Jaguar

JaguarVectorStore #

基类：BasePydanticVectorStore

Jaguar 向量存储。

参见 http://www.jaguardb.com 参见 http://github.com/fserv/jaguar-sdk

示例

pip install llama-index-vector-stores-jaguar

from llama_index.vector_stores.jaguar import JaguarVectorStore
vectorstore = JaguarVectorStore(
    pod = 'vdb',
    store = 'mystore',
    vector_index = 'v',
    vector_type = 'cosine_fraction_float',
    vector_dimension = 1536,
    url='http://192.168.8.88:8080/fwww/',
)

源码位于 llama-index-integrations/vector_stores/llama-index-vector-stores-jaguar/llama_index/vector_stores/jaguar/base.py

class JaguarVectorStore(BasePydanticVectorStore):
    """
    Jaguar vector store.

    See http://www.jaguardb.com
    See http://github.com/fserv/jaguar-sdk

    Examples:
        `pip install llama-index-vector-stores-jaguar`

        ```python
        from llama_index.vector_stores.jaguar import JaguarVectorStore
        vectorstore = JaguarVectorStore(
            pod = 'vdb',
            store = 'mystore',
            vector_index = 'v',
            vector_type = 'cosine_fraction_float',
            vector_dimension = 1536,
            url='http://192.168.8.88:8080/fwww/',
        )
        ```

    """

    stores_text: bool = True

    _pod: str = PrivateAttr()
    _store: str = PrivateAttr()
    _vector_index: str = PrivateAttr()
    _vector_type: str = PrivateAttr()
    _vector_dimension: int = PrivateAttr()
    _jag: JaguarHttpClient = PrivateAttr()
    _token: str = PrivateAttr()

    def __init__(
        self,
        pod: str,
        store: str,
        vector_index: str,
        vector_type: str,
        vector_dimension: int,
        url: str,
    ):
        """
        Constructor of JaguarVectorStore.

        Args:
            pod: str:  name of the pod (database)
            store: str:  name of vector store in the pod
            vector_index: str:  name of vector index of the store
            vector_type: str:  type of the vector index
            vector_dimension: int:  dimension of the vector index
            url: str:  URL end point of jaguar http server

        """
        super().__init__(stores_text=True)
        self._pod = self._sanitize_input(pod)
        self._store = self._sanitize_input(store)
        self._vector_index = self._sanitize_input(vector_index)
        self._vector_type = self._sanitize_input(vector_type)
        self._vector_dimension = vector_dimension
        self._jag = JaguarHttpClient(url)
        self._token = ""

    def __del__(self) -> None:
        pass

    @classmethod
    def class_name(cls) -> str:
        return "JaguarVectorStore"

    @property
    def client(self) -> Any:
        """Get client."""
        return self._jag

    def _sanitize_input(self, value: str) -> str:
        """Sanitize input to prevent SQL injection."""
        forbidden_chars = ['"', ";", "--", "/*", "*/"]
        sanitized = value.replace("'", "\\'")
        for char in forbidden_chars:
            sanitized = sanitized.replace(char, "")
        return sanitized

    def add(
        self,
        nodes: Sequence[BaseNode],
        **add_kwargs: Any,
    ) -> List[str]:
        """
        Add nodes to index.

        Args:
            nodes: List[BaseNode]: list of nodes with embeddings

        """
        use_node_metadata = add_kwargs.get("use_node_metadata", False)
        ids = []
        for node in nodes:
            text = node.get_text()
            embedding = node.get_embedding()
            if use_node_metadata is True:
                metadata = node.metadata
            else:
                metadata = None
            zid = self.add_text(text, embedding, metadata, **add_kwargs)
            ids.append(zid)

        return ids

    def delete(self, ref_doc_id: str, **delete_kwargs: Any) -> None:
        """
        Delete nodes using with ref_doc_id.

        Args:
            ref_doc_id (str): The doc_id of the document to delete.

        """
        podstore = self._pod + "." + self._store
        q = (
            "delete from "
            + podstore
            + " where zid='"
            + self._sanitize_input(ref_doc_id)
            + "'"
        )
        self.run(q)

    def query(self, query: VectorStoreQuery, **kwargs: Any) -> VectorStoreQueryResult:
        """
        Query index for top k most similar nodes.

        Args:
            query: VectorStoreQuery object
            kwargs:  may contain 'where', 'metadata_fields', 'args', 'fetch_k'

        """
        embedding = query.query_embedding
        k = query.similarity_top_k
        (nodes, ids, simscores) = self.similarity_search_with_score(
            embedding, k=k, form="node", **kwargs
        )
        return VectorStoreQueryResult(nodes=nodes, ids=ids, similarities=simscores)

    def load_documents(
        self, embedding: List[float], k: int, **kwargs: Any
    ) -> List[Document]:
        """
        Query index to load top k most similar documents.

        Args:
            embedding: a list of floats
            k: topK number
            kwargs:  may contain 'where', 'metadata_fields', 'args', 'fetch_k'

        """
        return cast(
            List[Document],
            self.similarity_search_with_score(embedding, k=k, form="doc", **kwargs),
        )

    def create(
        self,
        metadata_fields: str,
        text_size: int,
    ) -> None:
        """
        Create the vector store on the backend database.

        Args:
            metadata_fields (str):  exrta metadata columns and types
        Returns:
            True if successful; False if not successful

        """
        podstore = self._pod + "." + self._store

        """
        v:text column is required.
        """
        q = "create store "
        q += podstore
        q += f" ({self._vector_index} vector({self._vector_dimension},"
        q += f" '{self._vector_type}'),"
        q += f"  v:text char({text_size}),"
        q += self._sanitize_input(metadata_fields) + ")"
        self.run(q)

    def add_text(
        self,
        text: str,
        embedding: List[float],
        metadata: Optional[dict] = None,
        **kwargs: Any,
    ) -> str:
        """
        Add  texts through the embeddings and add to the vectorstore.

        Args:
          texts: text string to add to the jaguar vector store.
          embedding: embedding vector of the text, list of floats
          metadata: {'file_path': '../data/paul_graham/paul_graham_essay.txt',
                          'file_name': 'paul_graham_essay.txt',
                          'file_type': 'text/plain',
                          'file_size': 75042,
                          'creation_date': '2023-12-24',
                          'last_modified_date': '2023-12-24',
                          'last_accessed_date': '2023-12-28'}
          kwargs: vector_index=name_of_vector_index
                  file_column=name_of_file_column
                  metadata={...}

        Returns:
            id from adding the text into the vectorstore

        """
        text = self._sanitize_input(text)
        vcol = self._vector_index
        filecol = kwargs.get("file_column", "")
        text_tag = kwargs.get("text_tag", "")

        if text_tag != "":
            text = text_tag + " " + text

        podstorevcol = self._pod + "." + self._store + "." + vcol
        q = "textcol " + podstorevcol
        js = self.run(q)
        if js == "":
            return ""
        textcol = js["data"]

        zid = ""
        if metadata is None:
            ### no metadata and no files to upload
            str_vec = [str(x) for x in embedding]
            values_comma = self._sanitize_input(",".join(str_vec))
            podstore = self._pod + "." + self._store
            q = "insert into " + podstore + " ("
            q += vcol + "," + textcol + ") values ('" + values_comma
            q += "','" + text + "')"
            js = self.run(q, False)
            zid = js["zid"]
        else:
            str_vec = [str(x) for x in embedding]
            nvec, vvec, filepath = self._parseMeta(metadata, filecol)
            if filecol != "":
                rc = self._jag.postFile(self._token, filepath, 1)
                if not rc:
                    return ""
            names_comma = ",".join(nvec)
            names_comma += "," + vcol
            names_comma = self._sanitize_input(names_comma)
            ## col1,col2,col3,vecl

            if vvec is not None and len(vvec) > 0:
                values_comma = "'" + "','".join(vvec) + "'"
            else:
                values_comma = "'" + "','".join(vvec) + "'"

            ### 'va1','val2','val3'
            values_comma += ",'" + ",".join(str_vec) + "'"
            values_comma = self._sanitize_input(values_comma)
            ### 'v1,v2,v3'
            podstore = self._pod + "." + self._store
            q = "insert into " + podstore + " ("
            q += names_comma + "," + textcol + ") values (" + values_comma
            q += ",'" + text + "')"
            if filecol != "":
                js = self.run(q, True)
            else:
                js = self.run(q, False)
            zid = js["zid"]

        return zid

    def similarity_search_with_score(
        self,
        embedding: Optional[List[float]],
        k: int = 3,
        form: str = "node",
        **kwargs: Any,
    ) -> Union[Tuple[List[TextNode], List[str], List[float]], List[Document]]:
        """
        Return nodes most similar to query embedding, along with ids and scores.

        Args:
            embedding: embedding of text to look up.
            k: Number of nodes to return. Defaults to 3.
            form: if "node", return Tuple[List[TextNode], List[str], List[float]]
                  if "doc", return List[Document]
            kwargs: may have where, metadata_fields, args, fetch_k
        Returns:
            Tuple(list of nodes, list of ids, list of similaity scores)

        """
        where = kwargs.get("where")
        metadata_fields = kwargs.get("metadata_fields")

        args = kwargs.get("args")
        fetch_k = kwargs.get("fetch_k", -1)

        vcol = self._vector_index
        vtype = self._vector_type
        if embedding is None:
            return ([], [], [])
        str_embeddings = [str(f) for f in embedding]
        qv_comma = self._sanitize_input(",".join(str_embeddings))
        podstore = self._pod + "." + self._store
        q = (
            "select similarity("
            + vcol
            + ",'"
            + qv_comma
            + "','topk="
            + str(k)
            + ",fetch_k="
            + str(fetch_k)
            + ",type="
            + vtype
        )
        q += ",with_score=yes,with_text=yes"
        if args is not None:
            q += "," + args

        if metadata_fields is not None:
            x = "&".join(metadata_fields)
            q += ",metadata=" + x

        q += "') from " + podstore

        if where is not None:
            q += " where " + self._sanitize_input(where)

        jarr = self.run(q)

        if jarr is None:
            return ([], [], [])

        nodes = []
        ids = []
        simscores = []
        docs = []
        for js in jarr:
            score = js["score"]
            text = js["text"]
            zid = js["zid"]

            md = {}
            md["zid"] = zid
            if metadata_fields is not None:
                for m in metadata_fields:
                    mv = js[m]
                    md[m] = mv

            if form == "node":
                node = TextNode(
                    id_=zid,
                    text=text,
                    metadata=md,
                )
                nodes.append(node)
                ids.append(zid)
                simscores.append(float(score))
            else:
                doc = Document(
                    id_=zid,
                    text=text,
                    metadata=md,
                )
                docs.append(doc)

        if form == "node":
            return (nodes, ids, simscores)
        else:
            return docs

    def is_anomalous(
        self,
        node: BaseNode,
        **kwargs: Any,
    ) -> bool:
        """
        Detect if given text is anomalous from the dataset.

        Args:
            query: Text to detect if it is anomaly
        Returns:
            True or False

        """
        vcol = self._vector_index
        vtype = self._vector_type
        str_embeddings = [str(f) for f in node.get_embedding()]
        qv_comma = ",".join(str_embeddings)
        podstore = self._pod + "." + self._store
        q = "select anomalous(" + vcol + ", '" + qv_comma + "', 'type=" + vtype + "')"
        q += " from " + podstore

        js = self.run(q)
        if isinstance(js, list) and len(js) == 0:
            return False
        jd = json.loads(js[0])
        return jd["anomalous"] == "YES"


    def run(self, query: str, withFile: bool = False) -> dict:
        """
        Run any query statement in jaguardb.

        Args:
            query (str): query statement to jaguardb
        Returns:
            None for invalid token, or
            json result string

        """
        if self._token == "":
            logger.error(f"E0005 error run({query})")
            return {}

        resp = self._jag.post(query, self._token, withFile)
        txt = resp.text
        try:
            return json.loads(txt)
        except Exception:
            return {}

    def count(self) -> int:
        """
        Count records of a store in jaguardb.

        Args: no args
        Returns: (int) number of records in pod store
        """
        podstore = self._pod + "." + self._store
        q = "select count() from " + podstore
        js = self.run(q)
        if isinstance(js, list) and len(js) == 0:
            return 0
        jd = json.loads(js[0])
        return int(jd["data"])

    def clear(self) -> None:
        """
        Delete all records in jaguardb.

        Args: No args
        Returns: None
        """
        podstore = self._pod + "." + self._store
        q = "truncate store " + podstore
        self.run(q)

    def drop(self) -> None:
        """
        Drop or remove a store in jaguardb.

        Args: no args
        Returns: None
        """
        podstore = self._pod + "." + self._store
        q = "drop store " + podstore
        self.run(q)

    def prt(self, msg: str) -> None:
        nows = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
        with open("/tmp/debugjaguar.log", "a") as file:
            print(f"{nows} msg={msg}", file=file, flush=True)

    def login(
        self,
        jaguar_api_key: Optional[str] = "",
    ) -> bool:
        """
        Login to jaguar server with a jaguar_api_key or let self._jag find a key.

        Args:
            optional jaguar_api_key (str): API key of user to jaguardb server
        Returns:
            True if successful; False if not successful

        """
        if jaguar_api_key == "":
            jaguar_api_key = self._jag.getApiKey()
        self._jaguar_api_key = jaguar_api_key
        self._token = self._jag.login(jaguar_api_key)
        if self._token == "":
            logger.error("E0001 error init(): invalid jaguar_api_key")
            return False
        return True

    def logout(self) -> None:
        """
        Logout to cleanup resources.

        Args: no args
        Returns: None
        """
        self._jag.logout(self._token)

    def _parseMeta(self, nvmap: dict, filecol: str) -> Tuple[List[str], List[str], str]:
        filepath = ""
        if filecol == "":
            nvec = list(nvmap.keys())
            vvec = list(nvmap.values())
        else:
            nvec = []
            vvec = []
            if filecol in nvmap:
                nvec.append(filecol)
                vvec.append(nvmap[filecol])
                filepath = nvmap[filecol]

            for k, v in nvmap.items():
                if k != filecol:
                    nvec.append(k)
                    vvec.append(v)

        return nvec, vvec, filepath

client `property` #

client: Any

获取客户端。

add #

add(nodes: Sequence[BaseNode], **add_kwargs: Any) -> List[str]

将节点添加到索引。

参数

名称	类型	描述	默认值
`nodes`	`Sequence[BaseNode]`	List[BaseNode]：带有嵌入的节点列表	必需

源码位于 llama-index-integrations/vector_stores/llama-index-vector-stores-jaguar/llama_index/vector_stores/jaguar/base.py

def add(
    self,
    nodes: Sequence[BaseNode],
    **add_kwargs: Any,
) -> List[str]:
    """
    Add nodes to index.

    Args:
        nodes: List[BaseNode]: list of nodes with embeddings

    """
    use_node_metadata = add_kwargs.get("use_node_metadata", False)
    ids = []
    for node in nodes:
        text = node.get_text()
        embedding = node.get_embedding()
        if use_node_metadata is True:
            metadata = node.metadata
        else:
            metadata = None
        zid = self.add_text(text, embedding, metadata, **add_kwargs)
        ids.append(zid)

    return ids

delete #

delete(ref_doc_id: str, **delete_kwargs: Any) -> None

使用 ref_doc_id 删除节点。

参数

名称	类型	描述	默认值
`ref_doc_id`	`str`	要删除的文档的 doc_id。	必需

源码位于 llama-index-integrations/vector_stores/llama-index-vector-stores-jaguar/llama_index/vector_stores/jaguar/base.py

def delete(self, ref_doc_id: str, **delete_kwargs: Any) -> None:
    """
    Delete nodes using with ref_doc_id.

    Args:
        ref_doc_id (str): The doc_id of the document to delete.

    """
    podstore = self._pod + "." + self._store
    q = (
        "delete from "
        + podstore
        + " where zid='"
        + self._sanitize_input(ref_doc_id)
        + "'"
    )
    self.run(q)

query #

query(query: VectorStoreQuery, **kwargs: Any) -> VectorStoreQueryResult

查询索引以获取前 k 个最相似的节点。

参数

名称	类型	描述	默认值
`query`	`VectorStoreQuery`	VectorStoreQuery 对象	必需
`kwargs`	`Any`	可能包含 'where', 'metadata_fields', 'args', 'fetch_k'	`{}`

源码位于 llama-index-integrations/vector_stores/llama-index-vector-stores-jaguar/llama_index/vector_stores/jaguar/base.py

def query(self, query: VectorStoreQuery, **kwargs: Any) -> VectorStoreQueryResult:
    """
    Query index for top k most similar nodes.

    Args:
        query: VectorStoreQuery object
        kwargs:  may contain 'where', 'metadata_fields', 'args', 'fetch_k'

    """
    embedding = query.query_embedding
    k = query.similarity_top_k
    (nodes, ids, simscores) = self.similarity_search_with_score(
        embedding, k=k, form="node", **kwargs
    )
    return VectorStoreQueryResult(nodes=nodes, ids=ids, similarities=simscores)

load_documents #

load_documents(embedding: List[float], k: int, **kwargs: Any) -> List[Document]

查询索引以加载前 k 个最相似的文档。

参数

名称	类型	描述	默认值
`embedding`	`List[float]`	浮点数列表	必需
`k`	`int`	topK 数量	必需
`kwargs`	`Any`	可能包含 'where', 'metadata_fields', 'args', 'fetch_k'	`{}`

源码位于 llama-index-integrations/vector_stores/llama-index-vector-stores-jaguar/llama_index/vector_stores/jaguar/base.py

def load_documents(
    self, embedding: List[float], k: int, **kwargs: Any
) -> List[Document]:
    """
    Query index to load top k most similar documents.

    Args:
        embedding: a list of floats
        k: topK number
        kwargs:  may contain 'where', 'metadata_fields', 'args', 'fetch_k'

    """
    return cast(
        List[Document],
        self.similarity_search_with_score(embedding, k=k, form="doc", **kwargs),
    )

create #

create(metadata_fields: str, text_size: int) -> None

在后端数据库上创建向量存储。

参数

名称	类型	描述	默认值
`metadata_fields`	`str`	额外的元数据列和类型	必需

返回值：成功返回 True；失败返回 False

源码位于 llama-index-integrations/vector_stores/llama-index-vector-stores-jaguar/llama_index/vector_stores/jaguar/base.py

def create(
    self,
    metadata_fields: str,
    text_size: int,
) -> None:
    """
    Create the vector store on the backend database.

    Args:
        metadata_fields (str):  exrta metadata columns and types
    Returns:
        True if successful; False if not successful

    """
    podstore = self._pod + "." + self._store

    """
    v:text column is required.
    """
    q = "create store "
    q += podstore
    q += f" ({self._vector_index} vector({self._vector_dimension},"
    q += f" '{self._vector_type}'),"
    q += f"  v:text char({text_size}),"
    q += self._sanitize_input(metadata_fields) + ")"
    self.run(q)

add_text #

add_text(text: str, embedding: List[float], metadata: Optional[dict] = None, **kwargs: Any) -> str

通过嵌入添加文本并添加到向量存储。

参数

名称	类型	描述	默认值
`texts`		要添加到 jaguar 向量存储的文本字符串。	必需
`embedding`	`List[float]`	文本的嵌入向量，浮点数列表	必需
`metadata`	`Optional[dict]`	{'file_path': '../data/paul_graham/paul_graham_essay.txt', 'file_name': 'paul_graham_essay.txt', 'file_type': 'text/plain', 'file_size': 75042, 'creation_date': '2023-12-24', 'last_modified_date': '2023-12-24', 'last_accessed_date': '2023-12-28'}	`无`
`kwargs`	`Any`	vector_index=name_of_vector_index file_column=name_of_file_column metadata={...}	`{}`

返回值

类型	描述
`str`	将文本添加到向量存储后返回的 id

源码位于 llama-index-integrations/vector_stores/llama-index-vector-stores-jaguar/llama_index/vector_stores/jaguar/base.py

def add_text(
    self,
    text: str,
    embedding: List[float],
    metadata: Optional[dict] = None,
    **kwargs: Any,
) -> str:
    """
    Add  texts through the embeddings and add to the vectorstore.

    Args:
      texts: text string to add to the jaguar vector store.
      embedding: embedding vector of the text, list of floats
      metadata: {'file_path': '../data/paul_graham/paul_graham_essay.txt',
                      'file_name': 'paul_graham_essay.txt',
                      'file_type': 'text/plain',
                      'file_size': 75042,
                      'creation_date': '2023-12-24',
                      'last_modified_date': '2023-12-24',
                      'last_accessed_date': '2023-12-28'}
      kwargs: vector_index=name_of_vector_index
              file_column=name_of_file_column
              metadata={...}

    Returns:
        id from adding the text into the vectorstore

    """
    text = self._sanitize_input(text)
    vcol = self._vector_index
    filecol = kwargs.get("file_column", "")
    text_tag = kwargs.get("text_tag", "")

    if text_tag != "":
        text = text_tag + " " + text

    podstorevcol = self._pod + "." + self._store + "." + vcol
    q = "textcol " + podstorevcol
    js = self.run(q)
    if js == "":
        return ""
    textcol = js["data"]

    zid = ""
    if metadata is None:
        ### no metadata and no files to upload
        str_vec = [str(x) for x in embedding]
        values_comma = self._sanitize_input(",".join(str_vec))
        podstore = self._pod + "." + self._store
        q = "insert into " + podstore + " ("
        q += vcol + "," + textcol + ") values ('" + values_comma
        q += "','" + text + "')"
        js = self.run(q, False)
        zid = js["zid"]
    else:
        str_vec = [str(x) for x in embedding]
        nvec, vvec, filepath = self._parseMeta(metadata, filecol)
        if filecol != "":
            rc = self._jag.postFile(self._token, filepath, 1)
            if not rc:
                return ""
        names_comma = ",".join(nvec)
        names_comma += "," + vcol
        names_comma = self._sanitize_input(names_comma)
        ## col1,col2,col3,vecl

        if vvec is not None and len(vvec) > 0:
            values_comma = "'" + "','".join(vvec) + "'"
        else:
            values_comma = "'" + "','".join(vvec) + "'"

        ### 'va1','val2','val3'
        values_comma += ",'" + ",".join(str_vec) + "'"
        values_comma = self._sanitize_input(values_comma)
        ### 'v1,v2,v3'
        podstore = self._pod + "." + self._store
        q = "insert into " + podstore + " ("
        q += names_comma + "," + textcol + ") values (" + values_comma
        q += ",'" + text + "')"
        if filecol != "":
            js = self.run(q, True)
        else:
            js = self.run(q, False)
        zid = js["zid"]

    return zid

similarity_search_with_score #

similarity_search_with_score(embedding: Optional[List[float]], k: int = 3, form: str = 'node', **kwargs: Any) -> Union[Tuple[List[TextNode], List[str], List[float]], List[Document]]

返回与查询嵌入最相似的节点，以及它们的 id 和分数。

参数

名称	类型	描述	默认值
`embedding`	`Optional[List[float]]`	要查找文本的嵌入。	必需
`k`	`int`	要返回的节点数量。默认为 3。	`3`
`form`	`str`	如果是 "node"，返回 Tuple[List[TextNode], List[str], List[float]]；如果是 "doc"，返回 List[Document]	`'node'`
`kwargs`	`Any`	可能包含 where, metadata_fields, args, fetch_k	`{}`

返回值：Tuple (节点列表, id 列表, 相似度分数列表)

源码位于 llama-index-integrations/vector_stores/llama-index-vector-stores-jaguar/llama_index/vector_stores/jaguar/base.py

def similarity_search_with_score(
    self,
    embedding: Optional[List[float]],
    k: int = 3,
    form: str = "node",
    **kwargs: Any,
) -> Union[Tuple[List[TextNode], List[str], List[float]], List[Document]]:
    """
    Return nodes most similar to query embedding, along with ids and scores.

    Args:
        embedding: embedding of text to look up.
        k: Number of nodes to return. Defaults to 3.
        form: if "node", return Tuple[List[TextNode], List[str], List[float]]
              if "doc", return List[Document]
        kwargs: may have where, metadata_fields, args, fetch_k
    Returns:
        Tuple(list of nodes, list of ids, list of similaity scores)

    """
    where = kwargs.get("where")
    metadata_fields = kwargs.get("metadata_fields")

    args = kwargs.get("args")
    fetch_k = kwargs.get("fetch_k", -1)

    vcol = self._vector_index
    vtype = self._vector_type
    if embedding is None:
        return ([], [], [])
    str_embeddings = [str(f) for f in embedding]
    qv_comma = self._sanitize_input(",".join(str_embeddings))
    podstore = self._pod + "." + self._store
    q = (
        "select similarity("
        + vcol
        + ",'"
        + qv_comma
        + "','topk="
        + str(k)
        + ",fetch_k="
        + str(fetch_k)
        + ",type="
        + vtype
    )
    q += ",with_score=yes,with_text=yes"
    if args is not None:
        q += "," + args

    if metadata_fields is not None:
        x = "&".join(metadata_fields)
        q += ",metadata=" + x

    q += "') from " + podstore

    if where is not None:
        q += " where " + self._sanitize_input(where)

    jarr = self.run(q)

    if jarr is None:
        return ([], [], [])

    nodes = []
    ids = []
    simscores = []
    docs = []
    for js in jarr:
        score = js["score"]
        text = js["text"]
        zid = js["zid"]

        md = {}
        md["zid"] = zid
        if metadata_fields is not None:
            for m in metadata_fields:
                mv = js[m]
                md[m] = mv

        if form == "node":
            node = TextNode(
                id_=zid,
                text=text,
                metadata=md,
            )
            nodes.append(node)
            ids.append(zid)
            simscores.append(float(score))
        else:
            doc = Document(
                id_=zid,
                text=text,
                metadata=md,
            )
            docs.append(doc)

    if form == "node":
        return (nodes, ids, simscores)
    else:
        return docs

is_anomalous #

is_anomalous(node: BaseNode, **kwargs: Any) -> bool

检测给定文本是否是数据集中的异常值。

参数

名称	类型	描述	默认值
`query`		要检测是否为异常的文本	必需

返回值：True 或 False

源码位于 llama-index-integrations/vector_stores/llama-index-vector-stores-jaguar/llama_index/vector_stores/jaguar/base.py

def is_anomalous(
    self,
    node: BaseNode,
    **kwargs: Any,
) -> bool:
    """
    Detect if given text is anomalous from the dataset.

    Args:
        query: Text to detect if it is anomaly
    Returns:
        True or False

    """
    vcol = self._vector_index
    vtype = self._vector_type
    str_embeddings = [str(f) for f in node.get_embedding()]
    qv_comma = ",".join(str_embeddings)
    podstore = self._pod + "." + self._store
    q = "select anomalous(" + vcol + ", '" + qv_comma + "', 'type=" + vtype + "')"
    q += " from " + podstore

    js = self.run(q)
    if isinstance(js, list) and len(js) == 0:
        return False
    jd = json.loads(js[0])
    return jd["anomalous"] == "YES"

run #

run(query: str, withFile: bool = False) -> dict

在 jaguardb 中运行任何查询语句。

参数

名称	类型	描述	默认值
`query`	`str`	发送给 jaguardb 的查询语句	必需

返回值：无效 token 返回 None，否则返回 json 结果字符串

源码位于 llama-index-integrations/vector_stores/llama-index-vector-stores-jaguar/llama_index/vector_stores/jaguar/base.py

def run(self, query: str, withFile: bool = False) -> dict:
    """
    Run any query statement in jaguardb.

    Args:
        query (str): query statement to jaguardb
    Returns:
        None for invalid token, or
        json result string

    """
    if self._token == "":
        logger.error(f"E0005 error run({query})")
        return {}

    resp = self._jag.post(query, self._token, withFile)
    txt = resp.text
    try:
        return json.loads(txt)
    except Exception:
        return {}

count #

count() -> int

计算 jaguardb 中某个存储的记录数。

参数：无参数返回值：(int) pod 存储中的记录数

源码位于 llama-index-integrations/vector_stores/llama-index-vector-stores-jaguar/llama_index/vector_stores/jaguar/base.py

def count(self) -> int:
    """
    Count records of a store in jaguardb.

    Args: no args
    Returns: (int) number of records in pod store
    """
    podstore = self._pod + "." + self._store
    q = "select count() from " + podstore
    js = self.run(q)
    if isinstance(js, list) and len(js) == 0:
        return 0
    jd = json.loads(js[0])
    return int(jd["data"])

clear #

clear() -> None

删除 jaguardb 中的所有记录。

参数：无参数返回值：None

源码位于 llama-index-integrations/vector_stores/llama-index-vector-stores-jaguar/llama_index/vector_stores/jaguar/base.py

def clear(self) -> None:
    """
    Delete all records in jaguardb.

    Args: No args
    Returns: None
    """
    podstore = self._pod + "." + self._store
    q = "truncate store " + podstore
    self.run(q)

drop #

drop() -> None

在 jaguardb 中删除或移除某个存储。

参数：无参数返回值：None

源码位于 llama-index-integrations/vector_stores/llama-index-vector-stores-jaguar/llama_index/vector_stores/jaguar/base.py

def drop(self) -> None:
    """
    Drop or remove a store in jaguardb.

    Args: no args
    Returns: None
    """
    podstore = self._pod + "." + self._store
    q = "drop store " + podstore
    self.run(q)

login(jaguar_api_key: Optional[str] = '') -> bool

使用 jaguar_api_key 登录 jaguar 服务器，或让 self._jag 查找密钥。

参数

名称	类型	描述	默认值
`可选`	`jaguar_api_key (str`	用户访问 jaguardb 服务器的 API 密钥	必需

返回值：成功返回 True；失败返回 False

源码位于 llama-index-integrations/vector_stores/llama-index-vector-stores-jaguar/llama_index/vector_stores/jaguar/base.py

def login(
    self,
    jaguar_api_key: Optional[str] = "",
) -> bool:
    """
    Login to jaguar server with a jaguar_api_key or let self._jag find a key.

    Args:
        optional jaguar_api_key (str): API key of user to jaguardb server
    Returns:
        True if successful; False if not successful

    """
    if jaguar_api_key == "":
        jaguar_api_key = self._jag.getApiKey()
    self._jaguar_api_key = jaguar_api_key
    self._token = self._jag.login(jaguar_api_key)
    if self._token == "":
        logger.error("E0001 error init(): invalid jaguar_api_key")
        return False
    return True

logout #

logout() -> None

登出以清理资源。

参数：无参数返回值：None

源码位于 llama-index-integrations/vector_stores/llama-index-vector-stores-jaguar/llama_index/vector_stores/jaguar/base.py

def logout(self) -> None:
    """
    Logout to cleanup resources.

    Args: no args
    Returns: None
    """
    self._jag.logout(self._token)

Jaguar

JaguarVectorStore #

client property #

add #

delete #

query #

load_documents #

create #

add_text #

similarity_search_with_score #

is_anomalous #

run #

count #

clear #

drop #

login #

logout #

client `property` #