Documentation

English

Français

English

Français

LLM App Query Execution ​

The LLM App is the central concept of the solution. It is responsible for orchestrating Large Language Models (LLMs) and related tools (for example, semantic search of document fragments).

Question-Answer type LLM Apps have Data Sources to extract answer elements for users.

Authentication and Authorization ​

An API key is required to invoke the query execution API.

This key must be placed in a wikit-semantics-api-key HTTP header.

It is provided by Wikit.

Query Execution ​

POST /semantics/apps/{llm_app_id}/query-executions

bash
curl "https://apis.wikit.ai/semantics/apps/$SEMANTICS_APP_ID/query-executions" \
-H "Content-Type: application/json" \
-H "Wikit-Semantics-API-Key: $SEMANTICS_API_KEY" \
-H "X-Wikit-Response-Format: json" \
-H "X-Wikit-Organization-Id: $SEMANTICS_ORG_ID" \
-d '{
  "query": "How many days of special leave am I entitled to for my wedding?"
}'

Response ​

jsx
{
    "answer": "In France, you are entitled to 4 days of special leave for your marriage or PACS, according to the Labor Code (article L3142-1).",
    "queryId": "67c5646e05d2ed84ac20495d",
    "conversation_id": "67c5646e05d2ed84ac20495c",
    "metadata": null
}

Query Execution with Streaming ​

POST /semantics/apps/{llm_app_id}/query-executions?is_stream_mode=true

Response streaming is the standard way LLMs generate content token by token.

To enable it, simply set the is_stream_mode parameter to true in the query part of the URL:

bash
curl "https://apis.wikit.ai/semantics/apps/$SEMANTICS_APP_ID/query-executions?is_stream_mode=true" \
-H 'Content-Type: application/json' \
-H "Wikit-Semantics-API-Key: $SEMANTICS_API_KEY" \
-H "X-Wikit-Response-Format: json" \
-H "X-Wikit-Organization-Id: $SEMANTICS_ORG_ID" \
-d '{
  "query": "How many days of special leave am I entitled to for my wedding?"
}'

Response ​

jsx
data: {"queryId": "67c565542c8f2bb8fb52d43a", "chunk":"In "}STOP

data: {"queryId": "67c565542c8f2bb8fb52d43a", "chunk":"Fra"}STOP

data: {"queryId": "67c565542c8f2bb8fb52d43a", "chunk":" ce,"}STOP

data: {"queryId": "67c565542c8f2bb8fb52d43a", "chunk":" you "}STOP

data: {"queryId": "67c565542c8f2bb8fb52d43a", "chunk":" are"}STOP

Query Execution in an Existing Conversation ​

With the Conversational Question-Answer LLM App type, a conversation is created upon the first query execution via the /semantics/apps/{llm_app_id}/query-executions endpoint. The conversation ID is returned in the JSON output or in the response headers in streaming mode (X-wikit-Conversation-Id): it must be used in subsequent query executions to continue the conversation, via the X-Wikit-Conversation-Id HTTP header:

json
curl "https://apis.wikit.ai/semantics/apps/$SEMANTICS_APP_ID/query-executions?is_stream_mode=true" \
-H 'Content-Type: application/json' \
-H "Wikit-Semantics-API-Key: $SEMANTICS_API_KEY" \
-H "X-Wikit-Response-Format: json" \
-H "X-Wikit-Organization-Id: $SEMANTICS_ORG_ID" \
-H "X-Wikit-Conversation-Id: $SEMANTICS_CONVERSATION_ID" \
-d '{
  "query": "What about my daughter's birthday?"
}'