knowledge_experiments.py

import os, json
from openai import OpenAI




textfile = "data/alice_in_wonderland_small.txt"
with open(textfile, 'r') as f:
    text = f.read()

client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])
def ask_gpt(request, json=True, temp=2, seed=10, max_print=300):
    info_message = request
    user_message = text
    print("SYS:\n" + info_message[:max_print])
    print("USER:\n" + user_message[:max_print])
    response = client.chat.completions.create(
        model="gpt-3.5-turbo-0125",
        response_format={ "type": "json_object" if json else "text"},
        messages=[
            {"role": "system", "content": info_message},
            {"role": "user", "content": user_message}
        ],
        temperature=1,
        seed=seed
    )
    for i,c in enumerate(response.choices):
        print("GPT ({}): {}".format(i, c.message.content[:max_print]))
    print("\n\n")
    response = response.choices[0].message.content
    return response

# entity_types = ["character", "location", "special item", "event", "key relationship"]
# example = """{"characters": [...], "locations": [...], "special_items": [...], "relationships": [...], ...}"""
#   "locations": ["Jack's Diner", "blu on 4th"...],
# example = """{
# }"""

entities_list_request = " ".join([
    "You are a helpful assistant designed to process pages from a novel into a JSON file.",
    "You are preparing data to be used in a wiki about the novel, including articles on characters, special items, and locations.",
    "Use specific proper nouns and succinct unique descriptors rather than general terms.",
    "Read the following chapter and generate a list of named entities that would appear as titles of wiki pages in a wiki about the story.",
    # "Each entity should be under a  a type like one of {}".format(entity_types),
    """The output should be in a format like {"characters": [...], "locations": [...], "special_items": [...], "relationships": [...], ...}""",
])


entities_json = json.loads(ask_gpt(entities_list_request, json=True))
print(entities_json)

wiki = {}
for entity_type in entities_json:
    entity_article_request = " ".join([
        "You are a helpful assistant designed to process a novel into a comprehensive wiki in JSON format.",
        f"Read the following chapter and generate a concise wiki article about each the following {entity_type}, structured as",
        str({entity_type: {entity: "..." for entity in entities_json[entity_type]}}),
        # "{" + entity_type + ": [...]"
        # "You are preparing data to be used in a wiki about the novel, including articles on characters, special items, and locations.",
        # "Use specific proper nouns and succinct unique descriptors rather than general terms.",
        # "Read the following chapter and generate a list of named entities that would appear as titles of wiki pages in a wiki about the story.",
        # # "Each entity should be under a  a type like one of {}".format(entity_types),
        # """The output should be in a format like {"characters": [...], "locations": [...], "special_items": [...], "relationships": [...], ...}""",
    ])
    wiki[entity_type] = json.loads(ask_gpt(entity_article_request, json=True))[entity_type]

print(wiki)
savefile = 'wiki2.json'
with open(savefile, 'w', encoding='utf-8') as f:
    json.dump(wiki, f, ensure_ascii=False, indent=4)

import json