Hi all!

As you may have seen in the BIOVIA Pipeline Pilot 2026 Product Release Document, there are now two official ways to leverage LLMs:

• The new Generative Data Processing Collection runs models locally by downloading from Hugging Face

• The new Answers from LLM Chatbot (Openrouter) component (in the Document & Text Collection) accesses a variety of hosted models via OpenRouter.ai

For those who joined @GP, @JI or myself at BIOVIA Live (Replays available!), you saw us demoing LLM use cases developed before these official collections were finalized. Because Pipeline Pilot is so flexible, we created a third, "developer-first" approach: calling models via OpenAI Python API.

This allows you to simply provide an endpoint URL, API key, and certificate (if needed), for rapid testing. For example, the Investigation Report initiated by Large Language Models use case protocol, shown at BIOVIA Live and shared in the BIOVIA Discoverant community was built using these exact components.

We want to hear from you! Please reach out with questions or share your use cases in the comments, even if things are working perfectly, we’d love to know how you’re using them!

⚠️ Disclaimer ⚠️

Even though these components might become part of the collection in the future, they are only a Proof of Concept today. While they work well for learning and demos, please note:

• They are not production-ready and have limited documentation
• They have been tested primarily on Google Gemini, OpenRouter, and internal endpoints
• Data Privacy: Be mindful of the data you send to public endpoints. Third-party services are not controlled by 3DS and are governed by their own privacy policies.

Resources & Download

We have included an example protocol for each one and this little tutorial video: