Add OpenAI o1

6d79cc015ab8e9049a4b09b0d197c30aa709ef09

Cohee <18619528+Cohee1207@users.noreply.github.com>

Parents: 854541f4ba58ccca1f5513c9465031d7a3781566
6 files changed, +48 -10Ignore whitespace
package-lock.json+5 -4
@@ -46,7 +46,7 @@
46 "sanitize-filename": "^1.6.3",46 "sanitize-filename": "^1.6.3",
47 "sillytavern-transformers": "2.14.6",47 "sillytavern-transformers": "2.14.6",
48 "simple-git": "^3.19.1",48 "simple-git": "^3.19.1",
49 "tiktoken": "^1.0.15",49 "tiktoken": "^1.0.16",
50 "vectra": "^0.2.2",50 "vectra": "^0.2.2",
51 "wavefile": "^11.0.0",51 "wavefile": "^11.0.0",
52 "write-file-atomic": "^5.0.1",52 "write-file-atomic": "^5.0.1",
@@ -5751,9 +5751,10 @@
5751 "license": "MIT"5751 "license": "MIT"
5752 },5752 },
5753 "node_modules/tiktoken": {5753 "node_modules/tiktoken": {
5754 "version": "1.0.15",5754 "version": "1.0.16",
5755 "resolved": "https://registry.npmjs.org/tiktoken/-/tiktoken-1.0.15.tgz",5755 "resolved": "https://registry.npmjs.org/tiktoken/-/tiktoken-1.0.16.tgz",
5756 "integrity": "sha512-sCsrq/vMWUSEW29CJLNmPvWxlVp7yh2tlkAjpJltIKqp5CKf98ZNpdeHRmAlPVFlGEbswDc6SmI8vz64W/qErw=="5756 "integrity": "sha512-hRcORIGF2YlAgWx3nzrGJOrKSJwLoc81HpXmMQk89632XAgURc7IeV2FgQ2iXo9z/J96fCvpsHg2kWoHcbj9fg==",
5757 "license": "MIT"
5757 },5758 },
5758 "node_modules/timm": {5759 "node_modules/timm": {
5759 "version": "1.7.1",5760 "version": "1.7.1",
package.json+1 -1
@@ -36,7 +36,7 @@
36 "sanitize-filename": "^1.6.3",36 "sanitize-filename": "^1.6.3",
37 "sillytavern-transformers": "2.14.6",37 "sillytavern-transformers": "2.14.6",
38 "simple-git": "^3.19.1",38 "simple-git": "^3.19.1",
39 "tiktoken": "^1.0.15",39 "tiktoken": "^1.0.16",
40 "vectra": "^0.2.2",40 "vectra": "^0.2.2",
41 "wavefile": "^11.0.0",41 "wavefile": "^11.0.0",
42 "write-file-atomic": "^5.0.1",42 "write-file-atomic": "^5.0.1",
public/index.html+5 -1
@@ -383,7 +383,7 @@
383 Max Response Length (tokens)383 Max Response Length (tokens)
384 </div>384 </div>
385 <div class="wide100p">385 <div class="wide100p">
386 <input type="number" id="openai_max_tokens" name="openai_max_tokens" class="text_pole" min="1" max="16384">386 <input type="number" id="openai_max_tokens" name="openai_max_tokens" class="text_pole" min="1" max="65536 ">
387 </div>387 </div>
388 </div>388 </div>
389 <div class="range-block" data-source="openai,custom">389 <div class="range-block" data-source="openai,custom">
@@ -2611,6 +2611,10 @@
2611 <option value="gpt-4-0125-preview">gpt-4-0125-preview (2024)</option>2611 <option value="gpt-4-0125-preview">gpt-4-0125-preview (2024)</option>
2612 <option value="gpt-4-1106-preview">gpt-4-1106-preview (2023)</option>2612 <option value="gpt-4-1106-preview">gpt-4-1106-preview (2023)</option>
2613 </optgroup>2613 </optgroup>
2614 <optgroup label="o1">
2615 <option value="o1-preview">o1-preview</option>
2616 <option value="o1-mini">o1-mini</option>
2617 </optgroup>
2614 <optgroup label="Other">2618 <optgroup label="Other">
2615 <option value="text-davinci-003">text-davinci-003</option>2619 <option value="text-davinci-003">text-davinci-003</option>
2616 <option value="text-davinci-002">text-davinci-002</option>2620 <option value="text-davinci-002">text-davinci-002</option>
public/script.js+6 -1
@@ -2862,7 +2862,12 @@ export function getCharacterCardFields() {
28622862
2863export function isStreamingEnabled() {2863export function isStreamingEnabled() {
2864 const noStreamSources = [chat_completion_sources.SCALE];2864 const noStreamSources = [chat_completion_sources.SCALE];
2865 return ((main_api == 'openai' && oai_settings.stream_openai && !noStreamSources.includes(oai_settings.chat_completion_source) && !(oai_settings.chat_completion_source == chat_completion_sources.MAKERSUITE && oai_settings.google_model.includes('bison')))2865 return (
2866 (main_api == 'openai' &&
2867 oai_settings.stream_openai &&
2868 !noStreamSources.includes(oai_settings.chat_completion_source) &&
2869 !(oai_settings.chat_completion_source == chat_completion_sources.OPENAI && oai_settings.openai_model.startsWith('o1-')) &&
2870 !(oai_settings.chat_completion_source == chat_completion_sources.MAKERSUITE && oai_settings.google_model.includes('bison')))
2866 || (main_api == 'kobold' && kai_settings.streaming_kobold && kai_flags.can_use_streaming)2871 || (main_api == 'kobold' && kai_settings.streaming_kobold && kai_flags.can_use_streaming)
2867 || (main_api == 'novel' && nai_settings.streaming_novel)2872 || (main_api == 'novel' && nai_settings.streaming_novel)
2868 || (main_api == 'textgenerationwebui' && textgen_settings.streaming));2873 || (main_api == 'textgenerationwebui' && textgen_settings.streaming));
public/scripts/openai.js+27 -3
@@ -1797,7 +1797,7 @@ async function sendOpenAIRequest(type, messages, signal) {
1797 const isQuiet = type === 'quiet';1797 const isQuiet = type === 'quiet';
1798 const isImpersonate = type === 'impersonate';1798 const isImpersonate = type === 'impersonate';
1799 const isContinue = type === 'continue';1799 const isContinue = type === 'continue';
1800 const stream = oai_settings.stream_openai && !isQuiet && !isScale && !(isGoogle && oai_settings.google_model.includes('bison'));1800 const stream = oai_settings.stream_openai && !isQuiet && !isScale && !(isGoogle && oai_settings.google_model.includes('bison')) && !(isOAI && oai_settings.openai_model.startsWith('o1-'));
1801 const useLogprobs = !!power_user.request_token_probabilities;1801 const useLogprobs = !!power_user.request_token_probabilities;
1802 const canMultiSwipe = oai_settings.n > 1 && !isContinue && !isImpersonate && !isQuiet && (isOAI || isCustom);1802 const canMultiSwipe = oai_settings.n > 1 && !isContinue && !isImpersonate && !isQuiet && (isOAI || isCustom);
18031803
@@ -1960,12 +1960,33 @@ async function sendOpenAIRequest(type, messages, signal) {
1960 generate_data['seed'] = oai_settings.seed;1960 generate_data['seed'] = oai_settings.seed;
1961 }1961 }
19621962
1963 await eventSource.emit(event_types.CHAT_COMPLETION_SETTINGS_READY, generate_data);
1964
1965 if (isFunctionCallingSupported() && !stream) {1963 if (isFunctionCallingSupported() && !stream) {
1966 await registerFunctionTools(type, generate_data);1964 await registerFunctionTools(type, generate_data);
1967 }1965 }
19681966
1967 if (isOAI && oai_settings.openai_model.startsWith('o1-')) {
1968 generate_data.messages.forEach((msg) => {
1969 if (msg.role === 'system') {
1970 msg.role = 'user';
1971 }
1972 });
1973 delete generate_data.stream;
1974 delete generate_data.logprobs;
1975 delete generate_data.top_logprobs;
1976 delete generate_data.n;
1977 delete generate_data.temperature;
1978 delete generate_data.top_p;
1979 delete generate_data.frequency_penalty;
1980 delete generate_data.presence_penalty;
1981 delete generate_data.tools;
1982 delete generate_data.tool_choice;
1983 // IDK if it supports it and I have no way to test it
1984 // delete generate_data.logit_bias;
1985 // delete generate_data.stop;
1986 }
1987
1988 await eventSource.emit(event_types.CHAT_COMPLETION_SETTINGS_READY, generate_data);
1989
1969 const generate_url = '/api/backends/chat-completions/generate';1990 const generate_url = '/api/backends/chat-completions/generate';
1970 const response = await fetch(generate_url, {1991 const response = await fetch(generate_url, {
1971 method: 'POST',1992 method: 'POST',
@@ -3905,6 +3926,9 @@ function getMaxContextOpenAI(value) {
3905 if (oai_settings.max_context_unlocked) {3926 if (oai_settings.max_context_unlocked) {
3906 return unlocked_max;3927 return unlocked_max;
3907 }3928 }
3929 else if (value.startsWith('o1-')) {
3930 return max_128k;
3931 }
3908 else if (value.includes('chatgpt-4o-latest') || value.includes('gpt-4-turbo') || value.includes('gpt-4o') || value.includes('gpt-4-1106') || value.includes('gpt-4-0125') || value.includes('gpt-4-vision')) {3932 else if (value.includes('chatgpt-4o-latest') || value.includes('gpt-4-turbo') || value.includes('gpt-4o') || value.includes('gpt-4-1106') || value.includes('gpt-4-0125') || value.includes('gpt-4-vision')) {
3909 return max_128k;3933 return max_128k;
3910 }3934 }
src/endpoints/tokenizers.js+4 -0
@@ -350,6 +350,10 @@ function getWebTokenizersChunks(tokenizer, ids) {
350 * @returns {string} Tokenizer model to use350 * @returns {string} Tokenizer model to use
351 */351 */
352function getTokenizerModel(requestModel) {352function getTokenizerModel(requestModel) {
353 if (requestModel.includes('o1-preview') || requestModel.includes('o1-mini')) {
354 return 'gpt-4o';
355 }
356
353 if (requestModel.includes('gpt-4o')) {357 if (requestModel.includes('gpt-4o')) {
354 return 'gpt-4o';358 return 'gpt-4o';
355 }359 }