Ask HN: What tools for building AI products do you wish you'd discovered sooner?
Here's what I have so far of tools for building ai products:
You'll probably start with an LLM API - OpenAI - Cohere and others aren't as good - Anthropic's isn't available
If you're using embeddings - If you're working with a lot of items, you'll want a vector database, like Pinecone, or Weaviate, or pgvector
If you're building Q&A over a document - I'd suggest using GPT Index
If you need to be able to interact with external data sources, do google searches, database lookups, python REPL - I'd suggest using langchain
If you're doing chained prompts - Check out dust tt and langchain
If you want to deploy a little app quickly - Check out Streamlit
If you need to use something like stable diffusion or whisper in your product - banana dev, modal, replicate, tiyaro ai, beam cloud, inferrd, or pipeline ai
If you need something to optimize your prompts - Check out Humanloop and Everyprompt
If you're building models and need an ml framework - PyTorch, Keras, TensorFlow
If you're deploying models to production - Check out MLOps tools like MLflow, Kubeflow, Metaflow, Airflow, Seldon Core, TFServing
If you need to check out example projects for inspiration - Check out the pinecone op stack, the langchain gallery, the gpt index showcase, and the openai cookbook
If you want to browse the latest research, check out arXiv, of course