anthropic captioning

This commit is contained in:
based
2024-03-05 07:07:38 +10:00
parent cbda133d75
commit 64d9c9dc5d
6 changed files with 83 additions and 4 deletions

View File

@@ -284,6 +284,7 @@ jQuery(function () {
(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 === 'google' && secret_state[SECRET_KEYS.MAKERSUITE]) ||
(extension_settings.caption.source === 'multimodal' && extension_settings.caption.multimodal_api === 'anthropic' && secret_state[SECRET_KEYS.CLAUDE]) ||
(extension_settings.caption.source === 'multimodal' && extension_settings.caption.multimodal_api === 'ollama' && textgenerationwebui_settings.server_urls[textgen_types.OLLAMA]) ||
(extension_settings.caption.source === 'multimodal' && extension_settings.caption.multimodal_api === 'llamacpp' && textgenerationwebui_settings.server_urls[textgen_types.LLAMACPP]) ||
(extension_settings.caption.source === 'multimodal' && extension_settings.caption.multimodal_api === 'ooba' && textgenerationwebui_settings.server_urls[textgen_types.OOBA]) ||
@@ -343,7 +344,7 @@ jQuery(function () {
<label for="caption_source">Source</label>
<select id="caption_source" class="text_pole">
<option value="local">Local</option>
<option value="multimodal">Multimodal (OpenAI / llama / Google)</option>
<option value="multimodal">Multimodal (OpenAI / Anthropic / llama / Google)</option>
<option value="extras">Extras</option>
<option value="horde">Horde</option>
</select>
@@ -355,6 +356,7 @@ jQuery(function () {
<option value="ooba">Text Generation WebUI (oobabooga)</option>
<option value="ollama">Ollama</option>
<option value="openai">OpenAI</option>
<option value="anthropic">Anthropic</option>
<option value="openrouter">OpenRouter</option>
<option value="google">Google MakerSuite</option>
<option value="custom">Custom (OpenAI-compatible)</option>
@@ -364,6 +366,8 @@ jQuery(function () {
<label for="caption_multimodal_model">Model</label>
<select id="caption_multimodal_model" class="flex1 text_pole">
<option data-type="openai" value="gpt-4-vision-preview">gpt-4-vision-preview</option>
<option data-type="anthropic" value="claude-3-opus-20240229">claude-3-opus-20240229</option>
<option data-type="anthropic" value="claude-3-sonnet-20240229">claude-3-sonnet-20240229</option>
<option data-type="google" value="gemini-pro-vision">gemini-pro-vision</option>
<option data-type="openrouter" value="openai/gpt-4-vision-preview">openai/gpt-4-vision-preview</option>
<option data-type="openrouter" value="haotian-liu/llava-13b">haotian-liu/llava-13b</option>

View File

@@ -23,6 +23,7 @@ export async function getMultimodalCaption(base64Img, prompt) {
// OpenRouter has a payload limit of ~2MB. Google is 4MB, but we love democracy.
// Ooba requires all images to be JPEGs.
const isGoogle = extension_settings.caption.multimodal_api === 'google';
const isClaude = extension_settings.caption.multimodal_api === 'anthropic';
const isOllama = extension_settings.caption.multimodal_api === 'ollama';
const isLlamaCpp = extension_settings.caption.multimodal_api === 'llamacpp';
const isCustom = extension_settings.caption.multimodal_api === 'custom';
@@ -39,7 +40,7 @@ export async function getMultimodalCaption(base64Img, prompt) {
}
const useReverseProxy =
extension_settings.caption.multimodal_api === 'openai'
(extension_settings.caption.multimodal_api === 'openai' || extension_settings.caption.multimodal_api === 'anthropic')
&& extension_settings.caption.allow_reverse_proxy
&& oai_settings.reverse_proxy
&& isValidUrl(oai_settings.reverse_proxy);
@@ -87,6 +88,8 @@ export async function getMultimodalCaption(base64Img, prompt) {
switch (extension_settings.caption.multimodal_api) {
case 'google':
return '/api/google/caption-image';
case 'anthropic':
return '/api/anthropic/caption-image';
case 'llamacpp':
return '/api/backends/text-completions/llamacpp/caption-image';
case 'ollama':