🚀 feat: GPT-4.5, Anthropic Tool Header, and OpenAPI Ref Resolution (#6118)

* 🔧 refactor: Update settings to use 'as const' for improved type safety and make gpt-4o-mini default model (cheapest)

* 📖 docs: Update README to reflect support for GPT-4.5 in image analysis feature

* 🔧 refactor: Update model handling to use default settings and improve encoding logic

* 🔧 refactor: Enhance model version extraction logic for improved compatibility with future GPT and omni models

* feat: GPT-4.5 tx/token update, vision support

* fix: $ref resolution logic in OpenAPI handling

* feat: add new 'anthropic-beta' header for Claude 3.7 to include token-efficient tools; ref: https://docs.anthropic.com/en/docs/build-with-claude/tool-use/token-efficient-tool-use

api/app/clients/OpenAIClient.js

@@ -298,7 +298,9 @@ class OpenAIClient extends BaseClient {
   }
 
   getEncoding() {
-    return this.model?.includes('gpt-4o') ? 'o200k_base' : 'cl100k_base';
+    return this.modelOptions?.model && /gpt-4[^-\s]/.test(this.modelOptions.model)
+      ? 'o200k_base'
+      : 'cl100k_base';
   }
 
   /**
@@ -605,7 +607,7 @@ class OpenAIClient extends BaseClient {
   }
 
   initializeLLM({
-    model = 'gpt-4o-mini',
+    model = openAISettings.model.default,
     modelName,
     temperature = 0.2,
     max_tokens,
@@ -706,7 +708,7 @@ class OpenAIClient extends BaseClient {
 
     const { OPENAI_TITLE_MODEL } = process.env ?? {};
 
-    let model = this.options.titleModel ?? OPENAI_TITLE_MODEL ?? 'gpt-4o-mini';
+    let model = this.options.titleModel ?? OPENAI_TITLE_MODEL ?? openAISettings.model.default;
     if (model === Constants.CURRENT_MODEL) {
       model = this.modelOptions.model;
     }
@@ -899,7 +901,7 @@ ${convo}
     let prompt;
 
     // TODO: remove the gpt fallback and make it specific to endpoint
-    const { OPENAI_SUMMARY_MODEL = 'gpt-4o-mini' } = process.env ?? {};
+    const { OPENAI_SUMMARY_MODEL = openAISettings.model.default } = process.env ?? {};
     let model = this.options.summaryModel ?? OPENAI_SUMMARY_MODEL;
     if (model === Constants.CURRENT_MODEL) {
       model = this.modelOptions.model;

api/models/tx.js

@@ -79,6 +79,7 @@ const tokenValues = Object.assign(
     'o1-mini': { prompt: 1.1, completion: 4.4 },
     'o1-preview': { prompt: 15, completion: 60 },
     o1: { prompt: 15, completion: 60 },
+    'gpt-4.5': { prompt: 75, completion: 150 },
     'gpt-4o-mini': { prompt: 0.15, completion: 0.6 },
     'gpt-4o': { prompt: 2.5, completion: 10 },
     'gpt-4o-2024-05-13': { prompt: 5, completion: 15 },
@@ -167,6 +168,8 @@ const getValueKey = (model, endpoint) => {
     return 'o1-mini';
   } else if (modelName.includes('o1')) {
     return 'o1';
+  } else if (modelName.includes('gpt-4.5')) {
+    return 'gpt-4.5';
   } else if (modelName.includes('gpt-4o-2024-05-13')) {
     return 'gpt-4o-2024-05-13';
   } else if (modelName.includes('gpt-4o-mini')) {

api/models/tx.spec.js

@@ -50,6 +50,16 @@ describe('getValueKey', () => {
     expect(getValueKey('gpt-4-0125')).toBe('gpt-4-1106');
   });
 
+  it('should return "gpt-4.5" for model type of "gpt-4.5"', () => {
+    expect(getValueKey('gpt-4.5-preview')).toBe('gpt-4.5');
+    expect(getValueKey('gpt-4.5-2024-08-06')).toBe('gpt-4.5');
+    expect(getValueKey('gpt-4.5-2024-08-06-0718')).toBe('gpt-4.5');
+    expect(getValueKey('openai/gpt-4.5')).toBe('gpt-4.5');
+    expect(getValueKey('openai/gpt-4.5-2024-08-06')).toBe('gpt-4.5');
+    expect(getValueKey('gpt-4.5-turbo')).toBe('gpt-4.5');
+    expect(getValueKey('gpt-4.5-0125')).toBe('gpt-4.5');
+  });
+
   it('should return "gpt-4o" for model type of "gpt-4o"', () => {
     expect(getValueKey('gpt-4o-2024-08-06')).toBe('gpt-4o');
     expect(getValueKey('gpt-4o-2024-08-06-0718')).toBe('gpt-4o');

api/server/services/Endpoints/anthropic/helpers.js

@@ -48,7 +48,8 @@ function getClaudeHeaders(model, supportsCacheControl) {
     };
   } else if (/claude-3[-.]7/.test(model)) {
     return {
-      'anthropic-beta': 'output-128k-2025-02-19,prompt-caching-2024-07-31',
+      'anthropic-beta':
+        'token-efficient-tools-2025-02-19,output-128k-2025-02-19,prompt-caching-2024-07-31',
     };
   } else {
     return {

api/utils/tokens.js

@@ -13,6 +13,7 @@ const openAIModels = {
   'gpt-4-32k-0613': 32758, // -10 from max
   'gpt-4-1106': 127500, // -500 from max
   'gpt-4-0125': 127500, // -500 from max
+  'gpt-4.5': 127500, // -500 from max
   'gpt-4o': 127500, // -500 from max
   'gpt-4o-mini': 127500, // -500 from max
   'gpt-4o-2024-05-13': 127500, // -500 from max

api/utils/tokens.spec.js

@@ -103,6 +103,16 @@ describe('getModelMaxTokens', () => {
     );
   });
 
+  test('should return correct tokens for gpt-4.5 matches', () => {
+    expect(getModelMaxTokens('gpt-4.5')).toBe(maxTokensMap[EModelEndpoint.openAI]['gpt-4.5']);
+    expect(getModelMaxTokens('gpt-4.5-preview')).toBe(
+      maxTokensMap[EModelEndpoint.openAI]['gpt-4.5'],
+    );
+    expect(getModelMaxTokens('openai/gpt-4.5-preview')).toBe(
+      maxTokensMap[EModelEndpoint.openAI]['gpt-4.5'],
+    );
+  });
+
   test('should return correct tokens for Anthropic models', () => {
     const models = [
       'claude-2.1',