Skip to main content

Filtres de recherche pour la base de connaissances

L'API Retrieve renvoie des segments de texte issus d'une base de connaissances en se basant sur la similarité sémantique. Lorsque la recherche sémantique retourne trop de résultats non pertinents, utilisez SearchFilters pour appliquer un filtrage basé sur les métadonnées aux résultats sémantiques. Cette approche est efficace pour les données structurées, telles que les tableaux de données disposant de champs bien définis.

Fonctionnement de SearchFilters

Sans SearchFilters, l'API Retrieve renvoie tous les segments de texte sémantiquement similaires, ce qui peut inclure des correspondances non pertinentes. SearchFilters effectue d'abord une recherche sémantique, puis filtre les résultats selon des conditions de champ. Sans SearchFilters :
{
  "indexId": "o73yjlxxxx",
  "query": "Employees in the company named San Zhang"
}
La récupération sémantique renvoie des segments de texte peu pertinents par rapport à « San Zhang ».
image
Avec SearchFilters :
{
  "indexId": "o73yjlxxxx",
  "query": "Employees in the company named San Zhang",
  "searchFilters": [
    {
      "Name": "San Zhang"
    }
  ]
}
SearchFilters supprime des résultats sémantiques les segments de texte qui ne correspondent pas à Name: "San Zhang".
image

Syntaxe

SearchFilters est un tableau de sous-groupes. Chaque sous-groupe est un objet JSON contenant une ou plusieurs paires field: value. Les sous-groupes sont toujours combinés avec une logique AND.
{
  "searchFilters": [
    {
      "Name": "San Zhang",
      "Sex": "Male"
    },
    {
      "Position": "Engineer"
    }
  ]
}
Dans cet exemple, les résultats doivent correspondre à Name = "San Zhang" ET Sex = "Male" (au sein du sous-groupe 1) ET Position = "Engineer" (sous-groupe 2).

Types de requête pris en charge

Référence des opérateurs

Type de requêteOpérateurTypes de champ pris en chargeDescription
Valeur unique*(valeur directe)*chaîne, numérique (long, double)Correspond à une seule valeur exacte. Exemple : {"Name": "San Zhang"}
Valeurs multiples*(tableau JSON)*tableau de chaînes, tableau numériqueCorrespond à n'importe quelle valeur du tableau (similaire à SQL IN). Exemple : {"Name": "[\"San Zhang\",\"Si Li\"]"}
Égalitéeqchaîne, numériqueÉgal à. Exemple : {"Name": {"eq": "San Zhang"}}
Égaliténeqchaîne, numériqueDifférent de. Exemple : {"Name": {"neq": "San Zhang"}}
IntervallegtnumériqueSupérieur à. Exemple : {"Age": {"gt": 20}}
IntervallegtenumériqueSupérieur ou égal à. Exemple : {"Age": {"gte": 20}}
IntervalleltnumériqueInférieur à. Exemple : {"Age": {"lt": 30}}
IntervalleltenumériqueInférieur ou égal à. Exemple : {"Age": {"lte": 30}}
FlouelikechaîneCorrespondance de motif avec caractères génériques. Exemple : {"Position": {"like": "E%r"}}
Tagtagstableau de chaînesFiltrer par tags de document (bases de connaissances de recherche de documents uniquement). Exemple : {"tags": "[\"tag1\",\"tag2\"]"}
Les requêtes d'égalité et d'intervalle sont des sous-types de requêtes de plage. Les requêtes d'égalité ne sont pas sensibles à la casse et n'acceptent qu'une seule valeur par champ.

Caractères génériques pour les requêtes floues

CaractèreDescriptionExemple
%Correspond à zéro ou plusieurs caractèresE%r correspond à Engineer
_Correspond à exactement un caractèreE_gineer correspond à Engineer

Logique des requêtes de tag

  • Plusieurs tags au sein d'un même sous-groupe utilisent une logique OR : les fichiers correspondant à n'importe quel tag sont renvoyés.
  • Pour appliquer une logique AND entre les tags, placez chaque tag dans un sous-groupe distinct.

Prérequis

Vous avez besoin des éléments suivants :
Les utilisateurs RAM peuvent gérer les bases de connaissances uniquement dans les espaces de travail auxquels ils ont adhéré. Les comptes Alibaba Cloud peuvent gérer tous les espaces de travail.

Configuration des exemples de données

Les exemples utilisent le fichier EmployeeInformation.xlsx (3 enregistrements). Créez une base de connaissances avec les paramètres suivants :
ParamètreValeur
Knowledge base typeData Query
Data sourceUpload Data Table
FieldsName (string), Sex (string), Position (string), Age (double)
Index settingsTous les champs utilisés pour la récupération et les réponses du modèle
image

Exemples de code complets

Les exemples Python et Java suivants couvrent tous les types de requête.
Avant d'exécuter l'exemple de code, configurez votre AccessKey et votre AccessKey secret en tant que variables d'environnement .
  • Python
  • Java
# The sample code is for reference only. Do not use it in a production environment.
import json
import os
import sys
from typing import List

from alibabacloud_bailian20231229 import models as bailian_20231229_models
from alibabacloud_bailian20231229.client import Client as bailian20231229Client
from alibabacloud_tea_openapi import models as open_api_models
from alibabacloud_tea_util.client import Client as UtilClient

class SearchFiltersFullExample:
    class QueryObject:
        def __init__(self, prefix):
            self.like = prefix

        def to_dict(self):
            return {
                "like": self.like
            }

    class Range:
        def __init__(self, gte, lte):
            self.gte = gte
            self.lte = lte

        def to_dict(self):
            return {
                "gte": self.gte,
                "lte": self.lte
            }

    def __init__(self):
        pass

    @staticmethod
    def create_client() -> bailian20231229Client:
        """
        Initializes a client using an AccessKey pair.
        @return: Client
        @throws Exception
        """
        # For production, use STS or other secure credential management. See https://www.alibabacloud.com/help/sdk/developer-reference/v2-manage-access-credentials
        config = open_api_models.Config(
            # Required. Make sure that the ALIBABA_CLOUD_ACCESS_KEY_ID environment variable is set.
            access_key_id=os.environ['ALIBABA_CLOUD_ACCESS_KEY_ID'],
            # Required. Make sure that the ALIBABA_CLOUD_ACCESS_KEY_SECRET environment variable is set.
            access_key_secret=os.environ['ALIBABA_CLOUD_ACCESS_KEY_SECRET']
        )
        # For more information about endpoints, see https://www.alibabacloud.com/help/model-studio/api-bailian-2023-12-29-endpoint
        config.endpoint = f'bailian.ap-southeast-1.aliyuncs.com'
        return bailian20231229Client(config)

    @staticmethod
    def sub_group_query() -> None:
        try:
            client = SearchFiltersFullExample.create_client()
            # Create a retrieve_request object.
            retrieve_request = bailian_20231229_models.RetrieveRequest()
            # Required. You can enter the actual prompt.
            retrieve_request.query = 'Employees in the company named San Zhang'
            # Required. Enter the actual knowledge base ID.
            retrieve_request.index_id = 'Enter the actual knowledge base ID'
            # Group 1: Name = "San Zhang", Group 2: Sex = "Female"
            retrieve_request.search_filters = [
                {"Name": "San Zhang"},
                {"Sex": "Female"}
            ]
            # Perform a retrieval. Enter the workspace ID and the retrieve_request object.
            resp = client.retrieve('Enter the actual workspace ID', retrieve_request)
            print(UtilClient.to_jsonstring(resp))
        except Exception as error:
            # Handle exceptions appropriately in production
            print(error.message)

    @staticmethod
    def single_query() -> None:
        try:
            client = SearchFiltersFullExample.create_client()
            # Create a retrieve_request object.
            retrieve_request = bailian_20231229_models.RetrieveRequest()
            # Required. You can enter the actual prompt.
            retrieve_request.query = 'Employees in the company named San Zhang'
            # Required. Enter the actual knowledge base ID.
            retrieve_request.index_id = 'Enter the actual knowledge base ID'
            # Single-value filter: Name = "San Zhang"
            retrieve_request.search_filters = [
                {"Name": "San Zhang"}
            ]
            # Perform a retrieval. Enter the workspace ID and the retrieve_request object.
            resp = client.retrieve('Enter the actual workspace ID', retrieve_request)
            print(UtilClient.to_jsonstring(resp.body))
        except Exception as error:
            # Handle exceptions appropriately in production
            print(error.message)

    @staticmethod
    def multi_query() -> None:
        try:
            client = SearchFiltersFullExample.create_client()
            # Create a retrieve_request object.
            retrieve_request = bailian_20231229_models.RetrieveRequest()
            # Required. You can enter the actual prompt.
            retrieve_request.query = 'All employees in the company'
            # Required. Enter the actual knowledge base ID.
            retrieve_request.index_id = 'Enter the actual knowledge base ID'
            # Create a list to store multiple values.
            names = ["San Zhang", "Si Li"]
            # Multi-value filter: Name IN ["San Zhang", "Si Li"]
            retrieve_request.search_filters = [{"Name": json.dumps(names)}]
            # Perform a retrieval. Enter the workspace ID and the retrieve_request object.
            resp = client.retrieve('Enter the actual workspace ID', retrieve_request)
            print(UtilClient.to_jsonstring(resp.body))
        except Exception as error:
            # Handle exceptions appropriately in production
            print(error.message)

    @staticmethod
    def range_query() -> None:
        try:
            client = SearchFiltersFullExample.create_client()
            # Create a retrieve_request object.
            retrieve_request = bailian_20231229_models.RetrieveRequest()
            # Required. You can enter the actual prompt.
            retrieve_request.query = 'All employees in the company'
            # Required. Enter the actual knowledge base ID.
            retrieve_request.index_id = 'Enter the actual knowledge base ID'
            # Range filter: Age
            age_range = SearchFiltersFullExample.Range(20, 27)
            retrieve_request.search_filters = [
                {"Position": "Engineer"},  # Group 1: Position filter condition
                {"Age": json.dumps(age_range.to_dict())}  # Group 2: Age range filter condition
            ]
            # Perform a retrieval. Enter the workspace ID and the retrieve_request object.
            resp = client.retrieve('Enter the actual workspace ID', retrieve_request)
            print(UtilClient.to_jsonstring(resp.body))
        except Exception as error:
            # Handle exceptions appropriately in production
            print(error.message)

    @staticmethod
    def wildcard_query() -> None:
        try:
            client = SearchFiltersFullExample.create_client()
            # Create a retrieve_request object.
            retrieve_request = bailian_20231229_models.RetrieveRequest()
            # Required. You can enter the actual prompt.
            retrieve_request.query = 'Male employees in the company'
            # Required. Enter the actual knowledge base ID.
            retrieve_request.index_id = 'Enter the actual knowledge base ID'
            # Fuzzy filter: Position LIKE "E%r" (% = zero or more chars)
            position = SearchFiltersFullExample.QueryObject('E%r')
            retrieve_request.search_filters = [
                {"Name": "San Zhang"},  # Name filter condition. You can replace it with the actual property and value.
                {"Position": json.dumps(position.to_dict())}  # Position (fuzzy query) filter condition. You can replace it with the actual property and value.
            ]
            # Perform a retrieval. Enter the workspace ID and the retrieve_request object.
            resp = client.retrieve('Enter the actual workspace ID', retrieve_request)
            print(UtilClient.to_jsonstring(resp.body))
        except Exception as error:
            # Handle exceptions appropriately in production
            print(error.message)

    @staticmethod
    def tag_query() -> None:
        try:
            client = SearchFiltersFullExample.create_client()
            # Create a retrieve_request object.
            retrieve_request = bailian_20231229_models.RetrieveRequest()
            # Required. You can enter the actual prompt.
            retrieve_request.query = 'Provide some candidates'
            # Required. Enter the actual knowledge base ID.
            retrieve_request.index_id = 'Enter the actual knowledge base ID'
            # Tag filter (OR logic within array)
            tags = ["University A", "Student Union President"]
            retrieve_request.search_filters = [
                {"tags": json.dumps(tags)}
            ]
            # Perform a retrieval. Enter the workspace ID and the retrieve_request object.
            resp = client.retrieve('Enter the actual workspace ID', retrieve_request)
            print(UtilClient.to_jsonstring(resp.body))
        except Exception as error:
            # Handle exceptions appropriately in production
            print(error.message)

    @staticmethod
    def tag_query2() -> None:
        try:
            client = SearchFiltersFullExample.create_client()
            # Create a retrieve_request object.
            retrieve_request = bailian_20231229_models.RetrieveRequest()
            # Required. You can enter the actual prompt.
            retrieve_request.query = 'Provide some candidates'
            # Required. Enter the actual knowledge base ID.
            retrieve_request.index_id = 'Enter the actual knowledge base ID'
            # Tag filters (AND logic between groups)
            tag1 = ["University A"]
            tag2 = ["Sports Specialist Student"]
            retrieve_request.search_filters = [
                {"tags": json.dumps(tag1)},
                {"tags": json.dumps(tag2)}
            ]
            # Perform a retrieval. Enter the workspace ID and the retrieve_request object.
            resp = client.retrieve('Enter the actual workspace ID', retrieve_request)
            print(UtilClient.to_jsonstring(resp.body))
        except Exception as error:
            # Handle exceptions appropriately in production
            print(error.message)

    @staticmethod
    def main(args: List[str]) -> None:
        SearchFiltersFullExample.sub_group_query()
        SearchFiltersFullExample.single_query()
        SearchFiltersFullExample.multi_query()
        SearchFiltersFullExample.range_query()
        SearchFiltersFullExample.wildcard_query()
        SearchFiltersFullExample.tag_query()
        SearchFiltersFullExample.tag_query2()

if __name__ == '__main__':
    SearchFiltersFullExample.main(sys.argv[1:])

Exemples par type de requête

Chaque type de requête inclut la syntaxe JSON ainsi qu'un exemple de requête et de réponse. Remplacez ces espaces réservés par vos valeurs réelles :
Espace réservéDescriptionExemple
<knowledge-base-id>ID d'index de la base de connaissances27ubwxxxxx
<workspace-id>ID de l'espace de travailllm-4u5xpd1xdjxxxxxx

Requête par sous-groupe

Chaque élément du tableau searchFilters constitue un sous-groupe combinant des conditions de filtrage. Tous les sous-groupes doivent correspondre (logique AND). Filtrer les enregistrements où Name est « San Zhang » ET Sex est « Female » :
{
  "searchFilters": [
    {
      "Name": "San Zhang"
    },
    {
      "Sex": "Female"
    }
  ]
}
Aucun enregistrement ne correspond aux deux conditions, donc nodes est vide.
  • Python
  • Java
# The sample code is for reference only. Do not use it in a production environment.
import os
import sys
from typing import List

from alibabacloud_bailian20231229 import models as bailian_20231229_models
from alibabacloud_bailian20231229.client import Client as bailian20231229Client
from alibabacloud_tea_openapi import models as open_api_models
from alibabacloud_tea_util.client import Client as UtilClient

class SubGroupQueryExample:
    def __init__(self):
        pass

    @staticmethod
    def create_client() -> bailian20231229Client:
        """
        Initializes a client using an AccessKey pair.
        @return: Client
        @throws Exception
        """
        # For production, use STS or other secure credential management. See https://www.alibabacloud.com/help/sdk/developer-reference/v2-manage-access-credentials
        config = open_api_models.Config(
            # Required. Make sure that the ALIBABA_CLOUD_ACCESS_KEY_ID environment variable is set.
            access_key_id=os.environ['ALIBABA_CLOUD_ACCESS_KEY_ID'],
            # Required. Make sure that the ALIBABA_CLOUD_ACCESS_KEY_SECRET environment variable is set.
            access_key_secret=os.environ['ALIBABA_CLOUD_ACCESS_KEY_SECRET']
        )
        # For more information about endpoints, see https://www.alibabacloud.com/help/model-studio/api-bailian-2023-12-29-endpoint
        config.endpoint = f'bailian.ap-southeast-1.aliyuncs.com'
        return bailian20231229Client(config)

    @staticmethod
    def main(
            args: List[str],
    ) -> None:
        try:
            client = SubGroupQueryExample.create_client()
            # Create a retrieve_request object.
            retrieve_request = bailian_20231229_models.RetrieveRequest()
            # Required. You can enter the actual prompt.
            retrieve_request.query = 'Employees in the company named San Zhang'
            # Required. Enter the actual knowledge base ID.
            retrieve_request.index_id = 'Enter the actual knowledge base ID'
            # Group 1: Name = "San Zhang", Group 2: Sex = "Female"
            retrieve_request.search_filters = [
                {"Name": "San Zhang"},
                {"Sex": "Female"}
            ]
            # Perform a retrieval. Enter the workspace ID and the retrieve_request object.
            resp = client.retrieve('Enter the actual workspace ID', retrieve_request)
            print(UtilClient.to_jsonstring(resp))
        except Exception as error:
            # Handle exceptions appropriately in production
            print(error.message)

if __name__ == '__main__':
    SubGroupQueryExample.main(sys.argv[1:])
Paramètres de requête
{
  "indexId": "27ubwxxxxx",
  "WorkspaceId":"llm-4u5xpd1xdjxxxxxx",
  "query": "Employees in the company named San Zhang",
  "searchFilters": [
    {
      "Name": "San Zhang"
    },
    {
      "Sex": "Female"
    }
  ]
}
Paramètres de réponse
{
  "code": "Success",
  "data": {
    "nodes": []
  },
  "message": "success",
  "requestId": "5BA30772-xxxx-560C-B1F7-C1DA737A9D80",
  "status": "200",
  "success": true
}

Requête à valeur unique

Transmettez une valeur unique pour un champ afin de faire correspondre les enregistrements ayant exactement cette valeur. Filtrer les enregistrements où Name est « San Zhang » :
{
  "searchFilters": [
    {
      "Name": "San Zhang"
    }
  ]
}
  • Python
  • Java
# The sample code is for reference only. Do not use it in a production environment.
import os
import sys
from typing import List

from alibabacloud_bailian20231229 import models as bailian_20231229_models
from alibabacloud_bailian20231229.client import Client as bailian20231229Client
from alibabacloud_tea_openapi import models as open_api_models
from alibabacloud_tea_util.client import Client as UtilClient

class SingleQueryExample:
    def __init__(self):
        pass

    @staticmethod
    def create_client() -> bailian20231229Client:
        """
        Initializes a client using an AccessKey pair.
        @return: Client
        @throws Exception
        """
        # For production, use STS or other secure credential management. See https://www.alibabacloud.com/help/sdk/developer-reference/v2-manage-access-credentials
        config = open_api_models.Config(
            # Required. Make sure that the ALIBABA_CLOUD_ACCESS_KEY_ID environment variable is set.
            access_key_id=os.environ['ALIBABA_CLOUD_ACCESS_KEY_ID'],
            # Required. Make sure that the ALIBABA_CLOUD_ACCESS_KEY_SECRET environment variable is set.
            access_key_secret=os.environ['ALIBABA_CLOUD_ACCESS_KEY_SECRET']
        )
        # For more information about endpoints, see https://www.alibabacloud.com/help/model-studio/api-bailian-2023-12-29-endpoint
        config.endpoint = f'bailian.ap-southeast-1.aliyuncs.com'
        return bailian20231229Client(config)

    @staticmethod
    def main(
            args: List[str],
    ) -> None:
        client = SingleQueryExample.create_client()
        # Create a retrieve_request object.
        retrieve_request = bailian_20231229_models.RetrieveRequest()
        # Required. You can enter the actual prompt.
        retrieve_request.query = 'Employees in the company named San Zhang'
        # Required. Enter the actual knowledge base ID.
        retrieve_request.index_id = 'Enter the actual knowledge base ID'
        # The property for the single-value query is Name and the value is San Zhang. You can replace them with the actual property and value.
        retrieve_request.search_filters = [
            {"Name": "San Zhang"}
        ]
        try:
            # Perform a retrieval. Enter the workspace ID and the retrieve_request object.
            resp = client.retrieve('Enter the actual workspace ID', retrieve_request)
            print(UtilClient.to_jsonstring(resp.body))
        except Exception as error:
            # Handle exceptions appropriately in production
            print(error.message)

if __name__ == '__main__':
    SingleQueryExample.main(sys.argv[1:])
Paramètres de requête
{
  "indexId": "27ubwxxxxx",
  "WorkspaceId":"llm-4u5xpd1xdjxxxxxx",
  "query": "Employees in the company named San Zhang",
  "searchFilters": [
    {
      "Name": "San Zhang"
    }
  ]
}
Paramètres de réponse
{
  "code": "Success",
  "data": {
    "nodes": [
      {
        "metadata": {
          "_rc_v_score": 0.32581159472465515,
          "_q_score": 1,
          "source": "0",
          "_score": 0.4556944966316223,
          "doc_id": "table_xxxx75507aab4bd9a24c18d098b2e8ac_10285263_1",
          "Sex": "Male",
          "_rc_score": 0,
          "Name": "San Zhang",
          "doc_name": "Employee Sheet",
          "_id": "llm-xxxxpd1xdjqp8itj_27ubwxxxxx_table_200275507aab4bd9a24c18d098b2e8ac_10285263_1",
          "Age": "25",
          "Position": "Engineer"
        },
        "score": 0.4556944966316223,
        "text": "Name:San Zhang Age:25 Position:Engineer Sex:Male"
      }
    ]
  },
  "message": "success",
  "requestId": "2FA4113E-xxxx-59C1-BDB2-5B930D8C9B1C",
  "status": "200",
  "success": true
}

Requête à valeurs multiples

Transmettez plusieurs valeurs pour un champ afin de faire correspondre les enregistrements contenant l'une des valeurs spécifiées, similairement à l'opérateur SQL IN. Sérialisez le tableau de valeurs sous forme de chaîne JSON. Filtrer les enregistrements où Name est « San Zhang » ou « Si Li » :
{
  "searchFilters": [
    {
      "Name": "[\"San Zhang\",\"Si Li\"]"
    }
  ]
}
  • Python
  • Java
# The sample code is for reference only. Do not use it in a production environment.
import json
import os
import sys
from typing import List

from alibabacloud_bailian20231229 import models as bailian_20231229_models
from alibabacloud_bailian20231229.client import Client as bailian20231229Client
from alibabacloud_tea_openapi import models as open_api_models
from alibabacloud_tea_util.client import Client as UtilClient

class MultiQueryExample:
    def __init__(self):
        pass

    @staticmethod
    def create_client() -> bailian20231229Client:
        """
        Initializes a client using an AccessKey pair.
        @return: Client
        @throws Exception
        """
        # For production, use STS or other secure credential management. See https://www.alibabacloud.com/help/sdk/developer-reference/v2-manage-access-credentials
        config = open_api_models.Config(
            # Required. Make sure that the ALIBABA_CLOUD_ACCESS_KEY_ID environment variable is set.
            access_key_id=os.environ['ALIBABA_CLOUD_ACCESS_KEY_ID'],
            # Required. Make sure that the ALIBABA_CLOUD_ACCESS_KEY_SECRET environment variable is set.
            access_key_secret=os.environ['ALIBABA_CLOUD_ACCESS_KEY_SECRET']
        )
        # For more information about endpoints, see https://www.alibabacloud.com/help/model-studio/api-bailian-2023-12-29-endpoint
        config.endpoint = f'bailian.ap-southeast-1.aliyuncs.com'
        return bailian20231229Client(config)

    @staticmethod
    def main(
            args: List[str],
    ) -> None:
        try:
            client = MultiQueryExample.create_client()
            # Create a retrieve_request object.
            retrieve_request = bailian_20231229_models.RetrieveRequest()
            # Required. You can enter the actual prompt.
            retrieve_request.query = 'All employees in the company'
            # Required. Enter the actual knowledge base ID.
            retrieve_request.index_id = 'Enter the actual knowledge base ID'
            # Create a list to store multiple values.
            names = ["San Zhang", "Si Li"]
            # Multi-value filter: Name IN ["San Zhang", "Si Li"]
            retrieve_request.search_filters = [{"Name": json.dumps(names)}]
            # Perform a retrieval. Enter the workspace ID and the retrieve_request object.
            resp = client.retrieve('Enter the actual workspace ID', retrieve_request)
            print(UtilClient.to_jsonstring(resp.body))
        except Exception as error:
            # Handle exceptions appropriately in production
            print(error.message)

if __name__ == '__main__':
    MultiQueryExample.main(sys.argv[1:])
Paramètres de requête
{
  "indexId": "27ubwxxxxx",
  "WorkspaceId":"llm-4u5xpd1xdjxxxxxx",
  "query": "All employees in the company",
  "searchFilters": [
    {
      "Name": "[\"San Zhang\",\"Si Li\"]"
    }
  ]
}
Paramètres de réponse
{
  "code": "Success",
  "data": {
    "nodes": [
      {
        "metadata": {
          "_rc_v_score": 0.3016361088048254,
          "_q_score": 1,
          "source": "0",
          "_score": 0.3322954773902893,
          "doc_id": "table_xxxxad1c331c424780a8023376a73fac_10285263_1",
          "Sex": "Male",
          "_rc_score": 0,
          "Name": "San Zhang",
          "doc_name": "Employee Sheet",
          "_id": "llm-xxxxpd1xdjqp8itj_27ubwxxxxx_table_xxxxad1c331c424780a8023376a73fac_10285263_1",
          "Age": "25.0",
          "Position": "Engineer"
        },
        "score": 0.3322954773902893,
        "text": "Name:San Zhang Age:25.0 Position:Engineer Sex:Male"
      },
      {
        "metadata": {
          "_rc_v_score": 0.2531493306159973,
          "_q_score": 0.8392540654998252,
          "source": "0",
          "_score": 0.25632044672966003,
          "doc_id": "table_xxxxad1c331c424780a8023376a73fac_10285263_2",
          "Sex": "Female",
          "_rc_score": 0,
          "Name": "Si Li",
          "doc_name": "Employee Sheet",
          "_id": "llm-xxxxpd1xdjqp8itj_27ubwxxxxx_table_xxxxad1c331c424780a8023376a73fac_10285263_2",
          "Age": "31.0",
          "Position": "Sales"
        },
        "score": 0.25632044672966003,
        "text": "Name:Si Li Age:31.0 Position:Sales Sex:Female"
      }
    ]
  },
  "message": "success",
  "requestId": "1DFE5E9E-xxxx-5C37-8011-8FA2E2875309",
  "status": "200",
  "success": true
}

Requête de plage

Utilisez les requêtes de plage pour trouver les enregistrements dont un champ numérique se situe dans un intervalle spécifié. Combinez les requêtes de plage avec d'autres types de requête en les plaçant dans des sous-groupes distincts. Filtrer les enregistrements où Position est « Engineer » (valeur unique) ET Age est compris entre 20 et 25 (plage) :
{
  "searchFilters": [
    {
      "Position": "Engineer"
    },
    {
      "Age": {
        "gte": 20,
        "lte": 25
      }
    }
  ]
}
  • Python
  • Java
# The sample code is for reference only. Do not use it in a production environment.
import json
import os
import sys
from typing import List

from alibabacloud_bailian20231229 import models as bailian_20231229_models
from alibabacloud_bailian20231229.client import Client as bailian20231229Client
from alibabacloud_tea_openapi import models as open_api_models
from alibabacloud_tea_util.client import Client as UtilClient

class RangeQueryExample:
    class Range:
        def __init__(self, gte, lte):
            self.gte = gte
            self.lte = lte

        def to_dict(self):
            return {
                "gte": self.gte,
                "lte": self.lte
            }

    def __init__(self):
        pass

    @staticmethod
    def create_client() -> bailian20231229Client:
        """
        Initializes a client using an AccessKey pair.
        @return: Client
        @throws Exception
        """
        # For production, use STS or other secure credential management. See https://www.alibabacloud.com/help/sdk/developer-reference/v2-manage-access-credentials
        config = open_api_models.Config(
            # Required. Make sure that the ALIBABA_CLOUD_ACCESS_KEY_ID environment variable is set.
            access_key_id=os.environ['ALIBABA_CLOUD_ACCESS_KEY_ID'],
            # Required. Make sure that the ALIBABA_CLOUD_ACCESS_KEY_SECRET environment variable is set.
            access_key_secret=os.environ['ALIBABA_CLOUD_ACCESS_KEY_SECRET']
        )
        # For more information about endpoints, see https://www.alibabacloud.com/help/model-studio/api-bailian-2023-12-29-endpoint
        config.endpoint = f'bailian.ap-southeast-1.aliyuncs.com'
        return bailian20231229Client(config)

    @staticmethod
    def main(
            args: List[str],
    ) -> None:
        try:
            client = RangeQueryExample.create_client()
            # Create a retrieve_request object.
            retrieve_request = bailian_20231229_models.RetrieveRequest()
            # Required. You can enter the actual prompt.
            retrieve_request.query = 'All employees in the company'
            # Required. Enter the actual knowledge base ID.
            retrieve_request.index_id = 'Enter the actual knowledge base ID'
            # Range filter: Age
            age_range = RangeQueryExample.Range(20, 27)
            retrieve_request.search_filters = [
                {"Position": "Engineer"},  # Group 1: Position filter condition
                {"Age": json.dumps(age_range.to_dict())}  # Group 2: Age range filter condition
            ]
            # Perform a retrieval. Enter the workspace ID and the retrieve_request object.
            resp = client.retrieve('Enter the actual workspace ID', retrieve_request)
            print(UtilClient.to_jsonstring(resp.body))
        except Exception as error:
            # Handle exceptions appropriately in production
            print(error.message)

if __name__ == '__main__':
    RangeQueryExample.main(sys.argv[1:])
Paramètres de requête
{
  "indexId": "27ubwxxxxx",
  "WorkspaceId":"llm-4u5xpd1xdjxxxxxx",
  "query": "All employees in the company",
  "searchFilters": [
    {
      "Sex": "Male"
    },
    {
      "Age": "{\"gte\":20.0,\"lte\":27.0}"
    }
  ]
}
Paramètres de réponse
{
  "code": "Success",
  "data": {
    "nodes": [
      {
        "metadata": {
          "_rc_v_score": 0.3016361088048254,
          "_q_score": 1,
          "source": "0",
          "_score": 0.3322954773902893,
          "doc_id": "table_xxxxad1c331c424780a8023376a73fac_10285263_1",
          "Sex": "Male",
          "_rc_score": 0,
          "Name": "San Zhang",
          "doc_name": "Employee Sheet",
          "_id": "llm-xxxxpd1xdjqp8itj_27ubwxxxxx_table_xxxxad1c331c424780a8023376a73fac_10285263_1",
          "Age": "25.0",
          "Position": "Engineer"
        },
        "score": 0.3322954773902893,
        "text": "Name:San Zhang Age:25.0 Position:Engineer Sex:Male"
      }
    ]
  },
  "message": "success",
  "requestId": "AE0B5ABC-xxxx-54A1-9ED4-91865B859DF6",
  "status": "200",
  "success": true
}

Requête floue

Utilisez des motifs avec caractères génériques pour faire correspondre partiellement les champs de chaîne, similairement à l'opérateur SQL LIKE. Enveloppez le motif dans un objet {"like": "<pattern>"} et sérialisez-le sous forme de chaîne JSON. Filtrer les enregistrements où Name est « San Zhang » ET Position commence par « E » et se termine par « r » :
{
  "searchFilters": [
    {
      "Name": "San Zhang"
    },
    {
      "Position": {
        "like": "E%r"
      }
    }
  ]
}
  • Python
  • Java
# The sample code is for reference only. Do not use it in a production environment.
import json
import os
import sys
from typing import List

from alibabacloud_bailian20231229 import models as bailian_20231229_models
from alibabacloud_bailian20231229.client import Client as bailian20231229Client
from alibabacloud_tea_openapi import models as open_api_models
from alibabacloud_tea_util.client import Client as UtilClient

class WildcardQueryExample:
    class QueryObject:
        def __init__(self, prefix):
            self.like = prefix

        def to_dict(self):
            return {
                "like": self.like
            }

    def __init__(self):
        pass

    @staticmethod
    def create_client() -> bailian20231229Client:
        """
        Initializes a client using an AccessKey pair.
        @return: Client
        @throws Exception
        """
        # For production, use STS or other secure credential management. See https://www.alibabacloud.com/help/sdk/developer-reference/v2-manage-access-credentials
        config = open_api_models.Config(
            # Required. Make sure that the ALIBABA_CLOUD_ACCESS_KEY_ID environment variable is set.
            access_key_id=os.environ['ALIBABA_CLOUD_ACCESS_KEY_ID'],
            # Required. Make sure that the ALIBABA_CLOUD_ACCESS_KEY_SECRET environment variable is set.
            access_key_secret=os.environ['ALIBABA_CLOUD_ACCESS_KEY_SECRET']
        )
        # For more information about endpoints, see https://www.alibabacloud.com/help/model-studio/api-bailian-2023-12-29-endpoint
        config.endpoint = f'bailian.ap-southeast-1.aliyuncs.com'
        return bailian20231229Client(config)

    @staticmethod
    def main(
            args: List[str],
    ) -> None:
        try:
            client = WildcardQueryExample.create_client()
            # Create a retrieve_request object.
            retrieve_request = bailian_20231229_models.RetrieveRequest()
            # Required. You can enter the actual prompt.
            retrieve_request.query = 'Male employees in the company'
            # Required. Enter the actual knowledge base ID.
            retrieve_request.index_id = 'Enter the actual knowledge base ID'
            # Fuzzy filter: Position LIKE "E%r" (% = zero or more chars)
            position = WildcardQueryExample.QueryObject('E%r')
            retrieve_request.search_filters = [
                {"Name": "San Zhang"},  # Name filter condition. You can replace it with the actual property and value.
                {"Position": json.dumps(position.to_dict())}  # Position (fuzzy query) filter condition. You can replace it with the actual property and value.
            ]
            # Perform a retrieval. Enter the workspace ID and the retrieve_request object.
            resp = client.retrieve('Enter the actual workspace ID', retrieve_request)
            print(UtilClient.to_jsonstring(resp.body))
        except Exception as error:
            # Handle exceptions appropriately in production
            print(error.message)

if __name__ == '__main__':
    WildcardQueryExample.main(sys.argv[1:])
Paramètres de requête
{
  "indexId": "27ubwxxxxx",
  "WorkspaceId":"llm-4u5xpd1xdjxxxxxx",
  "query": "Male employees in the company",
  "searchFilters": [
    {
      "Name": "San Zhang"
    },
    {
      "Position": "{\"like\":\"E%r\"}"
    }
  ]
}
Paramètres de réponse
{
  "code": "Success",
  "data": {
    "nodes": [
      {
        "metadata": {
          "_rc_v_score": 0.41137335930387275,
          "_q_score": 1,
          "source": "0",
          "_score": 0.46098726987838745,
          "doc_id": "table_xxxxad1c331c424780a8023376a73fac_10285263_1",
          "Sex": "Male",
          "_rc_score": 0,
          "Name": "San Zhang",
          "doc_name": "Employee Sheet",
          "_id": "llm-xxxxpd1xdjqp8itj_27ubwxxxxx_table_xxxxad1c331c424780a8023376a73fac_10285263_1",
          "Age": "25.0",
          "Position": "Engineer"
        },
        "score": 0.46098726987838745,
        "text": "Name:San Zhang Age:25.0 Position:Engineer Sex:Male"
      }
    ]
  },
  "message": "success",
  "requestId": "FA759FEC-xxxx-50B7-A64D-BE49A7DF56B8",
  "status": "200",
  "success": true
}

Requête de tag

Filtrez par tags de document dans une base de connaissances de recherche de documents . Les tags restreignent la récupération aux segments de texte provenant de fichiers correspondant à des valeurs de tag spécifiques. Configuration de l'exemple : Une base de connaissances de recherche de documents contient trois fichiers de CV avec les tags suivants :
FichierTags
CV de San ZhangUniversity A, Sports Specialist Student
CV de Si LiUniversity B
CV de Wu WangUniversity B, Student Union President

Exemple 1 : Logique OR (plusieurs tags dans un seul sous-groupe)

Renvoyer les segments de texte des fichiers tagués avec University A OU Student Union President :
Plusieurs tags au sein d'un même sous-groupe utilisent une logique OR. Pour appliquer une logique AND, placez chaque tag dans un sous-groupe distinct (voir l'exemple 2).
{
  "searchFilters": [
    {
      "tags": "[\"University A\",\"Student Union President\"]"
    }
  ]
}
# The sample code is for reference only. Do not use it in a production environment.
import json
import os
import sys
from typing import List

from alibabacloud_bailian20231229 import models as bailian_20231229_models
from alibabacloud_bailian20231229.client import Client as bailian20231229Client
from alibabacloud_tea_openapi import models as open_api_models
from alibabacloud_tea_util.client import Client as UtilClient

class TagQueryExample:
    def __init__(self):
        pass

    @staticmethod
    def create_client() -> bailian20231229Client:
        """
        Initializes a client using an AccessKey pair.
        @return: Client
        @throws Exception
        """
        # Leaking code that contains an AccessKey pair threatens the security of all resources in your account. The following sample code is for reference only.
        # We recommend that you use a more secure method, such as Security Token Service (STS), to manage access credentials. For more information, see https://www.alibabacloud.com/help/en/sdk/developer-reference/v2-manage-python-access-credentials.
        config = open_api_models.Config(
            # Required. Make sure that the ALIBABA_CLOUD_ACCESS_KEY_ID environment variable is set.
            access_key_id=os.environ['ALIBABA_CLOUD_ACCESS_KEY_ID'],
            # Required. Make sure that the ALIBABA_CLOUD_ACCESS_KEY_SECRET environment variable is set.
            access_key_secret=os.environ['ALIBABA_CLOUD_ACCESS_KEY_SECRET']
        )
        # For more information about endpoints, see https://www.alibabacloud.com/help/model-studio/api-bailian-2023-12-29-endpoint
        config.endpoint = f'bailian.ap-southeast-1.aliyuncs.com'
        return bailian20231229Client(config)

    @staticmethod
    def main(
            args: List[str],
    ) -> None:
        client = TagQueryExample.create_client()
        # Create a retrieve_request object.
        retrieve_request = bailian_20231229_models.RetrieveRequest()
        # Required. You can enter the actual prompt.
        retrieve_request.query = 'Provide some candidates'
        # Required. Enter the actual knowledge base ID.
        retrieve_request.index_id = 'Enter the actual knowledge base ID'
        # Create a list to store tags. The logical relationship between multiple tags is OR, not AND.
        tags = ["University A", "Student Union President"]
        retrieve_request.search_filters = [
            {"tags": json.dumps(tags)}
        ]
        try:
            # Perform a retrieval. Enter the workspace ID and the retrieve_request object.
            resp = client.retrieve('Enter the actual workspace ID', retrieve_request)
            print(UtilClient.to_jsonstring(resp.body))
        except Exception as error:
            # Handle exceptions appropriately in production
            print(error.message)

if __name__ == '__main__':
    TagQueryExample.main(sys.argv[1:])
// The sample code is for reference only. Do not use it in a production environment.
import com.aliyun.bailian20231229.models.RetrieveRequest;
import com.aliyun.bailian20231229.models.RetrieveResponse;
import com.google.gson.Gson;
import com.google.gson.JsonArray;

import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import java.util.Map;

public class TagQueryExample {
    /**
     * description
Paramètres de requête
{
  "IndexId": "8mbtdxxxxx",
  "WorkspaceId": "llm-4u5xpd1xdjxxxxxx",
  "Query": "Provide some candidates",
  "SearchFilters": "[{\"tags\":\"[\\\"University A\\\", \\\"Student Union President\\\"]\"}]"
}
Paramètres de réponse
{
  "Code": "Success",
  "Data": {
    "Nodes": [
      {
        "Metadata": {
          "file_path": "https://bailian-datahub-data-prod.oss-cn-beijing.aliyuncs.com/10285263/multimodal/docJson/Zhang_San_Resume_1746760910599.json?Expires=1747020348&OSSAccessKeyId=YOUR_ACCESS_KEY_ID&Signature=YOUR_SIGNATURE",
          "is_displayed_chunk_content": "true",
          "_rc_v_score": 0.1422617520249937,
          "image_url": [],
          "nid": "ba8a14099f43308734538c29271cc7cd|b99e98835c3c6d8f6496df5a43de0ba5|aa3ed8fc4aae8bbb78872994b01e0fda",
          "_q_score": 0.8935035934804278,
          "source": "0",
          "_score": 0.1736905574798584,
          "title": "",
          "doc_id": "file_e787926158704f95aad6bc967619f176_10285263",
          "content": "Name: San Zhang Sex: Male Age: 23",
          "_rc_score": 0,
          "workspace_id": "llm-4u5xpd1xdjxxxxxx",
          "hier_title": "",
          "page_number": [
            0
          ],
          "doc_name": "San Zhang's Resume",
          "pipeline_id": "8mbtdxxxxx",
          "_id": "llm-4u5xpd1xdjxxxxxx_8mbtdxxxxx_file_e787926158704f95aad6bc967619f176_10285263_0_0"
        },
        "Score": 0.1736905574798584,
        "Text": "Name: San Zhang Sex: Male Age: 23"
      },
      {
        "Metadata": {
          "file_path": "https://bailian-datahub-data-prod.oss-cn-beijing.aliyuncs.com/10285263/multimodal/docJson/Wang_Wu_Resume_1746760946844.json?Expires=1747020348&OSSAccessKeyId=YOUR_ACCESS_KEY_ID&Signature=YOUR_SIGNATURE",
          "is_displayed_chunk_content": "true",
          "_rc_v_score": 0.1592178845871759,
          "image_url": [],
          "nid": "ba8a14099f43308734538c29271cc7cd|d4a048b6799ce07e08430f018af091a0|fe90f8248ea64f70ea37c0df7afcfc12",
          "_q_score": 1,
          "source": "0",
          "_score": 0.15737050771713257,
          "title": "",
          "doc_id": "file_63563df5df66488cb8e28bfff11e40eb_10285263",
          "content": "Name: Wu Wang Sex: Male Age: 23",
          "_rc_score": 0,
          "workspace_id": "llm-4u5xpd1xdjxxxxxx",
          "hier_title": "",
          "page_number": [
            0
          ],
          "doc_name": "Wu Wang's Resume",
          "pipeline_id": "8mbtdxxxxx",
          "_id": "llm-4u5xpd1xdjxxxxxx_8mbtdxxxxx_file_63563df5df66488cb8e28bfff11e40eb_10285263_0_0"
        },
        "Score": 0.15737050771713257,
        "Text": "Name: Wu Wang Sex: Male Age: 23"
      }
    ]
  },
  "Message": "success",
  "RequestId": "12A5F9C6-xxxx-5593-8955-86D52585EE27",
  "Status": 200,
  "Success": true
}
Cela renvoie des résultats issus à la fois du CV de San Zhang (tagué University A) et du CV de Wu Wang (tagué Student Union President).

Exemple 2 : Logique AND (tags dans des sous-groupes distincts)

Renvoyer les segments de texte des fichiers tagués à la fois avec University A ET Sports Specialist Student :
{
  "searchFilters": [
    {
      "tags": "[\"University A\"]"
    },
    {
      "tags": "[\"Sports Specialist Student\"]"
    }
  ]
}
import json
import os
import sys
from typing import List

from alibabacloud_bailian20231229 import models as bailian_20231229_models
from alibabacloud_bailian20231229.client import Client as bailian20231229Client
from alibabacloud_tea_openapi import models as open_api_models
from alibabacloud_tea_util.client import Client as UtilClient

class TagQueryExample2:
    def __init__(self):
        pass

    @staticmethod
    def create_client() -> bailian20231229Client:
        """
        Initializes a client using an AccessKey pair.
        @return: Client
        @throws Exception
        """
        # Leaking code that contains an AccessKey pair threatens the security of all resources in your account. The following sample code is for reference only.
        # We recommend that you use a more secure method, such as Security Token Service (STS), to manage access credentials. For more information, see https://www.alibabacloud.com/help/en/sdk/developer-reference/v2-manage-python-access-credentials.
        config = open_api_models.Config(
            # Required. Make sure that the ALIBABA_CLOUD_ACCESS_KEY_ID environment variable is set.
            access_key_id=os.environ['ALIBABA_CLOUD_ACCESS_KEY_ID'],
            # Required. Make sure that the ALIBABA_CLOUD_ACCESS_KEY_SECRET environment variable is set.
            access_key_secret=os.environ['ALIBABA_CLOUD_ACCESS_KEY_SECRET']
        )
        # For more information about endpoints, see https://www.alibabacloud.com/help/model-studio/api-bailian-2023-12-29-endpoint
        config.endpoint = f'bailian.ap-southeast-1.aliyuncs.com'
        return bailian20231229Client(config)

    @staticmethod
    def main(
            args: List[str],
    ) -> None:
        client = TagQueryExample.create_client()
        # Create a retrieve_request object.
        retrieve_request = bailian_20231229_models.RetrieveRequest()
        # Required. You can enter the actual prompt.
        retrieve_request.query = 'Provide some candidates'
        # Required. Enter the actual knowledge base ID.
        retrieve_request.index_id = 'Enter the actual knowledge base ID'
        # Create two lists to store tags.
        tag1 = ["University A"]
        tag2 = ["Sports Specialist Student"]
        retrieve_request.search_filters = [
            {"tags": json.dumps(tag1)},
            {"tags": json.dumps(tag2)}
        ]
        try:
            # Perform a retrieval. Enter the workspace ID and the retrieve_request object.
            resp = client.retrieve('Enter the actual workspace ID', retrieve_request)
            print(UtilClient.to_jsonstring(resp.body))
        except Exception as error:
            # Handle exceptions appropriately in production
            print(error.message)

if __name__ == '__main__':
    TagQueryExample2.main(sys.argv[1:])
import com.aliyun.bailian20231229.models.RetrieveRequest;
import com.aliyun.bailian20231229.models.RetrieveResponse;
import com.google.gson.Gson;
import com.google.gson.JsonArray;

import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import java.util.Map;

public class TagQueryExample2 {
    /**
     * description
Paramètres de requête
{
  "IndexId": "8mbtdxxxxx",
  "WorkspaceId": "llm-4u5xpd1xdjxxxxxx",
  "Query": "Provide some candidates",
  "SearchFilters": "[{\"tags\":\"[\\\"University A\\\"]\"},{\"tags\":\"[\\\"Sports Specialist Student\\\"]\"}]"
}
Paramètres de réponse
{
  "Code": "Success",
  "Data": {
    "Nodes": [
      {
        "Metadata": {
          "file_path": "https://bailian-datahub-data-prod.oss-cn-beijing.aliyuncs.com/10285263/multimodal/docJson/Zhang_San_Resume_1746760910599.json?Expires=1747020348&OSSAccessKeyId=YOUR_ACCESS_KEY_ID&Signature=YOUR_SIGNATURE",
          "is_displayed_chunk_content": "true",
          "_rc_v_score": 0.1422617520249937,
          "image_url": [],
          "nid": "ba8a14099f43308734538c29271cc7cd|b99e98835c3c6d8f6496df5a43de0ba5|aa3ed8fc4aae8bbb78872994b01e0fda",
          "_q_score": 1,
          "source": "0",
          "_score": 0.1736905574798584,
          "title": "",
          "doc_id": "file_e787926158704f95aad6bc967619f176_10285263",
          "content": "Name: San Zhang Sex: Male Age: 23",
          "_rc_score": 0,
          "workspace_id": "llm-4u5xpd1xdjxxxxxx",
          "hier_title": "",
          "page_number": [
            0
          ],
          "doc_name": "San Zhang's Resume",
          "pipeline_id": "8mbtdxxxxx",
          "_id": "llm-4u5xpd1xdjxxxxxx_8mbtdxxxxx_file_e787926158704f95aad6bc967619f176_10285263_0_0"
        },
        "Score": 0.1736905574798584,
        "Text": "Name: San Zhang Sex: Male Age: 23"
      }
    ]
  },
  "Message": "success",
  "RequestId": "1ED6CECE-xxxx-5B21-91DB-410E0219412A",
  "Status": 200,
  "Success": true
}
Seul le CV de San Zhang est renvoyé, car il s'agit du seul fichier possédant les deux tags.

Références

RessourceDescription
Créer et utiliser une base de connaissancesGuide d'utilisation de la base de connaissances
API RetrieveRécupérer des segments de texte depuis une base de connaissances
Accorder des autorisations API à un utilisateur RAMConfigurer l'accès utilisateur RAM pour l'API Retrieve
Codes d'erreurRésoudre les échecs d'appel API