Conversations and Request Executions

Interactions with Wikit Semantics LLM apps can take two distinct forms: individual request executions and full conversations. Understanding these concepts is essential to get the most out of the platform.

Request Executions

A request execution represents a one-time, autonomous interaction with an LLM app. It consists of the following elements:

• The initial request submitted by the user
• The processing performed by the LLM app
• The generated response
• Associated metadata (timestamp, identifiers, etc.)
• Any user feedback

Each execution is independent and constitutes a complete unit of interaction in itself. This approach is particularly suitable for use cases requiring one-off responses or specific processing.

Conversations

A conversation represents a series of connected interactions between a user and a conversational LLM app. It is characterized by:

• A conversational context maintained throughout the exchange
• A chronological succession of messages
• The preservation of interaction history
• The possibility of cross-references between messages
• The maintenance of thematic coherence

Conversations allow for more natural and elaborate interactions, where each message can build on the context of previous exchanges.

Nota bene: Technically speaking, a conversation is a set of request executions.

Traceability and Analysis

Wikit Semantics systematically retains the history of interactions, whether individual executions or complete conversations. This traceability allows for:

• Analysis of LLM app performance
• Identification of usage patterns
• Continuous improvement of responses
• Monitoring user satisfaction via feedback
• Export of data for in-depth analysis

Management and Exploitation

The platform offers several features to effectively exploit this data:

• Search and filtering of interactions
• Analysis of user feedback
• Export of data for external analysis
• Visualization of usage trends
• Identification of improvement opportunities

These management and analysis capabilities are essential for optimizing LLM app performance and ensuring a quality user experience.