Skip to main content
Tyk AI Studio caters to two primary user personas, each with different needs and workflows: the Technical User (AI Developer/Consumer) and the Non-technical User (Business User). An administrator can configure what each user can access. During user onboarding, an admin can assign permissions to view the AI Portal, the Chat interface, or both, tailoring the experience to the user’s role.

​
Technical User (AI Developer)

The Technical User, also known as the AI Developer or AI Consumer, is the primary user of the capabilities exposed by Tyk AI Studio. This persona builds end-user applications, chatbots, and internal tools that leverage the LLMs and data sources configured by the Admin. The rest of this page is written for this technical user.

​
Lifecycle

The workflow for a Technical User typically follows this path:
  1. Discover: Browsing the AI Portal to see available models, tools, and data sources they have access to.
  2. Experiment: Using the Model Playground (Chat Interface) to test prompts, compare model responses, and validate RAG retrieval relevance.
  3. Build: Creating Apps, generating API Keys, and integrating the AI capabilities into their code via the Tyk AI Gateway.
  4. Optimize: Analyzing app-specific metrics (latency, token usage) to refine prompts and improve performance.

​
Core Features

​
Model Playground

An interactive sandbox environment for rapid experimentation.
  • Prompt Testing: Iteratively refine system prompts and user messages.
  • Model Comparison: Switch between different models (e.g., GPT-4 vs. Claude 3) to compare quality and latency.
  • Tool Testing: Verify that Tools (e.g., “Get Weather”) are correctly invoked by the model.

​
API Management

Self-service management of credentials for applications.
  • API Keys: Generate and revoke long-lived API keys for application access.
  • SDK Integration: Use standard OpenAI-compatible SDKs, pointing the base URL to Tyk AI Studio.

​
Prompt Library

A repository for saving and sharing effective prompts.
  • Versioning: Track changes to prompts over time.
  • Templates: Use parameterized prompts to standardize interactions across applications.

​
App Analytics

Granular visibility into application performance.
  • Usage Metrics: View token consumption, request count, and error rates for specific apps.
  • Latency Tracking: Monitor the end-to-end response time of LLM interactions.
  • Debug Traces: Inspect individual request/response pairs to diagnose issues with model outputs or tool execution.

​
Non-technical User (Business User)

The Non-technical User, such as product managers or sales staff, can leverage the power of LLMs without any coding. Administrators can configure a powerful, ready-to-use LLM chat interface complete with all the necessary tools and data sources. This allows business users to directly query the system and get answers to their questions, for example, “Summarise customer feedback for product X in the last quarter” or “What are the key features of our competitor Y?”. They can do this without needing to create an application or manage API keys, making AI accessible to everyone in the organization.