Choose LLM and Embeddings Options

We’re excited to announce that Pyq can now embed text using the famed Instructor XL embeddings model, and also supports querying with Claude from Anthropic.

‍

New Embeddings Model Options

Instructor XL is a particularly powerful embedding model because of its instruction tuning.  It can embed text for domain and task specific use cases, allowing our users to unlock better performance from their queries.

When creating indices, you can now specify either MPTNET (default) or instructor XL with an instruction as the embedding model for that index.  All subsequent uploads and all queries will then follow the model the index was created with.

‍

New LLM Options

Similarly, when using our ask and ask-formatted endpoints, you can now specify “claude” as the model, if you want to use Claude as your querying model.  Omitting a specific model will result in unaffected behavior.  

Claude is a personal favorite model of ours, as it manages knitting patterns much better than OpenAI, due to its significantly larger context window (200k tokens vs 128k).

‍

Conclusion

Book a call here to learn more and get started! You can also take a look at our technical documentation here.

Thanks so much for tagging along for pyq’s first feature drop week! Stay tuned for more exciting updates.