Skip to main content

Proof of Concept for LLM Apps

PoC (Proof of Concept) should be done within days, not weeks or months. Meanwhile, testing LLM Apps with real-world users is challenging if the app is not easily accessible. This guide provides a step-by-step guide to building LLM Apps for PoC and comparing results.

Expectation​

This example demonstrates polishing English text using multiple prompts:

Step-by-Step Guide​

1. Creating new LLM Apps​

Create LLM Apps using any tool that works best for you. You can use OpenAI Assistants, GPTs, LangChain, or native Python applications. Visit other use cases to learn more about creating LLM Apps.

2. Adding the Apps to Runbear​

Add the LLM Apps to Runbear with just a few clicks. Visit the Integrations page, choose the tool you used, and follow the instructions.

3. Connecting the Apps to a single channel​

Using Runbear, connecting the LLM Apps and channels is straightforward. Visit the Connecting Channels with LLM Apps page and follow the guide.

4. (Optional) Updating Configs for Comparison​

Make sure the "Description-Based LLM App Selection" option is disabled to receive answers from the LLM Apps rather than receiving an answer from the most relevant one. You can locate this option on the Channel Preferences page.

Testing​

Open the channel (e.g., Slack channel) and start testing the LLM Apps. For example, invoke the LLM Apps by mentioning @Runbear with a query.

All Done! 🎉​

Introduce your LLM Apps to the real world and start collecting feedback. If you have any questions, please ask us at Runbear.