discordToShareGPT.py

import json

filenames = []
out_name = "processed.json"
myname = "YOURDISCORDUSERNAME"

for filename in filenames:
    with open(filename, encoding="utf-8") as file:
        data = json.load(file)["messages"]
    newdata = [[data[i]["author"]["name"], data[i]["content"]] for i in range(len(data)) if data[i]["content"] != ""]
    
    merged = []
    current_name = newdata[0][0]
    acc = ""
    for i in newdata:
        if i[0] != current_name:
            merged += [acc]
            current_name = i[0]
            acc = ""
        acc += i[1] + "\n"

    if newdata[0][0] == myname:
        merged = merged[1:]

    # Iterate over merged, two messages at a time
    # Make an array of exchanges 10 pairs in length
    long_conversation = []
    for i in range(0, len(merged) - 1, 2):
        long_conversation += [{"from": "human", "value": merged[i]}, {"from": "gpt", "value": merged[i+1]}]
    jsons = []
    while len(long_conversation) > 0:
        chunk_size = 10
        chunk = long_conversation[:chunk_size * 2]
        jsons += [{"conversations": chunk}]
        long_conversation = long_conversation[len(chunk):]
    # Add Jsons to output file, with each json on a new line
    with open(out_name, "a", encoding="utf-8") as f:
        for this_json in jsons:
            JSON = json.dumps(this_json) + "\n"
            f.write(JSON)