Lab
Playground
In the Playground section, you can experiment with various language models simultaneously and test them with any text you desire. You can adjust settings such as creativity level (temperature) and output length (max tokens).
In this guide we will cover the following:
- How to select models: Choose from a variety of models, both open-source and closed-source, from multiple providers.
- Adjust settings: Experiment with temperature, max tokens, and system prompts to tailor the AI models’s behavior.
- Testing Text: Enter the text you want the AI to continue or generate new content from.
- Viewing Results: Analyze the output generated by the AI models.
- Sharing: Share the results with others for feedback or further analysis.
Get started
Create a new playground or use the default playground to get started.
Add more language models so you can test them simultaneously.
Prem gives you the ability to observe how different AI models behave under various conditions.
Customize the models in the Params section.
The Params section is your go-to place to configure language models and adjust system prompts for the language model of your choice.
Param settings
Max Tokens: Limit for generated tokens in a chat, constrained by the model’s context length (value 0 means no limit).
Temperature: This parameter in language models affects the creativity or diversity of the generated text. You can set the randomness level (0 to 1). Higher values increase randomness, and the results will be more creative, with the model generating more diverse output. Lower values result in more focused and deterministic output (great for working with repositories).
System Prompt: The system prompt sets the initial context and direction for the model’s responses. It should be carefully selected to align with the specific topic and desired style of the conversation.
Add a repository to grant the language models access to your documents for easy setup of RAG.
Repository settings
Limit: Sets the maximum number of top-matching documents to retrieve based on similarity.
Similarity: Measures how closely a query matches document embeddings. Higher values indicate greater similarity, ranging from 0 to 1.
Trace & Launch from the lab
-
Use the Trace tool to explore the results and sub components of the language model’s response.
-
While in the lab, you can fast-track your way to launch. If you decide to launch a model from the lab, the model you choose will automatically load in the lab along with the params and repositories you’ve set with it.
Share your experiemts
Feel free to share your experiments with your colleagues or online communities.
Don’t worry about losing your tests; all activities in the Lab are saved. You can revisit them to analyze what worked well and what didn’t.
Chat
In the Chat section, you can interact with a language model in a conversational manner. Unlike the Playground, where you can experiment with multiple models simultaneously, here you can focus on testing out one model in a chat-like manner.
Similar to the Playground, you can adjust settings such as max tokens, temperature, and system prompts to customize the AI’s responses. Additionally, you can share the chat sessions with others for collaboration or feedback.
Status Page
We continuously monitor the performance and availability of all model providers. For detailed information on inference metrics and uptime, please refer to our status page.
