bulevar / steps.md
jonny's picture
Upload gradio app (manual upload)
a93809e
|
raw
history blame
443 Bytes

Steps

  1. Pre-process the knowledge base (docs)
  2. Store it in a database (embeddings)
  3. Inject content into GPT-3 prompt

App building specs

  1. langchain - for chaining prompts
  2. langflow - displays prompting with database
  3. gradio - integrates with huggingface (important, to easily demo the work being done, and let people pay for the openai results)
  4. pinecone (cloud vectorstore database, instead of chromadb which runs local)