Simple caching of openai Embeddings

cache.py

import openai

text_to_embeddings_cache = {}

def get_embeddings_for_instructions(instructions: list[str]):
        response = openai.Embedding.create(
            input=instructions,
            model="text-embedding-ada-002"
        )
        return response['data']

i = 0
instruction_batch = []
with tqdm(total=len(language_instructions_to_actions)) as pbar:
        while i < len(language_instructions_to_actions):
                pbar.update(1)
                # skip instructions that are already in the cache
                if language_instructions_to_actions[i]["instruction"] in text_to_embeddings_cache:
                        i += 1
                        continue
                # add instruction to batch
                instruction_batch.append(language_instructions_to_actions[i]["instruction"])

                # batch up to 100 instructions at a time
                if len(instruction_batch) == 100:
                        embeddings = get_embeddings_for_instructions(instruction_batch)
                        assert len(embeddings) == len(instruction_batch)
                        for j in range(len(instruction_batch)):
                                text_to_embeddings_cache[instruction_batch[j]] = embeddings[j]["embedding"]
                        instruction_batch = []
                i += 1

# process the remaining instructions
if len(instruction_batch) > 0:
        embeddings = get_embeddings_for_instructions(instruction_batch)
        assert len(embeddings) == len(instruction_batch)
        for j in range(len(instruction_batch)):
                text_to_embeddings_cache[instruction_batch[j]] = embeddings[j]["embedding"]
        instruction_batch = []