This script processes arbitrary yaml output (eg: helm or kubectl commands) sends it to OpenAI for text embedding and stores the embedding in a ChromaDB database. Then we query the ChromaDB for similar documents based on the embedded vector which we can provide in the form of context when asking questions to OpenAI's GPT-4 model.

1. Requirements
2. Usage
3. Functions
4. Error Handling

• Python 3.6 or above
• Libraries: sys, re, time, os, argparse, yaml, json, requests, subprocess, chromadb, hashlib, openai

python vector_embed_yaml.py [-h] [-p] [-e] [-c COMMAND] [-q QUERY] [--n_results N_RESULTS]

• -p or --parse: Only parse the helm output.
• -e or --embed: Send the parsed helm output to OpenAI for embedding.
• -c or --command: Command that outputs YAML to be processed.
• -q or --query: Query the ChromaDB for similar documents.
• --n_results: Number of similar results to fetch. Default is 2.

• get_gpt4_response(embeddings, user_question, model="gpt-4"): Fetches a response from the GPT-4 model based on the user's question and the document embeddings.
• query_chromadb(query_text, n_results=10, metadata_filter=None, document_filter=None): Queries the ChromaDB for similar documents based on the embedded vector.
• command_already_embedded(command, collection): Checks if the given command is already in the metadata of any stored embeddings.
• display_chromadb_contents(): Displays the contents of the ChromaDB database for debugging purposes.
• compute_id_from_command(command): Computes a unique ID for each command using SHA-256 hashing.
• store_embedding_in_chromadb(embedding_response, command, helm_output): Stores the embedding in ChromaDB with associated metadata.
• parse_helm_output(output): Parses the helm output into separate documents.
• get_openai_embedding(text, model_id="text-embedding-ada-002", api_key=None): Sends text to OpenAI for embedding.
• main(args): Main function that processes command-line arguments and orchestrates the script's workflow.

The script exits with an error message if:

• No action (parse or embed) is chosen.
• No helm command is provided.