumentEmbedder class that uses Xinference for embeddings.

Args:

used (bool, optional): whether RAG has been used (frontend setting).

chunk_size (int, optional): size of chunks to split text into.

chunk_overlap (int, optional): overlap between chunks.

split_by_characters (bool, optional): whether to split by characters
or tokens.

separators (Optional[list], optional): list of separators to use when
splitting by characters.

n_results (int, optional): number of results to return from
similarity search.

model (Optional[str], optional): name of model to use for embeddings.
Can be "auto" to use the first available model.

vector_db_vendor (Optional[str], optional): name of vector database
to use.

connection_args (Optional[dict], optional): arguments to pass to
vector database connection.

embedding_collection_name (Optional[str], optional): name of
collection to store embeddings in.

metadata_collection_name (Optional[str], optional): name of
collection to store metadata in.

api_key (Optional[str], optional): Xinference API key.

base_url (Optional[str], optional): base url of Xinference API.

documentids_workspace (Optional[List[str]], optional): a list of document IDs
that defines the scope within which rag operations (remove, similarity search,
and get all) occur. Defaults to None, which means the operations will be
performed across all documents in the database.
Source code in biochatter/vectorstore.py 
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
def __init__(
    self,
    used: bool = False,
    chunk_size: int = 1000,
    chunk_overlap: int = 0,
    split_by_characters: bool = True,
    separators: Optional[list] = None,
    n_results: int = 3,
    model: Optional[str] = "auto",
    vector_db_vendor: Optional[str] = None,
    connection_args: Optional[dict] = None,
    embedding_collection_name: Optional[str] = None,
    metadata_collection_name: Optional[str] = None,
    api_key: Optional[str] = "none",
    base_url: Optional[str] = None,
    documentids_workspace: Optional[list[str]] = None,
):
    """
    Extension of the DocumentEmbedder class that uses Xinference for
    embeddings.

    Args:

        used (bool, optional): whether RAG has been used (frontend setting).

        chunk_size (int, optional): size of chunks to split text into.

        chunk_overlap (int, optional): overlap between chunks.

        split_by_characters (bool, optional): whether to split by characters
        or tokens.

        separators (Optional[list], optional): list of separators to use when
        splitting by characters.

        n_results (int, optional): number of results to return from
        similarity search.

        model (Optional[str], optional): name of model to use for embeddings.
        Can be "auto" to use the first available model.

        vector_db_vendor (Optional[str], optional): name of vector database
        to use.

        connection_args (Optional[dict], optional): arguments to pass to
        vector database connection.

        embedding_collection_name (Optional[str], optional): name of
        collection to store embeddings in.

        metadata_collection_name (Optional[str], optional): name of
        collection to store metadata in.

        api_key (Optional[str], optional): Xinference API key.

        base_url (Optional[str], optional): base url of Xinference API.

        documentids_workspace (Optional[List[str]], optional): a list of document IDs
        that defines the scope within which rag operations (remove, similarity search,
        and get all) occur. Defaults to None, which means the operations will be
        performed across all documents in the database.

    """
    from xinference.client import Client

    self.model_name = model
    self.client = Client(base_url=base_url)
    self.models = {}
    self.load_models()

    if self.model_name is None or self.model_name == "auto":
        self.model_name = self.list_models_by_type("embedding")[0]
    self.model_uid = self.models[self.model_name]["id"]

    super().__init__(
        used=used,
        online=True,
        chunk_size=chunk_size,
        chunk_overlap=chunk_overlap,
        split_by_characters=split_by_characters,
        separators=separators,
        n_results=n_results,
        model=model,
        vector_db_vendor=vector_db_vendor,
        connection_args=connection_args,
        embedding_collection_name=embedding_collection_name,
        metadata_collection_name=metadata_collection_name,
        api_key=api_key,
        base_url=base_url,
        embeddings=XinferenceEmbeddings(
            server_url=base_url, model_uid=self.model_uid
        ),
        documentids_workspace=documentids_workspace,
    )

list_models_by_type(type)

Return all models of a certain type that are currently available on the Xinference server.

Parameters:

Name Type Description Default
type str

type of model to list (e.g. "embedding", "chat")

 required

Returns:

Type Description
list[str]

List[str]: list of model names
Source code in biochatter/vectorstore.py 
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
def list_models_by_type(self, type: str) -> list[str]:
    """
    Return all models of a certain type that are currently available on the
    Xinference server.

    Args:
        type (str): type of model to list (e.g. "embedding", "chat")

    Returns:
        List[str]: list of model names
    """
    names = []
    for name, model in self.models.items():
        if "model_ability" in model:
            if type in model["model_ability"]:
                names.append(name)
        elif model["model_type"] == type:
            names.append(name)
    return names

load_models()

Get all models that are currently available on the Xinference server and write them to self.models.

Source code in biochatter/vectorstore.py 
337
338
339
340
341
342
343
344
def load_models(self) -> None:
    """
    Get all models that are currently available on the Xinference server and
    write them to `self.models`.
    """
    for id, model in self.client.list_models().items():
        model["id"] = id
        self.models[model["model_name"]] = model

Vectorstore Agent

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 
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
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
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.
        """
        MAX_LENGTH = 10000
        doc_id = FieldSchema(
            name="id", dtype=DataType.INT64, is_primary=True, auto_id=True
        )
        doc_name = FieldSchema(
            name="name", dtype=DataType.VARCHAR, max_length=MAX_LENGTH
        )
        doc_author = FieldSchema(
            name="author", dtype=DataType.VARCHAR, max_length=MAX_LENGTH
        )
        doc_title = FieldSchema(
            name="title", dtype=DataType.VARCHAR, max_length=MAX_LENGTH
        )
        doc_format = FieldSchema(
            name="format", dtype=DataType.VARCHAR, max_length=255
        )
        doc_subject = FieldSchema(
            name="subject", dtype=DataType.VARCHAR, max_length=MAX_LENGTH
        )
        doc_creator = FieldSchema(
            name="creator", dtype=DataType.VARCHAR, max_length=MAX_LENGTH
        )
        doc_producer = FieldSchema(
            name="producer", dtype=DataType.VARCHAR, max_length=MAX_LENGTH
        )
        doc_creationDate = FieldSchema(
            name="creationDate", dtype=DataType.VARCHAR, max_length=1024
        )
        doc_modDate = FieldSchema(
            name="modDate", dtype=DataType.VARCHAR, max_length=1024
        )
        doc_source = FieldSchema(
            name="source", dtype=DataType.VARCHAR, max_length=MAX_LENGTH
        )
        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: Optional[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 (Optional[list[str]]): 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 (Optional[list[str]]): 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 obtaining 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

    def get_description(self, doc_ids: Optional[list[str]] = None):
        def get_name(meta: dict[str, str]):
            name_col = ["title", "name", "subject", "source"]
            for col in name_col:
                if meta[col] is not None and len(meta[col]) > 0:
                    return meta[col]
            return ""

        expr = VectorDatabaseAgentMilvus._build_meta_col_query_expr_for_all_documents(
            doc_ids
        )
        result = self._col_metadata.query(
            expr=expr,
            output_fields=METADATA_FIELDS,
        )
        names = list(map(get_name, result))
        names_set = set(names)
        desc = f"This vector store contains the following articles: {names_set}"
        return desc[:MAX_AGENT_DESC_LENGTH]

__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 
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
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 
163
164
165
166
167
168
169
170
171
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 which the operation of obtaining all documents occurs

 None

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 
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
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 obtaining 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 Optional[list[str]]

the list of document ids, defines documents scope within which remove operation occurs.

 None

Returns:

Name Type Description
bool bool

True if the document is deleted, False otherwise
Source code in biochatter/vectorstore_agent.py 
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
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 (Optional[list[str]]): 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 Optional[list[str]]

the list of document ids, do similarity search across the specified documents

 None

Returns:

Type Description
list[Document]

List[Document]: search results
Source code in biochatter/vectorstore_agent.py 
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
def similarity_search(
    self, query: str, k: int = 3, doc_ids: Optional[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 (Optional[list[str]]): 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 
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
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 
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
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 
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
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