|
|
|
import json |
|
import typing |
|
|
|
|
|
from modules.common.file_ensurer import FileEnsurer |
|
from modules.common.logger import Logger |
|
from modules.common.toolkit import Toolkit |
|
|
|
|
|
|
|
class JsonHandler: |
|
|
|
""" |
|
|
|
Handles the Kijiku Rules.json file and interactions with it. |
|
|
|
""" |
|
|
|
current_kijiku_rules = dict() |
|
|
|
kijiku_settings_message = """ |
|
---------------------------------------------------------------------------------- |
|
Kijiku Settings: |
|
|
|
prompt_assembly_mode : 1 or 2. 1 means the system message will actually be treated as a system message. 2 means it'll be treated as a user message. 1 is recommend for gpt-4 otherwise either works. For Gemini, this setting is ignored. |
|
|
|
number_of_lines_per_batch : The number of lines to be built into a prompt at once. Theoretically, more lines would be more cost effective, but other complications may occur with higher lines. So far been tested up to 48. |
|
|
|
sentence_fragmenter_mode : 1 or 2 (1 - via regex and other nonsense) 2 - None (Takes formatting and text directly from API return)) the API can sometimes return a result on a single line, so this determines the way Kijiku fragments the sentences if at all. Use 2 for newer models. |
|
|
|
je_check_mode : 1 or 2, 1 will print out the jap then the english below separated by ---, 2 will attempt to pair the english and jap sentences, placing the jap above the eng. If it cannot, it will default to 1. Use 2 for newer models. |
|
|
|
number_of_malformed_batch_retries : (Malformed batch is when je-fixing fails) How many times Kijiku will attempt to mend a malformed batch (mending is resending the request), only for gpt4. Be careful with increasing as cost increases at (cost * length * n) at worst case. This setting is ignored if je_check_mode is set to 1. |
|
|
|
batch_retry_timeout : How long Kijiku will try to translate a batch in seconds, if a requests exceeds this duration, Kijiku will leave it untranslated. |
|
|
|
number_of_concurrent_batches : How many translations batches Kijiku will send to the translation API at a time. For OpenAI, be conservative as rate-limiting is aggressive, I'd suggest 3-5. For Gemini, do not exceed 60. |
|
---------------------------------------------------------------------------------- |
|
Open AI Settings: |
|
See https://platform.openai.com/docs/api-reference/chat/create for further details |
|
---------------------------------------------------------------------------------- |
|
openai_model : ID of the model to use. Kijiku only works with 'chat' models. |
|
|
|
openai_system_message : Instructions to the model. Basically tells the model how to translate. |
|
|
|
openai_temperature : What sampling temperature to use, between 0 and 2. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. Lower values are typically better for translation. |
|
|
|
openai_top_p : An alternative to sampling with temperature, called nucleus sampling, where the model considers the results of the tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. I generally recommend altering this or temperature but not both. |
|
|
|
openai_n : How many chat completion choices to generate for each input message. Do not change this. |
|
|
|
openai_stream : If set, partial message deltas will be sent, like in ChatGPT. Tokens will be sent as data-only server-sent events as they become available, with the stream terminated by a data: [DONE] message. See the OpenAI python library on GitHub for example code. Do not change this. |
|
|
|
openai_stop : Up to 4 sequences where the API will stop generating further tokens. Do not change this. |
|
|
|
openai_logit_bias : Modifies the likelihood of specified tokens appearing in the completion. Do not change this. |
|
|
|
openai_max_tokens : The maximum number of tokens to generate in the chat completion. The total length of input tokens and generated tokens is limited by the model's context length. I wouldn't recommend changing this. Is none by default. If you change to an integer, make sure it doesn't exceed that model's context length or your request will fail and repeat till timeout. |
|
|
|
openai_presence_penalty : Number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far, increasing the model's likelihood to talk about new topics. While negative values encourage repetition. Should leave this at 0.0. |
|
|
|
openai_frequency_penalty : Number between -2.0 and 2.0. Positive values penalize new tokens based on their existing frequency in the text so far, decreasing the model's likelihood to repeat the same line verbatim. Negative values encourage repetition. Should leave this at 0.0. |
|
---------------------------------------------------------------------------------- |
|
openai_stream, openai_logit_bias, openai_stop and openai_n are included for completion's sake, current versions of Kudasai will hardcode their values when validating the Kijiku_rule.json to their default values. As different values for these settings do not have a use case in Kudasai's current implementation. |
|
---------------------------------------------------------------------------------- |
|
Gemini Settings: |
|
https://ai.google.dev/docs/concepts#model-parameters for further details |
|
---------------------------------------------------------------------------------- |
|
gemini_model : The model to use. Currently only supports gemini-pro and gemini-pro-vision, the 1.0 model and it's aliases. |
|
|
|
gemini_prompt : Instructions to the model. Basically tells the model how to translate. |
|
|
|
gemini_temperature : What sampling temperature to use, between 0 and 2. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. Lower values are typically better for translation. |
|
|
|
gemini_top_p : An alternative to sampling with temperature, called nucleus sampling, where the model considers the results of the tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. I generally recommend altering this or temperature but not both. |
|
|
|
gemini_top_k : Determines the number of most probable tokens to consider for each selection step. A higher value increases diversity, a lower value makes the output more deterministic. |
|
|
|
gemini_candidate_count : The number of candidates to generate for each input message. Do not change this. |
|
|
|
gemini_stream : If set, partial message deltas will be sent, like in Gemini Chat. Tokens will be sent as data-only server-sent events as they become available, with the stream terminated by a data: [DONE] message. Do not change this. |
|
|
|
gemini_stop_sequences : Up to 4 sequences where the API will stop generating further tokens. Do not change this. |
|
|
|
gemini_max_output_tokens : The maximum number of tokens to generate in the chat completion. The total length of input tokens and generated tokens is limited by the model's context length. I wouldn't recommend changing this. Is none by default. If you change to an integer, make sure it doesn't exceed that model's context length or your request will fail and repeat till timeout. |
|
---------------------------------------------------------------------------------- |
|
gemini_stream, gemini_stop_sequences and gemini_candidate_count are included for completion's sake, current versions of Kudasai will hardcode their values when validating the Kijiku_rule.json to their default values. As different values for these settings do not have a use case in Kudasai's current implementation. |
|
---------------------------------------------------------------------------------- |
|
""" |
|
|
|
|
|
|
|
@staticmethod |
|
def validate_json() -> None: |
|
|
|
""" |
|
|
|
Validates the Kijiku Rules.json file. |
|
|
|
""" |
|
|
|
base_kijiku_keys = [ |
|
"prompt_assembly_mode", |
|
"number_of_lines_per_batch", |
|
"sentence_fragmenter_mode", |
|
"je_check_mode", |
|
"number_of_malformed_batch_retries", |
|
"batch_retry_timeout", |
|
"number_of_concurrent_batches" |
|
] |
|
|
|
openai_keys = [ |
|
"openai_model", |
|
"openai_system_message", |
|
"openai_temperature", |
|
"openai_top_p", |
|
"openai_n", |
|
"openai_stream", |
|
"openai_stop", |
|
"openai_logit_bias", |
|
"openai_max_tokens", |
|
"openai_presence_penalty", |
|
"openai_frequency_penalty" |
|
] |
|
|
|
gemini_keys = [ |
|
"gemini_model", |
|
"gemini_prompt", |
|
"gemini_temperature", |
|
"gemini_top_p", |
|
"gemini_top_k", |
|
"gemini_candidate_count", |
|
"gemini_stream", |
|
"gemini_stop_sequences", |
|
"gemini_max_output_tokens" |
|
] |
|
|
|
validation_rules = { |
|
"prompt_assembly_mode": lambda x: isinstance(x, int) and 1 <= x <= 2, |
|
"number_of_lines_per_batch": lambda x: isinstance(x, int) and x > 0, |
|
"sentence_fragmenter_mode": lambda x: isinstance(x, int) and 1 <= x <= 2, |
|
"je_check_mode": lambda x: isinstance(x, int) and 1 <= x <= 2, |
|
"number_of_malformed_batch_retries": lambda x: isinstance(x, int) and x >= 0, |
|
"batch_retry_timeout": lambda x: isinstance(x, int) and x >= 0, |
|
"number_of_concurrent_batches": lambda x: isinstance(x, int) and x >= 0, |
|
"openai_model": lambda x: isinstance(x, str) and x in FileEnsurer.ALLOWED_OPENAI_MODELS, |
|
"openai_system_message": lambda x: x not in ["", "None", None], |
|
"openai_temperature": lambda x: isinstance(x, float) and 0 <= x <= 2, |
|
"openai_top_p": lambda x: isinstance(x, float) and 0 <= x <= 1, |
|
"openai_max_tokens": lambda x: x is None or isinstance(x, int) and x > 0, |
|
"openai_presence_penalty": lambda x: isinstance(x, float) and -2 <= x <= 2, |
|
"gemini_model": lambda x: isinstance(x, str) and x in FileEnsurer.ALLOWED_GEMINI_MODELS, |
|
"gemini_prompt": lambda x: x not in ["", "None", None], |
|
"gemini_temperature": lambda x: isinstance(x, float) and 0 <= x <= 2, |
|
"gemini_top_p": lambda x: x is None or (isinstance(x, float) and 0 <= x <= 2), |
|
"gemini_top_k": lambda x: x is None or (isinstance(x, int) and x >= 0), |
|
"gemini_max_output_tokens": lambda x: x is None or isinstance(x, int), |
|
} |
|
|
|
try: |
|
|
|
assert "base kijiku settings" in JsonHandler.current_kijiku_rules |
|
assert "openai settings" in JsonHandler.current_kijiku_rules |
|
assert "gemini settings" in JsonHandler.current_kijiku_rules |
|
|
|
|
|
base_kijiku_settings = JsonHandler.current_kijiku_rules["base kijiku settings"] |
|
openai_settings = JsonHandler.current_kijiku_rules["openai settings"] |
|
gemini_settings = JsonHandler.current_kijiku_rules["gemini settings"] |
|
|
|
|
|
|
|
assert all(key in base_kijiku_settings for key in base_kijiku_keys) |
|
assert all(key in openai_settings for key in openai_keys) |
|
assert all(key in gemini_settings for key in gemini_keys) |
|
|
|
|
|
for key, validate in validation_rules.items(): |
|
if(key in openai_settings and not validate(openai_settings[key])): |
|
raise ValueError(f"Invalid value for {key}") |
|
elif(key in gemini_settings and not validate(gemini_settings[key])): |
|
raise ValueError(f"Invalid value for {key}") |
|
|
|
|
|
openai_settings["openai_stop"] = None |
|
openai_settings["openai_logit_bias"] = None |
|
openai_settings["openai_stream"] = False |
|
|
|
gemini_settings["gemini_stop_sequences"] = None |
|
gemini_settings["gemini_stream"] = False |
|
|
|
|
|
openai_settings["openai_n"] = 1 |
|
|
|
gemini_settings["gemini_candidate_count"] = 1 |
|
|
|
except Exception as e: |
|
Logger.log_action("Kijiku Rules.json is not valid, setting to invalid_placeholder, current:") |
|
Logger.log_action("Reason: " + str(e)) |
|
Logger.log_action(str(JsonHandler.current_kijiku_rules)) |
|
JsonHandler.current_kijiku_rules = FileEnsurer.INVALID_KIJIKU_RULES_PLACEHOLDER |
|
|
|
Logger.log_action("Kijiku Rules.json is valid, current:") |
|
Logger.log_action(str(JsonHandler.current_kijiku_rules)) |
|
|
|
|
|
|
|
@staticmethod |
|
def reset_kijiku_rules_to_default() -> None: |
|
|
|
""" |
|
|
|
Resets the kijiku_rules json to default. |
|
|
|
""" |
|
|
|
JsonHandler.current_kijiku_rules = FileEnsurer.DEFAULT_KIJIKU_RULES |
|
|
|
JsonHandler.dump_kijiku_rules() |
|
|
|
JsonHandler.load_kijiku_rules() |
|
|
|
|
|
|
|
@staticmethod |
|
def dump_kijiku_rules() -> None: |
|
|
|
""" |
|
|
|
Dumps the Kijiku Rules.json file to disk. |
|
|
|
""" |
|
|
|
with open(FileEnsurer.config_kijiku_rules_path, 'w+', encoding='utf-8') as file: |
|
json.dump(JsonHandler.current_kijiku_rules, file) |
|
|
|
|
|
|
|
@staticmethod |
|
def load_kijiku_rules() -> None: |
|
|
|
""" |
|
|
|
Loads the Kijiku Rules.json file into memory. |
|
|
|
""" |
|
|
|
with open(FileEnsurer.config_kijiku_rules_path, 'r', encoding='utf-8') as file: |
|
JsonHandler.current_kijiku_rules = json.load(file) |
|
|
|
|
|
|
|
@staticmethod |
|
def print_kijiku_rules(output:bool=False) -> None: |
|
|
|
""" |
|
|
|
Prints the Kijiku Rules.json file to the log. |
|
Logs by default, but can be set to print to console as well. |
|
|
|
Parameters: |
|
output (bool | optional | default=False) : Whether to print to console as well. |
|
|
|
""" |
|
|
|
print("-------------------") |
|
print("Base Kijiku Settings:") |
|
print("-------------------") |
|
|
|
for key,value in JsonHandler.current_kijiku_rules["base kijiku settings"].items(): |
|
Logger.log_action(key + " : " + str(value), output=output, omit_timestamp=output) |
|
|
|
print("-------------------") |
|
print("Open AI Settings:") |
|
print("-------------------") |
|
|
|
for key,value in JsonHandler.current_kijiku_rules["openai settings"].items(): |
|
Logger.log_action(key + " : " + str(value), output=output, omit_timestamp=output) |
|
|
|
print("-------------------") |
|
print("Gemini Settings:") |
|
print("-------------------") |
|
|
|
for key,value in JsonHandler.current_kijiku_rules["gemini settings"].items(): |
|
Logger.log_action(key + " : " + str(value), output=output, omit_timestamp=output) |
|
|
|
|
|
|
|
|
|
@staticmethod |
|
def change_kijiku_settings() -> None: |
|
|
|
""" |
|
|
|
Allows the user to change the settings of the Kijiku Rules.json file |
|
|
|
""" |
|
|
|
while(True): |
|
|
|
Toolkit.clear_console() |
|
|
|
settings_print_message = JsonHandler.kijiku_settings_message + SettingsChanger.generate_settings_change_menu() |
|
|
|
action = input(settings_print_message).lower() |
|
|
|
if(action == 'q'): |
|
break |
|
|
|
|
|
if(action == "c"): |
|
SettingsChanger.load_custom_json() |
|
|
|
elif(action == "d"): |
|
print("Resetting to default settings.") |
|
JsonHandler.reset_kijiku_rules_to_default() |
|
|
|
elif(action in JsonHandler.current_kijiku_rules["base kijiku settings"]): |
|
SettingsChanger.change_setting("base kijiku settings", action) |
|
|
|
elif(action in JsonHandler.current_kijiku_rules["openai settings"]): |
|
SettingsChanger.change_setting("openai settings", action) |
|
|
|
elif(action in JsonHandler.current_kijiku_rules["gemini settings"]): |
|
SettingsChanger.change_setting("gemini settings", action) |
|
|
|
else: |
|
print("Invalid setting name. Please try again.") |
|
|
|
Toolkit.pause_console("\nPress enter to continue.") |
|
|
|
JsonHandler.dump_kijiku_rules() |
|
|
|
|
|
|
|
@staticmethod |
|
def convert_to_correct_type(setting_name:str, initial_value:str) -> typing.Any: |
|
|
|
""" |
|
|
|
Converts the input string to the correct type based on the setting name. |
|
|
|
Parameters: |
|
setting_name (str) : The name of the setting to convert. |
|
initial_value (str) : The initial value to convert. |
|
|
|
Returns: |
|
(typing.Any) : The converted value. |
|
|
|
""" |
|
|
|
value = initial_value |
|
|
|
type_expectations = { |
|
"prompt_assembly_mode": {"type": int, "constraints": lambda x: 1 <= x <= 2}, |
|
"number_of_lines_per_batch": {"type": int, "constraints": lambda x: x > 0}, |
|
"sentence_fragmenter_mode": {"type": int, "constraints": lambda x: 1 <= x <= 2}, |
|
"je_check_mode": {"type": int, "constraints": lambda x: 1 <= x <= 2}, |
|
"number_of_malformed_batch_retries": {"type": int, "constraints": lambda x: x >= 0}, |
|
"batch_retry_timeout": {"type": int, "constraints": lambda x: x >= 0}, |
|
"number_of_concurrent_batches": {"type": int, "constraints": lambda x: x >= 0}, |
|
"openai_model": {"type": str, "constraints": lambda x: x in FileEnsurer.ALLOWED_OPENAI_MODELS}, |
|
"openai_system_message": {"type": str, "constraints": lambda x: x not in ["", "None", None]}, |
|
"openai_temperature": {"type": float, "constraints": lambda x: 0 <= x <= 2}, |
|
"openai_top_p": {"type": float, "constraints": lambda x: 0 <= x <= 2}, |
|
"openai_n": {"type": int, "constraints": lambda x: x == 1}, |
|
"openai_stream": {"type": bool, "constraints": lambda x: x is False}, |
|
"openai_stop": {"type": None, "constraints": lambda x: x is None}, |
|
"openai_logit_bias": {"type": None, "constraints": lambda x: x is None}, |
|
"openai_max_tokens": {"type": int, "constraints": lambda x: x is None or isinstance(x, int)}, |
|
"openai_presence_penalty": {"type": float, "constraints": lambda x: -2 <= x <= 2}, |
|
"openai_frequency_penalty": {"type": float, "constraints": lambda x: -2 <= x <= 2}, |
|
"gemini_model": {"type": str, "constraints": lambda x: x in FileEnsurer.ALLOWED_GEMINI_MODELS}, |
|
"gemini_prompt": {"type": str, "constraints": lambda x: x not in ["", "None", None]}, |
|
"gemini_temperature": {"type": float, "constraints": lambda x: 0 <= x <= 2}, |
|
"gemini_top_p": {"type": float, "constraints": lambda x: x is None or (isinstance(x, float) and 0 <= x <= 2)}, |
|
"gemini_top_k": {"type": int, "constraints": lambda x: x is None or x >= 0}, |
|
"gemini_candidate_count": {"type": int, "constraints": lambda x: x == 1}, |
|
"gemini_stream": {"type": bool, "constraints": lambda x: x is False}, |
|
"gemini_stop_sequences": {"type": None, "constraints": lambda x: x is None}, |
|
"gemini_max_output_tokens": {"type": int, "constraints": lambda x: x is None or isinstance(x, int)}, |
|
} |
|
|
|
if(setting_name not in type_expectations): |
|
raise ValueError("Invalid setting name") |
|
|
|
setting_info = type_expectations[setting_name] |
|
|
|
if("stream" in setting_name): |
|
value = Toolkit.string_to_bool(initial_value) |
|
|
|
elif(initial_value.lower() in ["none","null"]): |
|
value = None |
|
|
|
if(setting_info["type"] is None): |
|
converted_value = None |
|
|
|
elif(setting_info["type"] == int) or (setting_info["type"] == float): |
|
|
|
if(value is None or value is ''): |
|
converted_value = None |
|
|
|
elif(setting_info["type"] == int): |
|
converted_value = int(value) |
|
|
|
else: |
|
converted_value = float(value) |
|
|
|
else: |
|
converted_value = setting_info["type"](value) |
|
|
|
if("constraints" in setting_info and not setting_info["constraints"](converted_value)): |
|
raise ValueError(f"{setting_name} out of range") |
|
|
|
return converted_value |
|
|
|
|
|
|
|
class SettingsChanger: |
|
|
|
""" |
|
|
|
Handles changing the settings of the Kijiku Rules.json file. |
|
|
|
""" |
|
|
|
|
|
|
|
@staticmethod |
|
def generate_settings_change_menu() -> str: |
|
|
|
""" |
|
|
|
Generates the settings change menu. |
|
|
|
Returns: |
|
menu (str) : The settings change menu. |
|
|
|
""" |
|
|
|
menu = """ |
|
Current settings: |
|
---------------------------------------------------------------- |
|
|
|
""" |
|
|
|
for key,value in JsonHandler.current_kijiku_rules["base kijiku settings"].items(): |
|
menu += key + " : " + str(value) + "\n" |
|
|
|
print("\n") |
|
|
|
for key,value in JsonHandler.current_kijiku_rules["openai settings"].items(): |
|
menu += key + " : " + str(value) + "\n" |
|
|
|
print("\n") |
|
|
|
for key,value in JsonHandler.current_kijiku_rules["gemini settings"].items(): |
|
menu += key + " : " + str(value) + "\n" |
|
|
|
menu += """ |
|
It is recommended that you maximize the console window for this. You will have to to see the settings above. |
|
|
|
Enter the name of the setting you want to change, type d to reset to default, type c to load an external/custom json directly, or type 'q' to quit settings change : |
|
""" |
|
|
|
return menu |
|
|
|
|
|
|
|
@staticmethod |
|
def load_custom_json() -> None: |
|
|
|
""" |
|
|
|
Loads a custom json into the Kijiku Rules.json file. |
|
|
|
""" |
|
|
|
Toolkit.clear_console() |
|
|
|
|
|
old_kijiku_rules = JsonHandler.current_kijiku_rules |
|
|
|
try: |
|
|
|
|
|
with open(FileEnsurer.external_kijiku_rules_path, 'r', encoding='utf-8') as file: |
|
JsonHandler.current_kijiku_rules = json.load(file) |
|
|
|
JsonHandler.validate_json() |
|
|
|
|
|
assert JsonHandler.current_kijiku_rules != FileEnsurer.INVALID_KIJIKU_RULES_PLACEHOLDER |
|
|
|
JsonHandler.dump_kijiku_rules() |
|
|
|
print("Settings loaded successfully.") |
|
|
|
except AssertionError: |
|
print("Invalid JSON file. Please try again.") |
|
JsonHandler.current_kijiku_rules = old_kijiku_rules |
|
|
|
except FileNotFoundError: |
|
print("Missing JSON file. Make sure you have a json in the same directory as kudasai.py and that the json is named \"kijiku_rules.json\". Please try again.") |
|
JsonHandler.current_kijiku_rules = old_kijiku_rules |
|
|
|
|
|
|
|
@staticmethod |
|
def change_setting(setting_area:str, setting_name:str) -> None: |
|
|
|
""" |
|
|
|
Changes the setting of the Kijiku Rules.json file. |
|
|
|
Parameters: |
|
setting_area (str) : The area of the setting to change. |
|
setting_name (str) : The name of the setting to change. |
|
|
|
""" |
|
|
|
new_value = input(f"\nEnter a new value for {setting_name}: ") |
|
|
|
try: |
|
converted_value = JsonHandler.convert_to_correct_type(setting_name, new_value) |
|
|
|
JsonHandler.current_kijiku_rules[setting_area][setting_name] = converted_value |
|
print(f"Updated {setting_name} to {converted_value}.") |
|
|
|
except Exception as e: |
|
print(f"Invalid input. No changes made. {e}") |