Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

Code Block
breakoutModewide
{
  "languageModel": "AZURE_GPT_4o_2024_0513",
  "maxLoopIterations": 5,
  "tools": [
    {
      "name": "InternalSearch",
      "configuration": {
        "languageModel": "AZURE_GPT_4o_2024_0513",
        "scopeIds": ["scope_A", "scope_B"]
        "maxTokensForSources": 14000,
        "chunkRelevancySortConfig": {
          "enabled": true,
          "languageModel": "AZURE_GPT_4o_MINI_2024_0718"
        }
      }
    },
    {
      "name": "WebSearch",
      "configuration": {
        "languageModel": "AZURE_GPT_4o_2024_0513",
        "maxTokensForSources": 14000,
        "chunkRelevancySortConfig": {
          "enabled": true,
          "languageModel": "AZURE_GPT_4o_MINI_2024_0718"
        }
      }
    }
  ],
  "evaluationConfig": {
    "maxReviewSteps": 2,
    "evaluationPassValues": [
      "low",
      "medium"
    ]
  },
  "followUpQuestionsConfig": {
    "languageModel": "AZURE_GPT_4o_MINI_2024_0718",
    "numberOfFollowUpQuestions": 2
  }
}

...

Attribute

Description

Type

languageModel

Defines the default model used by the agent

Text enum (depending on availability): “AZURE_GPT_4o_2024_0513“,
”AZURE_GPT_4_0613”,

maxLoopIterations

Max. number of loop iterations done by the agent to answer the question

Number

tools

List of tools (see below for specific search types and configurations)

name

Name of search type

Name of search type:

  • InternalSearch

  • WebSearch

configuration

Configuration of search

Depends on search type. See below for more details

evaluationConfig

maxReviewSteps

Specifies the maximum number of review steps allowed if the evaluation check is not passed.

Number

evaluationPassValues

Defines the pass levels for evaluation checks. The levels can be categorised as low, medium, or high.

Possible Values:

  • ["low"]

  • ["low", "medium"]

  • ["low", "medium", "high"]

followUpQuestionsConfig

languageModel

Defines the model used for generation of follow-up questions

Text enum (depending on availability): “AZURE_GPT_4o_2024_0513“,
”AZURE_GPT_4_0613”,

numberOfFollowUpQuestions

Specifies the number of follow-up questions to generate

Number

Internal Knowledge Search

Search in information in the knowledge base

Attribute

Description

Type

languageModel

Specifies the model used in the internal search tool.

Text enum (depending on availability):

  • "AZURE_GPT_4o_2024_0513",

  • "AZURE_GPT_4_0613",

  • ...

maxTokensForSources

Defines the maximum number of tokens for the returned search results. The search results are provided in chunks.

Number

scopeIds

Defines scopes the search is performed on.

List of scope IDs

chunkRelevancySortConfig

enabled

Indicates whether chunk relevancy sorting is activated.

Boolean

languageModel

Specifies the model used for chunk relevancy sorting.

Text enum (depending on availability):

  • "AZURE_GPT_4o_2024_0513",

  • "AZURE_GPT_4_0613",

  • ...

Web Search

Search in the internet.

Attribute

Description

Type

languageModel

Specifies the model used in the internal search tool.

Text enum (depending on availability):

  • "AZURE_GPT_4o_2024_0513",

  • "AZURE_GPT_4_0613",

  • ...

maxTokensForSources

Defines the maximum number of tokens for the returned search results. The search results are provided in chunks.

Number

chunkRelevancySortConfig

enabled

Indicates whether chunk relevancy sorting is activated.

Boolean

languageModel

Specifies the model used for chunk relevancy sorting.

Text enum (depending on availability):

  • "AZURE_GPT_4o_2024_0513",

  • "AZURE_GPT_4_0613",

  • ...

(Tool) Definition

Note

Only adjust function definition if you are fully familiar with the code logic. Small changes can break the module or reduce the output quality.

...