Chutes integration (#4844)

* Chutes integration * Fix eslint * Fix key saving * Fix logo coloration * Fix tool checks * Unhide image inlining controls * Fix order of options * Fix type use in TTS extension script * Add Chutes as a vector storage source * Change log levels to debug * Fix streamed reasoning parsing * Skip remote models update * TTS: Fix API key highlight * Sort image models A-Z * TTS: Fixes * Remove unused SD endpoint * Skip setting context size if models list is not yet loaded * remove chutes quota / balance * Fix: streamed tool calling * Hide reasoning effort control * Add image request debug log * Fix: scroll down on media load in extensions * Unhide some samplers * Bring back reasoning effort * This code will never execute * Reformat else if cases * Add stop strings to request * Remove conditional from reasoning_effort body param * Preserve original pricing fields * Unhide logit bias setting * Pass repetition penalty and logit bias to backend * Swap llama tokenizer for llama3 * Pass min_p, remove supported_sampling_parameters checks * Enable logprobs --------- Co-authored-by: Cohee <18619528+Cohee1207@users.noreply.github.com>
2025-11-30 17:17:49 -05:00
parent dc06abb364
commit 55a07d445d
28 changed files with 1049 additions and 33 deletions
@@ -699,6 +699,26 @@ export function getTokenizerModel() {
        }
    }

+    if (oai_settings.chat_completion_source == chat_completion_sources.CHUTES && oai_settings.chutes_model) {
+        const model = oai_settings.chutes_model.toLowerCase();
+
+        if (model.includes('deepseek') || model.includes('mai-ds')) {
+            return deepseekTokenizer;
+        } else if (model.includes('qwen') || model.includes('qwq') || model.includes('tongyi') || model.includes('kimi')) {
+            return qwen2Tokenizer;
+        } else if (model.includes('llama') || model.includes('longcat') || model.includes('hermes')) {
+            return llama3Tokenizer;
+        } else if (model.includes('gemma')) {
+            return gemmaTokenizer;
+        } else if (model.includes('nemo')) {
+            return nemoTokenizer;
+        } else if (model.includes('mistral')) {
+            return mistralTokenizer;
+        } else if (model.includes('gpt-oss')) {
+            return gpt4oTokenizer;
+        }
+    }
+
    if (oai_settings.chat_completion_source == chat_completion_sources.COHERE) {
        if (oai_settings.cohere_model.includes('command-a')) {
            return commandATokenizer;
@@ -1206,3 +1226,4 @@ export async function initTokenizers() {
    await loadTokenCache();
    registerDebugFunction('resetTokenCache', 'Reset token cache', 'Purges the calculated token counts. Use this if you want to force a full re-tokenization of all chats or suspect the token counts are wrong.', resetTokenCache);
 }
+