modelmix 4.4.16 → 4.4.18

package/README.md

@@ -136,9 +136,9 @@ Here's a comprehensive list of available methods:
 | Method             | Provider   | Model                          | Price (I/O) per 1 M tokens |
 | ------------------ | ---------- | ------------------------------ | -------------------------- |
 | `gpt54()`          | OpenAI     | gpt-5.4                        | [\$2.50 / \$15.00][1]      |
+| `gpt53codex()`     | OpenAI     | gpt-5.3-codex                  | [\$1.25 / \$14.00][1]      |
 | `gpt52()`          | OpenAI     | gpt-5.2                        | [\$1.75 / \$14.00][1]      |
 | `gpt51()`          | OpenAI     | gpt-5.1                        | [\$1.25 / \$10.00][1]      |
-| `gpt53codex()`     | OpenAI     | gpt-5.3-codex                  | [\$1.25 / \$14.00][1]      |
 | `gpt5mini()`       | OpenAI     | gpt-5-mini                     | [\$0.25 / \$2.00][1]       |
 | `gpt5nano()`       | OpenAI     | gpt-5-nano                     | [\$0.05 / \$0.40][1]       |
 | `gpt41()`          | OpenAI     | gpt-4.1                        | [\$2.00 / \$8.00][1]       |
@@ -406,6 +406,7 @@ Every response from `raw()` now includes a `tokens` object with the following st
     input: 150,    // Number of tokens in the prompt/input
     output: 75,    // Number of tokens in the completion/output
     total: 225,    // Total tokens used (input + output)
+    cached: 100,   // Cached input tokens reported by the provider (0 when absent)
     cost: 0.0012,  // Estimated cost in USD (null if model not in pricing table)
     speed: 42      // Output tokens per second (int)
   }
@@ -419,10 +420,10 @@ After calling `message()` or `json()`, use `lastRaw` to access the complete resp
 ```javascript
 const text = await model.message();
 console.log(model.lastRaw.tokens);
-// { input: 122, output: 86, total: 541, cost: 0.000319, speed: 38 }
+// { input: 122, output: 86, total: 208, cached: 41, cost: 0.000319, speed: 38 }
 ```
 
-The `cost` field is the estimated cost in USD based on the model's pricing per 1M tokens (input/output). If the model is not found in the pricing table, `cost` will be `null`. The `speed` field is the generation speed measured in output tokens per second (integer).
+The `cached` field is a single aggregated count of cached input tokens reported by the provider. The `cost` field is the estimated cost in USD based on the model's pricing per 1M tokens (input/output). If the model is not found in the pricing table, `cost` will be `null`. The `speed` field is the generation speed measured in output tokens per second (integer).
 
 ## 🐛 Enabling Debug Mode
 
@@ -516,7 +517,7 @@ new ModelMix(args = { options: {}, config: {} })
   - `message`: The text response from the model
   - `think`: Reasoning/thinking content (if available)
   - `toolCalls`: Array of tool calls made by the model (if any)
-  - `tokens`: Object with `input`, `output`, `total` token counts, `cost` (USD), and `speed` (output tokens/sec)
+  - `tokens`: Object with `input`, `output`, `total`, and `cached` token counts, plus `cost` (USD) and `speed` (output tokens/sec)
   - `response`: The raw API response
 - `stream(callback)`: Sends the message and streams the response, invoking the callback with each streamed part.
 - `json(schemaExample, descriptions = {}, options = {})`: Forces the model to return a response in a specific JSON format.

package/demo/cache.js

@@ -0,0 +1,52 @@
+import { ModelMix } from '../index.js';
+try { process.loadEnvFile(); } catch {}
+
+console.log("\n" + '--------| gpt54() prompt cache |--------');
+
+// Keep the reusable prefix first and only vary the question at the end.
+const sharedPrefix = [
+    "You are a concise science tutor.",
+    "The repeated block below is intentionally long so OpenAI can reuse cached prompt tokens on the second request.",
+    Array.from({ length: 80 }, (_, index) =>
+        `Reference ${String(index + 1).padStart(3, '0')}: Quantum systems are described with probabilities, measurements collapse possibilities into outcomes, and explanations must stay concrete, brief, and easy to understand.`
+    ).join("\n")
+].join("\n\n");
+
+const buildPrompt = (question) => `${sharedPrefix}\n\nQuestion: ${question}`;
+
+const createModel = () => ModelMix.new({
+    config: {
+        debug: 3,
+    }
+}).gpt54({
+    options: {
+        reasoning_effort: "none",
+        verbosity: "low",
+        prompt_cache_key: "demo-gpt54-prompt-cache",
+        prompt_cache_retention: "24h"
+    }
+});
+
+const runRequest = async (label, question) => {
+    const model = createModel();
+    model.addText(buildPrompt(question));
+
+    const result = await model.raw();
+
+    console.log(`\n${label}`);
+    console.log("message:", result.message);
+    console.log("tokens:", result.tokens);
+
+    return result;
+};
+
+await runRequest(
+    "Request 1 (warms the cache)",
+    "Explain quantum entanglement in simple Spanish in 3 short bullet points."
+);
+
+await runRequest(
+    "Request 2 (reuses the cached prefix)",
+    "Now explain quantum entanglement in simple Spanish with a different analogy and 3 short bullet points."
+);
+

package/index.js

@@ -181,6 +181,15 @@ class ModelMix {
         return (tokens.input * inputPerMillion / 1_000_000) + (tokens.output * outputPerMillion / 1_000_000);
     }
 
+    static extractCacheTokens(usage = {}) {
+        return usage.input_tokens_details?.cached_tokens
+            || usage.prompt_tokens_details?.cached_tokens
+            || usage.cache_read_input_tokens
+            || usage.cachedContentTokenCount
+            || usage.cached_content_token_count
+            || 0;
+    }
+
     static formatInputSummary(messages, system, debug = 2) {
         const lastMessage = messages[messages.length - 1];
         let inputText = '';
@@ -250,15 +259,6 @@ class ModelMix {
     gpt41nano({ options = {}, config = {} } = {}) {
         return this.attach('gpt-4.1-nano', new MixOpenAI({ options, config }));
     }
-    o4mini({ options = {}, config = {} } = {}) {
-        return this.attach('o4-mini', new MixOpenAI({ options, config }));
-    }
-    o3({ options = {}, config = {} } = {}) {
-        return this.attach('o3', new MixOpenAI({ options, config }));
-    }
-    gpt45({ options = {}, config = {} } = {}) {
-        return this.attach('gpt-4.5-preview', new MixOpenAI({ options, config }));
-    }
     gpt5({ options = {}, config = {} } = {}) {
         return this.attach('gpt-5', new MixOpenAI({ options, config }));
     }
@@ -269,10 +269,10 @@ class ModelMix {
         return this.attach('gpt-5-nano', new MixOpenAI({ options, config }));
     }
     gpt51({ options = {}, config = {} } = {}) {
-        return this.attach('gpt-5.1', new MixOpenAI({ options, config }));
+        return this.attach('gpt-5.1', new MixOpenAIResponses({ options, config }));
     }
     gpt52({ options = {}, config = {} } = {}) {
-        return this.attach('gpt-5.2', new MixOpenAI({ options, config }));
+        return this.attach('gpt-5.2', new MixOpenAIResponses({ options, config }));
     }
     gpt54({ options = {}, config = {} } = {}) {
         return this.attach('gpt-5.4', new MixOpenAIResponses({ options, config }));
@@ -289,8 +289,8 @@ class ModelMix {
     gpt53codex({ options = {}, config = {} } = {}) {
         return this.attach('gpt-5.3-codex', new MixOpenAIResponses({ options, config }));
     }          
-    gpt52chat({ options = {}, config = {} } = {}) {
-        return this.attach('gpt-5.2-chat-latest', new MixOpenAI({ options, config }));
+    gpt53chat({ options = {}, config = {} } = {}) {
+        return this.attach('gpt-5.3-chat-latest', new MixOpenAIResponses({ options, config }));
     }
     gptOss({ options = {}, config = {}, mix = {} } = {}) {
         mix = { ...this.mix, ...mix };
@@ -963,7 +963,10 @@ class ModelMix {
                     // debug level 2: Readable summary of output
                     if (currentConfig.debug >= 2) {
                         const tokenInfo = result.tokens
-                            ? ` ${result.tokens.input} → ${result.tokens.output} tok` + (result.tokens.speed ? ` ${result.tokens.speed} t/s` : '') + (result.tokens.cost != null ? ` $${result.tokens.cost.toFixed(4)}` : '')
+                            ? ` ${result.tokens.input} → ${result.tokens.output} tok`
+                                + (result.tokens.cached ? ` (cached:${result.tokens.cached})` : '')
+                                + (result.tokens.speed ? `| ${result.tokens.speed} t/s` : '')
+                                + (result.tokens.cost != null ? ` $${result.tokens.cost.toFixed(4)}` : '')
                             : '';
                         console.log(`✓${tokenInfo}\n${ModelMix.formatOutputSummary(result, currentConfig.debug).trim()}`);
                     }
@@ -1327,7 +1330,7 @@ class MixCustom {
                 message: message.trim(),
                 toolCalls: [],
                 think: null,
-                tokens: raw.length > 0 ? MixCustom.extractTokens(raw[raw.length - 1]) : { input: 0, output: 0, total: 0 }
+                tokens: raw.length > 0 ? MixCustom.extractTokens(raw[raw.length - 1]) : { input: 0, output: 0, total: 0, cached: 0 }
             }));
             response.data.on('error', reject);
         });
@@ -1379,13 +1382,15 @@ class MixCustom {
             return {
                 input: data.usage.prompt_tokens || 0,
                 output: data.usage.completion_tokens || 0,
-                total: data.usage.total_tokens || 0
+                total: data.usage.total_tokens || 0,
+                cached: ModelMix.extractCacheTokens(data.usage)
             };
         }
         return {
             input: 0,
             output: 0,
-            total: 0
+            total: 0,
+            cached: 0
         };
     }
 
@@ -1569,6 +1574,8 @@ class MixOpenAIResponses extends MixOpenAI {
         if (typeof options.n === 'number') request.n = options.n;
         if (options.logit_bias !== undefined) request.logit_bias = options.logit_bias;
         if (options.user !== undefined) request.user = options.user;
+        if (options.prompt_cache_key !== undefined) request.prompt_cache_key = options.prompt_cache_key;
+        if (options.prompt_cache_retention !== undefined) request.prompt_cache_retention = options.prompt_cache_retention;
 
         return request;
     }
@@ -1589,13 +1596,15 @@ class MixOpenAIResponses extends MixOpenAI {
             return {
                 input: data.usage.input_tokens || 0,
                 output: data.usage.output_tokens || 0,
-                total: data.usage.total_tokens || ((data.usage.input_tokens || 0) + (data.usage.output_tokens || 0))
+                total: data.usage.total_tokens || ((data.usage.input_tokens || 0) + (data.usage.output_tokens || 0)),
+                cached: ModelMix.extractCacheTokens(data.usage)
             };
         }
         return {
             input: 0,
             output: 0,
-            total: 0
+            total: 0,
+            cached: 0
         };
     }
 
@@ -2038,13 +2047,15 @@ class MixAnthropic extends MixCustom {
             return {
                 input: data.usage.input_tokens || 0,
                 output: data.usage.output_tokens || 0,
-                total: (data.usage.input_tokens || 0) + (data.usage.output_tokens || 0)
+                total: (data.usage.input_tokens || 0) + (data.usage.output_tokens || 0),
+                cached: ModelMix.extractCacheTokens(data.usage)
             };
         }
         return {
             input: 0,
             output: 0,
-            total: 0
+            total: 0,
+            cached: 0
         };
     }
 
@@ -2573,13 +2584,15 @@ class MixGoogle extends MixCustom {
             return {
                 input: data.usageMetadata.promptTokenCount || 0,
                 output: data.usageMetadata.candidatesTokenCount || 0,
-                total: data.usageMetadata.totalTokenCount || 0
+                total: data.usageMetadata.totalTokenCount || 0,
+                cached: ModelMix.extractCacheTokens(data.usageMetadata)
             };
         }
         return {
             input: 0,
             output: 0,
-            total: 0
+            total: 0,
+            cached: 0
         };
     }
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "modelmix",
-  "version": "4.4.16",
+  "version": "4.4.18",
   "description": "🧬 Reliable interface with automatic fallback for AI LLMs.",
   "main": "index.js",
   "repository": {
@@ -50,7 +50,7 @@
     "@modelcontextprotocol/sdk": "^1.27.1",
     "axios": "^1.13.5",
     "bottleneck": "^2.19.5",
-    "file-type": "^16.5.4",
+    "file-type": "^21.3.1",
     "form-data": "^4.0.4",
     "lemonlog": "^1.2.0",
     "ws": "^8.19.0"
@@ -61,6 +61,10 @@
     "nock": "^14.0.9",
     "sinon": "^21.0.0"
   },
+  "overrides": {
+    "diff": ">=8.0.3",
+    "serialize-javascript": ">=7.0.3"
+  },
   "scripts": {
     "test": "mocha test/**/*.js --timeout 10000 --require test/setup.js",
     "test:watch": "mocha test/**/*.js --watch --timeout 10000 --require test/setup.js",
@@ -71,6 +75,7 @@
     "test:bottleneck": "mocha test/bottleneck.test.js --timeout 10000 --require test/setup.js",
     "test:live": "mocha test/live.test.js --timeout 10000 --require test/setup.js",
     "test:live.mcp": "mocha test/live.mcp.js --timeout 60000 --require test/setup.js",
-    "test:tokens": "mocha test/tokens.test.js --timeout 10000 --require test/setup.js"
+    "test:tokens": "mocha test/tokens.test.js --timeout 10000 --require test/setup.js",
+    "test:offline": "mocha test/json.test.js test/fallback.test.js test/templates.test.js test/images.test.js test/bottleneck.test.js test/tokens.test.js test/history.test.js --timeout 10000 --require test/setup.js"
   }
 }

package/test/bottleneck.test.js

@@ -74,20 +74,13 @@ describe('Rate Limiting with Bottleneck Tests', () => {
             
             model.gpt51();
             
-            // Mock API responses
+            // Mock API responses (gpt51 uses /v1/responses)
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .times(3)
                 .reply(function() {
                     startTimes.push(Date.now());
-                    return [200, {
-                        choices: [{
-                            message: {
-                                role: 'assistant',
-                                content: `Response ${startTimes.length}`
-                            }
-                        }]
-                    }];
+                    return [200, testUtils.createMockResponse('openai-responses', `Response ${startTimes.length}`)];
                 });
 
             // Start three requests sequentially to test rate limiting
@@ -125,9 +118,9 @@ describe('Rate Limiting with Bottleneck Tests', () => {
             
             model.gpt51();
             
-            // Mock API with delay to simulate concurrent requests
+            // Mock API with delay to simulate concurrent requests (gpt51 uses /v1/responses)
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .times(5)
                 .reply(function() {
                     concurrentCount++;
@@ -136,14 +129,7 @@ describe('Rate Limiting with Bottleneck Tests', () => {
                     return new Promise(resolve => {
                         setTimeout(() => {
                             concurrentCount--;
-                            resolve([200, {
-                                choices: [{
-                                    message: {
-                                        role: 'assistant',
-                                        content: 'Concurrent response'
-                                    }
-                                }]
-                            }]);
+                            resolve([200, testUtils.createMockResponse('openai-responses', 'Concurrent response')]);
                         }, 100);
                     });
                 });
@@ -188,18 +174,11 @@ describe('Rate Limiting with Bottleneck Tests', () => {
             model.gpt51();
             
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .times(2)
                 .reply(function() {
                     requestTimes.push(Date.now());
-                    return [200, {
-                        choices: [{
-                            message: {
-                                role: 'assistant',
-                                content: 'OpenAI rate limited response'
-                            }
-                        }]
-                    }];
+                    return [200, testUtils.createMockResponse('openai-responses', 'OpenAI rate limited response')];
                 });
 
             const start = Date.now();
@@ -271,7 +250,7 @@ describe('Rate Limiting with Bottleneck Tests', () => {
             model.gpt51();
             
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .reply(429, {
                     error: {
                         message: 'Rate limit exceeded',
@@ -294,7 +273,7 @@ describe('Rate Limiting with Bottleneck Tests', () => {
             
             // First request fails
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .reply(function() {
                     requestTimes.push(Date.now());
                     return [500, { error: 'Server error' }];
@@ -302,17 +281,10 @@ describe('Rate Limiting with Bottleneck Tests', () => {
             
             // Second request succeeds
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .reply(function() {
                     requestTimes.push(Date.now());
-                    return [200, {
-                        choices: [{
-                            message: {
-                                role: 'assistant',
-                                content: 'Success after error'
-                            }
-                        }]
-                    }];
+                    return [200, testUtils.createMockResponse('openai-responses', 'Success after error')];
                 });
 
             const start = Date.now();
@@ -352,18 +324,11 @@ describe('Rate Limiting with Bottleneck Tests', () => {
             let requestCount = 0;
             
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .times(5)
                 .reply(function() {
                     requestCount++;
-                    return [200, {
-                        choices: [{
-                            message: {
-                                role: 'assistant',
-                                content: `Response ${requestCount}`
-                            }
-                        }]
-                    }];
+                    return [200, testUtils.createMockResponse('openai-responses', `Response ${requestCount}`)];
                 });
 
             const startTime = Date.now();
@@ -400,19 +365,13 @@ describe('Rate Limiting with Bottleneck Tests', () => {
             const results = [];
             
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .times(3)
                 .reply(function(uri, body) {
-                    const content = body.messages[0].content;
-                    results.push(content);
-                    return [200, {
-                        choices: [{
-                            message: {
-                                role: 'assistant',
-                                content: `Processed: ${content}`
-                            }
-                        }]
-                    }];
+                    const lastInput = body.input[body.input.length - 1];
+                    const text = lastInput?.content?.[0]?.text ?? '';
+                    results.push(text);
+                    return [200, testUtils.createMockResponse('openai-responses', `Processed: ${text}`)];
                 });
 
             // Submit requests with different priorities
@@ -447,16 +406,9 @@ describe('Rate Limiting with Bottleneck Tests', () => {
             model.gpt51();
             
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .times(3)
-                .reply(200, {
-                    choices: [{
-                        message: {
-                            role: 'assistant',
-                            content: 'Statistics tracking response'
-                        }
-                    }]
-                });
+                .reply(200, testUtils.createMockResponse('openai-responses', 'Statistics tracking response'));
 
             // Make multiple requests
             await Promise.all([
@@ -496,15 +448,8 @@ describe('Rate Limiting with Bottleneck Tests', () => {
             model.gpt51();
             
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
-                .reply(200, {
-                    choices: [{
-                        message: {
-                            role: 'assistant',
-                            content: 'Event handling response'
-                        }
-                    }]
-                });
+                .post('/v1/responses')
+                .reply(200, testUtils.createMockResponse('openai-responses', 'Event handling response'));
 
             // Make a request to trigger events
             model.addText('Event test').message();

package/test/images.test.js

@@ -25,7 +25,8 @@ describe('Image Processing and Multimodal Support Tests', () => {
         it('should handle base64 image data correctly', async () => {
             const base64Image = 'data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAoAAAAKCAYAAACNMs+9AAAAFUlEQVR42mP8z8BQz0AEYBxVSF+FABJADveWkH6oAAAAAElFTkSuQmCC';
 
-            model.gpt52()
+            // Use gpt5mini (chat/completions) - gpt52 uses Responses API which has different image format
+            model.gpt5mini()
                 .addText('What do you see in this image?')
                 .addImageFromUrl(base64Image);
 
@@ -62,8 +63,6 @@ describe('Image Processing and Multimodal Support Tests', () => {
             nock('https://api.anthropic.com')
                 .post('/v1/messages')
                 .reply(function (uri, body) {
-                    console.log(body.messages);
-                    // body is already parsed as JSON by nock
                     expect(body.messages).to.be.an('array');
                     // Find the message with the image
                     const userMsg = body.messages.find(m => m.role === 'user');

package/test/json.test.js

@@ -357,23 +357,15 @@ describe('JSON Schema and Structured Output Tests', () => {
             
             model.gpt52().addText('List 3 countries');
             
-            // Mock the API response
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
-                .reply(200, {
-                    choices: [{
-                        message: {
-                            role: 'assistant',
-                            content: JSON.stringify({
-                                countries: [
-                                    { name: 'France', capital: 'Paris' },
-                                    { name: 'Germany', capital: 'Berlin' },
-                                    { name: 'Spain', capital: 'Madrid' }
-                                ]
-                            })
-                        }
-                    }]
-                });
+                .post('/v1/responses')
+                .reply(200, testUtils.createMockResponse('openai-responses', JSON.stringify({
+                    countries: [
+                        { name: 'France', capital: 'Paris' },
+                        { name: 'Germany', capital: 'Berlin' },
+                        { name: 'Spain', capital: 'Madrid' }
+                    ]
+                })));
 
             const result = await model.json(example);
             
@@ -429,17 +421,9 @@ describe('JSON Schema and Structured Output Tests', () => {
         it('should handle JSON parsing errors gracefully', async () => {
             model.gpt52().addText('Generate invalid JSON');
             
-            // Mock invalid JSON response
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
-                .reply(200, {
-                    choices: [{
-                        message: {
-                            role: 'assistant',
-                            content: 'This is not valid JSON'
-                        }
-                    }]
-                });
+                .post('/v1/responses')
+                .reply(200, testUtils.createMockResponse('openai-responses', 'This is not valid JSON'));
 
             try {
                 await model.json({ name: 'test' });
@@ -453,21 +437,14 @@ describe('JSON Schema and Structured Output Tests', () => {
             model.gpt52().addText('List 3 countries');
 
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
-                .reply(200, {
-                    choices: [{
-                        message: {
-                            role: 'assistant',
-                            content: JSON.stringify({
-                                out: [
-                                    { name: 'France' },
-                                    { name: 'Germany' },
-                                    { name: 'Spain' }
-                                ]
-                            })
-                        }
-                    }]
-                });
+                .post('/v1/responses')
+                .reply(200, testUtils.createMockResponse('openai-responses', JSON.stringify({
+                    out: [
+                        { name: 'France' },
+                        { name: 'Germany' },
+                        { name: 'Spain' }
+                    ]
+                })));
 
             const result = await model.json([{ name: 'France' }]);
 

package/test/setup.js

@@ -104,6 +104,18 @@ global.testUtils = {
                         }
                     }]
                 };
+            case 'openai-responses':
+                return {
+                    output: [{
+                        type: 'message',
+                        content: [{ type: 'output_text', text: content }]
+                    }],
+                    usage: {
+                        input_tokens: 10,
+                        output_tokens: 5,
+                        total_tokens: 15
+                    }
+                };
                 
             case 'anthropic':
                 return {

package/test/templates.test.js

@@ -36,19 +36,11 @@ describe('Template and File Operations Tests', () => {
                 .addText('Hello {{name}}, you are {{age}} years old and live in {{city}}.');
 
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .reply(function (uri, body) {
-
-                    expect(body.messages[1].content[0].text).to.equal('Hello Alice, you are 30 years old and live in New York.');
-
-                    return [200, {
-                        choices: [{
-                            message: {
-                                role: 'assistant',
-                                content: 'Template processed successfully'
-                            }
-                        }]
-                    }];
+                    const userMsg = body.input.find(m => m.role === 'user');
+                    expect(userMsg.content[0].text).to.equal('Hello Alice, you are 30 years old and live in New York.');
+                    return [200, testUtils.createMockResponse('openai-responses', 'Template processed successfully')];
                 });
 
             const response = await model.message();
@@ -63,18 +55,11 @@ describe('Template and File Operations Tests', () => {
                 .addText('{{greeting}} {{name}}, {{action}} to our platform!');
 
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .reply(function (uri, body) {
-                    expect(body.messages[1].content[0].text).to.equal('Hello Bob, welcome to our platform!');
-
-                    return [200, {
-                        choices: [{
-                            message: {
-                                role: 'assistant',
-                                content: 'Multiple templates replaced'
-                            }
-                        }]
-                    }];
+                    const userMsg = body.input.find(m => m.role === 'user');
+                    expect(userMsg.content[0].text).to.equal('Hello Bob, welcome to our platform!');
+                    return [200, testUtils.createMockResponse('openai-responses', 'Multiple templates replaced')];
                 });
 
             const response = await model.message();
@@ -92,18 +77,11 @@ describe('Template and File Operations Tests', () => {
                 .addText('User {{user_name}} with role {{user_role}} works at {{company_name}} ({{company_domain}})');
 
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .reply(function (uri, body) {
-                    expect(body.messages[1].content[0].text).to.equal('User Charlie with role admin works at TechCorp (techcorp.com)');
-
-                    return [200, {
-                        choices: [{
-                            message: {
-                                role: 'assistant',
-                                content: 'Nested templates working'
-                            }
-                        }]
-                    }];
+                    const userMsg = body.input.find(m => m.role === 'user');
+                    expect(userMsg.content[0].text).to.equal('User Charlie with role admin works at TechCorp (techcorp.com)');
+                    return [200, testUtils.createMockResponse('openai-responses', 'Nested templates working')];
                 });
 
             const response = await model.message();
@@ -116,18 +94,11 @@ describe('Template and File Operations Tests', () => {
                 .addText('Hello {{name}}, your ID is {{user_id}} and status is {{status}}');
 
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .reply(function (uri, body) {
-                    expect(body.messages[1].content[0].text).to.equal('Hello David, your ID is {{user_id}} and status is {{status}}');
-
-                    return [200, {
-                        choices: [{
-                            message: {
-                                role: 'assistant',
-                                content: 'Partial template replacement'
-                            }
-                        }]
-                    }];
+                    const userMsg = body.input.find(m => m.role === 'user');
+                    expect(userMsg.content[0].text).to.equal('Hello David, your ID is {{user_id}} and status is {{status}}');
+                    return [200, testUtils.createMockResponse('openai-responses', 'Partial template replacement')];
                 });
 
             const response = await model.message();
@@ -145,18 +116,11 @@ describe('Template and File Operations Tests', () => {
                 .addText('Empty: {{empty}}, Special: {{special}}, Number: {{number}}, Boolean: {{boolean}}');
 
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .reply(function (uri, body) {
-                    expect(body.messages[1].content[0].text).to.equal('Empty: , Special: Hello & "World" <test>, Number: 42, Boolean: true');
-
-                    return [200, {
-                        choices: [{
-                            message: {
-                                role: 'assistant',
-                                content: 'Special characters handled'
-                            }
-                        }]
-                    }];
+                    const userMsg = body.input.find(m => m.role === 'user');
+                    expect(userMsg.content[0].text).to.equal('Empty: , Special: Hello & "World" <test>, Number: 42, Boolean: true');
+                    return [200, testUtils.createMockResponse('openai-responses', 'Special characters handled')];
                 });
 
             const response = await model.message();
@@ -189,24 +153,16 @@ describe('Template and File Operations Tests', () => {
                 .addText('Process this template: {{template}}');
 
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .reply(function (uri, body) {
-                    const content = body.messages[1].content[0].text;
-
+                    const userMsg = body.input.find(m => m.role === 'user');
+                    const content = userMsg.content[0].text;
                     expect(content).to.include('Hello Eve, welcome to ModelMix!');
                     expect(content).to.include('Username: eve_user');
                     expect(content).to.include('Role: developer');
                     expect(content).to.include('Created: 2023-12-01');
                     expect(content).to.include('The AI Solutions Team');
-
-                    return [200, {
-                        choices: [{
-                            message: {
-                                role: 'assistant',
-                                content: 'Template file processed'
-                            }
-                        }]
-                    }];
+                    return [200, testUtils.createMockResponse('openai-responses', 'Template file processed')];
                 });
 
             const response = await model.message();
@@ -219,10 +175,10 @@ describe('Template and File Operations Tests', () => {
                 .addText('Process this data: {{data}}');
 
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .reply(function (uri, body) {
-                    const content = body.messages[1].content[0].text;
-
+                    const userMsg = body.input.find(m => m.role === 'user');
+                    const content = userMsg.content[0].text;
                     expect(content).to.include('Alice Smith');
                     expect(content).to.include('alice@example.com');
                     expect(content).to.include('admin');
@@ -230,15 +186,7 @@ describe('Template and File Operations Tests', () => {
                     expect(content).to.include('Carol Davis');
                     expect(content).to.include('"theme": "dark"');
                     expect(content).to.include('"version": "1.0.0"');
-
-                    return [200, {
-                        choices: [{
-                            message: {
-                                role: 'assistant',
-                                content: 'JSON data processed'
-                            }
-                        }]
-                    }];
+                    return [200, testUtils.createMockResponse('openai-responses', 'JSON data processed')];
                 });
 
             const response = await model.message();
@@ -251,19 +199,11 @@ describe('Template and File Operations Tests', () => {
                 .addText('This should contain: {{missing}}');
 
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .reply(function (uri, body) {
-                    // The template should remain unreplaced if file doesn't exist
-                    expect(body.messages[1].content[0].text).to.equal('This should contain: {{missing}}');
-
-                    return [200, {
-                        choices: [{
-                            message: {
-                                role: 'assistant',
-                                content: 'File not found handled'
-                            }
-                        }]
-                    }];
+                    const userMsg = body.input.find(m => m.role === 'user');
+                    expect(userMsg.content[0].text).to.equal('This should contain: {{missing}}');
+                    return [200, testUtils.createMockResponse('openai-responses', 'File not found handled')];
                 });
 
             const response = await model.message();
@@ -286,26 +226,15 @@ describe('Template and File Operations Tests', () => {
                 .addText('Template: {{template}}\n\nData: {{data}}');
 
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .reply(function (uri, body) {
-                    const content = body.messages[1].content[0].text;
-
-                    // Should contain processed template
+                    const userMsg = body.input.find(m => m.role === 'user');
+                    const content = userMsg.content[0].text;
                     expect(content).to.include('Hello Frank, welcome to TestPlatform!');
                     expect(content).to.include('Username: frank_test');
-
-                    // Should contain JSON data
                     expect(content).to.include('Alice Smith');
                     expect(content).to.include('"theme": "dark"');
-
-                    return [200, {
-                        choices: [{
-                            message: {
-                                role: 'assistant',
-                                content: 'Multiple files processed'
-                            }
-                        }]
-                    }];
+                    return [200, testUtils.createMockResponse('openai-responses', 'Multiple files processed')];
                 });
 
             const response = await model.message();
@@ -329,21 +258,13 @@ describe('Template and File Operations Tests', () => {
                 .addText('Absolute path content: {{absolute}}');
 
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .reply(function (uri, body) {
-                    const content = body.messages[1].content[0].text;
-
+                    const userMsg = body.input.find(m => m.role === 'user');
+                    const content = userMsg.content[0].text;
                     expect(content).to.include('Hello Grace, welcome to AbsolutePath!');
                     expect(content).to.include('The Absolute Corp Team');
-
-                    return [200, {
-                        choices: [{
-                            message: {
-                                role: 'assistant',
-                                content: 'Absolute path works'
-                            }
-                        }]
-                    }];
+                    return [200, testUtils.createMockResponse('openai-responses', 'Absolute path works')];
                 });
 
             const response = await model.message();
@@ -372,22 +293,14 @@ describe('Template and File Operations Tests', () => {
                 .addText('Please {{action}} the following {{target}} and generate a {{format}}:\n\n{{user_data}}');
 
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .reply(function (uri, body) {
-                    const content = body.messages[1].content[0].text;
-
+                    const userMsg = body.input.find(m => m.role === 'user');
+                    const content = userMsg.content[0].text;
                     expect(content).to.include('Please analyze the following user behavior patterns and generate a detailed report:');
                     expect(content).to.include('Alice Smith');
                     expect(content).to.include('total_users');
-
-                    return [200, {
-                        choices: [{
-                            message: {
-                                role: 'assistant',
-                                content: 'Complex template integration successful'
-                            }
-                        }]
-                    }];
+                    return [200, testUtils.createMockResponse('openai-responses', 'Complex template integration successful')];
                 });
 
             const response = await model.message();
@@ -408,23 +321,22 @@ describe('Template and File Operations Tests', () => {
                 .addText('{{instruction}} from this data: {{data}}');
 
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
+                .post('/v1/responses')
                 .reply(function (uri, body) {
-                    expect(body.messages[1].content[0].text).to.include('Count active users by role');
-                    expect(body.messages[1].content[0].text).to.include('Alice Smith');
-
+                    const userMsg = body.input.find(m => m.role === 'user');
+                    expect(userMsg.content[0].text).to.include('Count active users by role');
+                    expect(userMsg.content[0].text).to.include('Alice Smith');
                     return [200, {
-                        choices: [{
-                            message: {
-                                role: 'assistant',
-                                content: JSON.stringify({
-                                    summary: 'User analysis completed',
-                                    user_count: 3,
-                                    active_users: 2,
-                                    roles: ['admin', 'user', 'moderator']
-                                })
-                            }
-                        }]
+                        output: [{
+                            type: 'message',
+                            content: [{ type: 'output_text', text: JSON.stringify({
+                                summary: 'User analysis completed',
+                                user_count: 3,
+                                active_users: 2,
+                                roles: ['admin', 'user', 'moderator']
+                            }) }]
+                        }],
+                        usage: { input_tokens: 10, output_tokens: 5, total_tokens: 15 }
                     }];
                 });
 
@@ -461,15 +373,8 @@ describe('Template and File Operations Tests', () => {
                 .addText('Content: {{bad_file}}');
 
             nock('https://api.openai.com')
-                .post('/v1/chat/completions')
-                .reply(200, {
-                    choices: [{
-                        message: {
-                            role: 'assistant',
-                            content: 'Error handled gracefully'
-                        }
-                    }]
-                });
+                .post('/v1/responses')
+                .reply(200, testUtils.createMockResponse('openai-responses', 'Error handled gracefully'));
 
             const response = await model.message();
             expect(response).to.include('Error handled gracefully');

package/test/tokens.test.js

@@ -1,5 +1,5 @@
 import { expect } from 'chai';
-import { ModelMix } from '../index.js';
+import { ModelMix, MixAnthropic, MixCustom, MixGoogle, MixOpenAIResponses } from '../index.js';
 import { createRequire } from 'module';
 
 const require = createRequire(import.meta.url);
@@ -18,6 +18,64 @@ describe('Token Usage Tracking', () => {
         nock.activate();
     });
 
+    it('should extract cached tokens from supported provider usage formats', function () {
+        const openAIChatTokens = MixCustom.extractTokens({
+            usage: {
+                prompt_tokens: 120,
+                completion_tokens: 30,
+                total_tokens: 150,
+                prompt_tokens_details: {
+                    cached_tokens: 80
+                }
+            }
+        });
+        const openAIResponsesTokens = MixOpenAIResponses.extractResponsesTokens({
+            usage: {
+                input_tokens: 90,
+                output_tokens: 20,
+                total_tokens: 110,
+                input_tokens_details: {
+                    cached_tokens: 45
+                }
+            }
+        });
+        const anthropicTokens = MixAnthropic.extractTokens({
+            usage: {
+                input_tokens: 60,
+                output_tokens: 15,
+                cache_read_input_tokens: 25
+            }
+        });
+        const googleTokens = MixGoogle.extractTokens({
+            usageMetadata: {
+                promptTokenCount: 70,
+                candidatesTokenCount: 10,
+                totalTokenCount: 80,
+                cachedContentTokenCount: 35
+            }
+        });
+
+        expect(openAIChatTokens.cached).to.equal(80);
+        expect(openAIResponsesTokens.cached).to.equal(45);
+        expect(anthropicTokens.cached).to.equal(25);
+        expect(googleTokens.cached).to.equal(35);
+    });
+
+    it('should pass OpenAI Responses prompt cache options through the request body', function () {
+        const request = MixOpenAIResponses.buildResponsesRequest({
+            model: 'gpt-5.4',
+            messages: [{
+                role: 'user',
+                content: [{ type: 'text', text: 'Explain caching briefly.' }]
+            }],
+            prompt_cache_key: 'demo-gpt54-cache',
+            prompt_cache_retention: '24h'
+        });
+
+        expect(request.prompt_cache_key).to.equal('demo-gpt54-cache');
+        expect(request.prompt_cache_retention).to.equal('24h');
+    });
+
     it('should track tokens in OpenAI response', async function () {
         this.timeout(30000);
 
@@ -31,10 +89,12 @@ describe('Token Usage Tracking', () => {
         expect(result.tokens).to.have.property('input');
         expect(result.tokens).to.have.property('output');
         expect(result.tokens).to.have.property('total');
+        expect(result.tokens).to.have.property('cached');
         
         expect(result.tokens.input).to.be.a('number');
         expect(result.tokens.output).to.be.a('number');
         expect(result.tokens.total).to.be.a('number');
+        expect(result.tokens.cached).to.be.a('number');
         
         expect(result.tokens.input).to.be.greaterThan(0);
         expect(result.tokens.output).to.be.greaterThan(0);
@@ -54,6 +114,7 @@ describe('Token Usage Tracking', () => {
         expect(result.tokens).to.have.property('input');
         expect(result.tokens).to.have.property('output');
         expect(result.tokens).to.have.property('total');
+        expect(result.tokens).to.have.property('cached');
         
         expect(result.tokens.input).to.be.greaterThan(0);
         expect(result.tokens.output).to.be.greaterThan(0);
@@ -73,6 +134,7 @@ describe('Token Usage Tracking', () => {
         expect(result.tokens).to.have.property('input');
         expect(result.tokens).to.have.property('output');
         expect(result.tokens).to.have.property('total');
+        expect(result.tokens).to.have.property('cached');
         
         expect(result.tokens.input).to.be.greaterThan(0);
         expect(result.tokens.output).to.be.greaterThan(0);
@@ -140,6 +202,7 @@ describe('Token Usage Tracking', () => {
             expect(result.tokens.input, `${provider.name} should have input`).to.be.a('number');
             expect(result.tokens.output, `${provider.name} should have output`).to.be.a('number');
             expect(result.tokens.total, `${provider.name} should have total`).to.be.a('number');
+            expect(result.tokens.cached, `${provider.name} should have cached`).to.be.a('number');
             
             // Verify values are positive
             expect(result.tokens.input, `${provider.name} input should be > 0`).to.be.greaterThan(0);