RAG
Query a chosen LLM with a structured prompt and vectorized data and retrieve its response
💡
Learn More about Retrieval Augmented Generation (RAG) on Weaviate (opens in a new tab)
Input Parameters
- Input Field for Query:
- Vector DB: Choose vector database to use for RAG process
- Limit: Upper limit of how many documents to be fed to the model
- Certainty: Decimal value between 0 to 1 for how loose or strict to be on what data is given to the model
- Filters (JSON): Custom JSON to filter which documents are accessed by key attributes
- System Prompt: A prompt to give the model context as to what its purpose is in relation to the process
- User Template: A prompt for the model combining the data provided and the specific query for the workflow
- Embedding Model Name: Choose model to embed data as vectors
- Generative Model Name: Choose LLM from your activated models to use for the query
Expected Output
Generated text designed by the prompt template imformed by vectorized data from database
Example Use Case
In this example, a vector database of recipes is embedded using OpenAI's ADA-002 model and combined with a prompt and GPT 3.5 is queried and its response is the output