The author also discusses several new releases and updates to various plugins and tools. These include llm-gemini 0.1a2, which adds support for system prompts, and datasette-enrichments-gpt, which can enrich data in a table by running prompts through GPT-3.5, GPT-4 Turbo, or GPT-4 Vision. The author also provides a list of his recent blog entries, all of which discuss ways he uses LLM tools in his work. The article concludes with a list of the author's recent releases and updates related to Datasette Cloud, and a list of his recent "Today I Learned" (TIL) posts.
Key takeaways:
- Google Gemini Pro 1.5 is now generally available and offers a free tier that allows up to 50 requests a day, with rate limits of 2 per minute.
- OpenAI has released the non-preview version of GPT-4 Turbo, which now integrates GPT-4 Vision directly into the model and supports both functions and JSON output for image inputs.
- Mistral has released a new openly licensed model, Mixtral 8x22B, which is significantly larger than their previous model and is expected to perform extremely well.
- The author has been working on Datasette Extract, a plugin for Datasette that adds structured data extraction from unstructured text, powered by GPT-4 Turbo.