Add multimodal captioning for 01.ai

This commit is contained in:
Cohee
2024-08-01 01:34:49 +03:00
parent 7b9eb97c7f
commit 40ee236ca8
4 changed files with 17 additions and 3 deletions

View File

@ -8,13 +8,12 @@ import { textgen_types, textgenerationwebui_settings } from '../../textgen-setti
import { SlashCommandParser } from '../../slash-commands/SlashCommandParser.js';
import { SlashCommand } from '../../slash-commands/SlashCommand.js';
import { ARGUMENT_TYPE, SlashCommandArgument, SlashCommandNamedArgument } from '../../slash-commands/SlashCommandArgument.js';
import { SlashCommandEnumValue } from '../../slash-commands/SlashCommandEnumValue.js';
import { commonEnumProviders } from '../../slash-commands/SlashCommandCommonEnumsProvider.js';
export { MODULE_NAME };
const MODULE_NAME = 'caption';
const PROMPT_DEFAULT = 'Whats in this image?';
const PROMPT_DEFAULT = 'What\'s in this image?';
const TEMPLATE_DEFAULT = '[{{user}} sends {{char}} a picture that contains: {{caption}}]';
/**
@ -334,7 +333,7 @@ async function getCaptionForFile(file, prompt, quiet) {
}
catch (error) {
const errorMessage = error.message || 'Unknown error';
toastr.error(errorMessage, "Failed to caption image.");
toastr.error(errorMessage, 'Failed to caption image.');
console.error(error);
return '';
}
@ -399,6 +398,7 @@ jQuery(async function () {
(modules.includes('caption') && extension_settings.caption.source === 'extras') ||
(extension_settings.caption.source === 'multimodal' && extension_settings.caption.multimodal_api === 'openai' && (secret_state[SECRET_KEYS.OPENAI] || extension_settings.caption.allow_reverse_proxy)) ||
(extension_settings.caption.source === 'multimodal' && extension_settings.caption.multimodal_api === 'openrouter' && secret_state[SECRET_KEYS.OPENROUTER]) ||
(extension_settings.caption.source === 'multimodal' && extension_settings.caption.multimodal_api === 'zerooneai' && secret_state[SECRET_KEYS.ZEROONEAI]) ||
(extension_settings.caption.source === 'multimodal' && extension_settings.caption.multimodal_api === 'google' && (secret_state[SECRET_KEYS.MAKERSUITE] || extension_settings.caption.allow_reverse_proxy)) ||
(extension_settings.caption.source === 'multimodal' && extension_settings.caption.multimodal_api === 'anthropic' && (secret_state[SECRET_KEYS.CLAUDE] || extension_settings.caption.allow_reverse_proxy)) ||
(extension_settings.caption.source === 'multimodal' && extension_settings.caption.multimodal_api === 'ollama' && textgenerationwebui_settings.server_urls[textgen_types.OLLAMA]) ||