AI Knowledge Base Chat Connector: Functionality and Operation

The AI Knowledge Base Chat Connector generates AI-supported responses to user prompts and uses previously stored data to answer them.

Properties

Operation

Defines the operation executed by the adapter.

Possible values:

AI-Knowledgebase-Chat: Generates AI-supported responses to user prompts. Previously stored data is used for the response.

Parameters

`Adapter`	Main adapter class (do not change!) Possible values: de.softproject.integration.nativeadapter.NativeAdapter: Main class (Default)
`Repository`	The virtual repository of the adapter project. Possible values: `876cf76d-bcb1-463e-8433-68270d04c00e (Default)`
`protocol`	Used protocol (HTTP oder HTTPS) Possible values: `http`: Use HTTP (no encryption) (Default) `https`: Use HTTPS (SSL-encrypted)
`host`	Name of AI gateway host Possible values: Any valid character string without protocol specification (e.g. `localhost`)
`port`	(optional) Number of AI gateway port Possible values: Any integer (e.g. `9999`)
`accessToken`	Is required to establish a Keycloak connection. Possible values: unique access token The access token needs to be generated in Keycloak.
`configKey`	Configuration key for the chat service. The key originates from a valid X4AI configuration in the system. Possible values: Any string Examples: `default` `general-knowledge`
`datasets`	List of data records used for the response, comma-separated. Possible values: Any character string for existing data records in the system, comma-separated. Example: `x4.documentation.ver7-3-1,x4.documentation.ver7-2-3`
`backstory`	Initial context or background information for the LLM (Large Language Model). There are two options: a statement with the variables `{knowledgebase}` and `{prompt}` in the parameter `backstory` Text file (`.txt`) with an statement in the folder Resources Possible values (Examples) for a statement with the variables `{knowledgebase}` and `{prompt}`: `Complete the prompt based only on the following context: {knowledgebase} {prompt}` `You are a helpful assistant and an X4 BPMS expert: {knowledgebase} {prompt}` Possible value when using a text file: `Path to a txt file` Example: `%USER_COLLECTION_ABSOLUTE%/X4AI-Example-ChatAdapter/Resources/backstory.txt` Example contents of the text file: `You are a helpful assistant and an X4 BPMS expert: {knowledgebase} {prompt}`
`callbackLog`	Outputs information during execution, such as token consumption.

Status values

`1 (successful)`	The operation was executed successfully.
`0`	The operation was executed successfully, but without result.
`-1 (failed)`	The operation failed due to a technical error.

Input

The adapter expects an XML document with the following structure as input.

CODE

<x4ai_adapter>
  <prompt>your prompt</prompt>
</x4ai_adapter>

The following elements are included in the input:

Element	Description
`x4ai_adapter`	Enclosing element for the `AI-Knowledgebase-Chat` operation.
`prompt`	Prompt or question, i.e. instruction to the AI

Output

The adapter generates a XML document as output:

CODE

<output>
  <original_prompt>customer’s question</original_prompt>
  <response>AI Response</response>
  <sources>
    <metadata>
      <id>uuid</id>
      <source>document name</source>
      <page>page</page>
      <source_url>url</source_url>
      <version>document version</version>
      <similarity_score>number between (0; 1)</similarity_score>
    </metadata>
    <metadata>
      <id></id>
      <source>X4BPMS_731_ESB_EN</source>
      <page>588</page>
      <source_url>http://download.softproject.de/x4bpms/7-3-1/X4BPMS_731_ESB_EN.pdf#page=588</source_url>
      <version>7.3.1</version>
      <similarity_score>0.16465315713444784</similarity_score>
    </metadata>
  </sources>
  <callback_log>
    <total_tokens>30</total_tokens>
    <completion_tokens>13</completion_tokens>
    <prompt_tokens>17</prompt_tokens>
    <total_cost_usd>0.01327</total_cost_usd>
    <successful_requests>1</successful_requests>
  </callback_log>
</output>

The following elements are included in the output:

Element	Description
`Output`	Root element for the operation `AI-Semantic-Search`.
`original_prompt`	Original question to the AI
`response`	AI answer
`sources`	Enclosing element for the element `metadata`.
`metadata`	Meta data of document with the elements: `id`: Document ID `source`: unique name (`sourceKey`) of the document `page`: Document page `source_url`: Document URL `version`: Document version (Version: 7.3 1 7.4.0) `similarity_score`: Similarity score of answer to the question asked. The lower the value, the better the match with the question.
`callback_log`	Returns the token consumption during execution. Contains the elements: `total_tokens`: Total number of tokens `completion_tokens`: Number of tokens used for completions `prompt_tokens`: Number of tokens used for prompts `total_cost_usd`: Total cost of the request in USD `successful_requests`: Number of successful requests