SillyTavern/public/scripts/extensions/vectors/settings.html

194 lines
10 KiB
HTML

<div class="vectors_settings">
<div class="inline-drawer">
<div class="inline-drawer-toggle inline-drawer-header">
<b>Vector Storage</b>
<div class="inline-drawer-icon fa-solid fa-circle-chevron-down down"></div>
</div>
<div class="inline-drawer-content">
<div class="flex-container flexFlowColumn">
<label for="vectors_source">
Vectorization Source
</label>
<select id="vectors_source" class="text_pole">
<option value="transformers">Local (Transformers)</option>
<option value="extras">Extras</option>
<option value="openai">OpenAI</option>
<option value="palm">Google MakerSuite (PaLM)</option>
<option value="mistral">MistralAI</option>
<option value="togetherai">TogetherAI</option>
<option value="nomicai">NomicAI</option>
</select>
</div>
<div class="flex-container flexFlowColumn" id="openai_vectorsModel">
<label for="vectors_openai_model">
Vectorization Model
</label>
<select id="vectors_openai_model" class="text_pole">
<option value="text-embedding-ada-002">text-embedding-ada-002</option>
<option value="text-embedding-3-small">text-embedding-3-small</option>
<option value="text-embedding-3-large">text-embedding-3-large</option>
</select>
</div>
<div class="flex-container flexFlowColumn" id="together_vectorsModel">
<label for="vectors_togetherai_model">
Vectorization Model
</label>
<select id="vectors_togetherai_model" class="text_pole">
<option value="togethercomputer/m2-bert-80M-32k-retrieval">M2-BERT-Retrieval-32k</option>
<option value="togethercomputer/m2-bert-80M-8k-retrieval">M2-BERT-Retrieval-8k</option>
<option value="togethercomputer/m2-bert-80M-2k-retrieval">M2-BERT-Retrieval-2K</option>
<option value="WhereIsAI/UAE-Large-V1">UAE-Large-V1</option>
<option value="BAAI/bge-large-en-v1.5">BAAI-Bge-Large-1p5</option>
<option value="BAAI/bge-base-en-v1.5">BAAI-Bge-Base-1p5</option>
<option value="sentence-transformers/msmarco-bert-base-dot-v5">Sentence-BERT</option>
<option value="bert-base-uncased">Bert Base Uncased</option>
</select>
</div>
<small id="vectors_modelWarning">
<i class="fa-solid fa-exclamation-triangle"></i>
<span data-i18n="Vectors Model Warning">
It is recommended to purge vectors when changing the model mid-chat. Otherwise, it will lead to sub-par results.
</span>
</small>
<div class="flex-container flexFlowColumn" id="nomicai_apiKey">
<label for="api_key_nomicai">
<span>NomicAI API Key</span>
</label>
<div class="flex-container">
<input id="api_key_nomicai" name="api_key_nomicai" class="text_pole flex1 wide100p" maxlength="500" size="35" type="text" autocomplete="off">
<div title="Clear your API key" class="menu_button fa-solid fa-circle-xmark clear-api-key" data-key="api_key_nomicai">
</div>
</div>
<div data-for="api_key_nomicai" class="neutral_warning" data-i18n="For privacy reasons, your API key will be hidden after you reload the page.">
For privacy reasons, your API key will be hidden after you reload the page.
</div>
</div>
<div class="flex-container flexFlowColumn" title="How many last messages will be matched for relevance.">
<label for="vectors_query">
<span>Query messages</span>
</label>
<input type="number" id="vectors_query" class="text_pole widthUnset" min="1" max="99" />
</div>
<label class="checkbox_label" for="vectors_include_wi" title="Query results can activate World Info entries.">
<input id="vectors_include_wi" type="checkbox" class="checkbox">
Include in World Info Scanning
</label>
<hr>
<h4>
File vectorization settings
</h4>
<label class="checkbox_label" for="vectors_enabled_files">
<input id="vectors_enabled_files" type="checkbox" class="checkbox">
Enabled for files
</label>
<div id="vectors_files_settings">
<div class="flex-container">
<div class="flex1" title="Only files past this size will be vectorized.">
<label for="vectors_size_threshold">
<small>Size threshold (KB)</small>
</label>
<input id="vectors_size_threshold" type="number" class="text_pole widthUnset" min="1" max="99999" />
</div>
<div class="flex1" title="Chunk size for file splitting.">
<label for="vectors_chunk_size">
<small>Chunk size (chars)</small>
</label>
<input id="vectors_chunk_size" type="number" class="text_pole widthUnset" min="1" max="99999" />
</div>
<div class="flex1" title="How many chunks to retrieve when querying.">
<label for="vectors_chunk_count">
<small>Retrieve chunks</small>
</label>
<input id="vectors_chunk_count" type="number" class="text_pole widthUnset" min="1" max="99999" />
</div>
</div>
</div>
<hr>
<h4>
Chat vectorization settings
</h4>
<label class="checkbox_label" for="vectors_enabled_chats">
<input id="vectors_enabled_chats" type="checkbox" class="checkbox">
Enabled for chat messages
</label>
<div id="vectors_chats_settings">
<div id="vectors_advanced_settings">
<label for="vectors_template">
Insertion Template
</label>
<textarea id="vectors_template" class="text_pole textarea_compact" rows="3" placeholder="Use {{text}} macro to specify the position of retrieved text."></textarea>
<label for="vectors_position">Injection Position</label>
<div class="radio_group">
<label>
<input type="radio" name="vectors_position" value="2" />
Before Main Prompt / Story String
</label>
<!--Keep these as 0 and 1 to interface with the setExtensionPrompt function-->
<label>
<input type="radio" name="vectors_position" value="0" />
After Main Prompt / Story String
</label>
<label for="vectors_depth" title="How many messages before the current end of the chat." data-i18n="[title]How many messages before the current end of the chat.">
<input type="radio" name="vectors_position" value="1" />
In-chat @ Depth <input id="vectors_depth" class="text_pole widthUnset" type="number" min="0" max="999" />
</label>
</div>
<div class="flex-container">
<div class="flex1" title="Can increase the retrieval quality for the cost of processing. 0 = disabled.">
<label for="vectors_message_chunk_size">
<small>Chunk size (chars)</small>
</label>
<input id="vectors_message_chunk_size" type="number" class="text_pole widthUnset" min="0" max="9999" />
</div>
<div class="flex1" title="Prevents last N messages from being placed out of order.">
<label for="vectors_protect">
<small>Retain#</small>
</label>
<input type="number" id="vectors_protect" class="text_pole widthUnset" min="1" max="9999" />
</div>
<div class="flex1" title="How many past messages to insert as memories.">
<label for="vectors_insert">
<small>Insert#</small>
</label>
<input type="number" id="vectors_insert" class="text_pole widthUnset" min="1" max="9999" />
</div>
</div>
</div>
<small>
Old messages are vectorized gradually as you chat.
To process all previous messages, click the button below.
</small>
<div class="flex-container">
<div id="vectors_vectorize_all" class="menu_button menu_button_icon">
Vectorize All
</div>
<div id="vectors_purge" class="menu_button menu_button_icon">
Purge Vectors
</div>
<div id="vectors_view_stats" class="menu_button menu_button_icon">
View Stats
</div>
</div>
<div id="vectorize_progress" style="display: none;">
<small>
Processed <span id="vectorize_progress_percent">0</span>% of messages.
ETA: <span id="vectorize_progress_eta">...</span> seconds.
</small>
</div>
</div>
</div>
</div>
</div>