Open Source

---

After all, a Symphony is a complex workflow of musical notes flowing in harmony

A platform developed from an open framework to help AI researchers and developers to rapidly prototype, test and deploy workflows that include machine learning models as worker components. The workflow can be invoked on command line or web service, as well as integrated into UX/UI as part of a larger intelligent solution.

Early Github access to this platform and experiments: 

Code branch: https://github.com/2ndthoughts/symphonyGPT/tree/Early-Access-3

Release: https://github.com/2ndthoughts/symphonyGPT/releases/tag/EA3

This release introduces automatic embeddings using chromaDB for all API extractor classes. What this means is that large document retrievals in RAG symphonies now all have the advantage of using embeddings to not only pick relevant sections of a large API retrieval, but also pick relevant individual retrievals by rank. As a result, you get more accurate RAG that reads entire retrieved documents.

In this release, we also added an API extractor called WikipediaExtractor that, as you may have guessed, searches wikipedia using your prompt, and retrieves pertinent articles to flow into your LLM models for summarizing or additional generative activities.

As usual, contact us at dev@2ndthoughts.ai for any questions, or post in this forum if you need help setting up or have an idea on how to improve symphonies. Cheers.