@qvac/sdk 0.14.1 → 0.16.0

package/dist/examples/kv-cache-cleanup.js

@@ -1,43 +1,43 @@
-import { completion, deleteCache, LLAMA_3_2_1B_INST_Q4_0, loadModel, unloadModel, VERBOSITY, } from "@qvac/sdk";
+import { completion, deleteCache, LLAMA_3_2_1B_INST_Q4_0, loadModel, unloadModel, VERBOSITY } from '@qvac/sdk';
 try {
-    console.log("▸ KV Cache Cleanup Demo\n");
-    console.log("▸ Cache structure: {kvCacheKey}/{modelId}/{configHash}.bin\n");
-    console.log("▸ " + "=".repeat(70) + "\n");
+    console.log('▸ KV Cache Cleanup Demo\n');
+    console.log('▸ Cache structure: {kvCacheKey}/{modelId}/{configHash}.bin\n');
+    console.log('▸ ' + '='.repeat(70) + '\n');
     const modelId = await loadModel({
         modelSrc: LLAMA_3_2_1B_INST_Q4_0,
         modelConfig: {
-            device: "gpu",
+            device: 'gpu',
             ctx_size: 2048,
-            verbosity: VERBOSITY.ERROR,
-        },
+            verbosity: VERBOSITY.ERROR
+        }
     });
     // Create some cached sessions
-    const cacheKey1 = "user-alice-session";
-    const cacheKey2 = "user-bob-session";
+    const cacheKey1 = 'user-alice-session';
+    const cacheKey2 = 'user-bob-session';
     console.log(`▸ Creating cache for key: "${cacheKey1}"`);
     const result1 = completion({
         modelId,
-        history: [{ role: "user", content: "Hello, I'm Alice!" }],
+        history: [{ role: 'user', content: "Hello, I'm Alice!" }],
         stream: true,
-        kvCache: cacheKey1,
+        kvCache: cacheKey1
     });
     for await (const token of result1.tokenStream) {
         process.stdout.write(token);
     }
-    console.log("\n▸ Cache created\n");
+    console.log('\n▸ Cache created\n');
     console.log(`▸ Creating cache for key: "${cacheKey2}"`);
     const result2 = completion({
         modelId,
-        history: [{ role: "user", content: "Hello, I'm Bob!" }],
+        history: [{ role: 'user', content: "Hello, I'm Bob!" }],
         stream: true,
-        kvCache: cacheKey2,
+        kvCache: cacheKey2
     });
     for await (const token of result2.tokenStream) {
         process.stdout.write(token);
     }
-    console.log("\n▸ Cache created\n");
+    console.log('\n▸ Cache created\n');
     // Now cleanup specific cache key
-    console.log("▸ " + "=".repeat(70));
+    console.log('▸ ' + '='.repeat(70));
     console.log(`\n▸ Deleting cache key: "${cacheKey1}"`);
     const deleteResult1 = await deleteCache({ kvCacheKey: cacheKey1 });
     if (deleteResult1.success) {
@@ -48,7 +48,7 @@ try {
     console.log(`\n▸ Deleting specific model in cache key: "${cacheKey2}"`);
     const deleteResult2 = await deleteCache({
         kvCacheKey: cacheKey2,
-        modelId: modelId,
+        modelId: modelId
     });
     if (deleteResult2.success) {
         console.log(`▸ Deleted model "${modelId}" from cache key "${cacheKey2}"`);
@@ -56,31 +56,31 @@ try {
     }
     // Try to delete non-existent cache key (should succeed silently)
     console.log(`\n▸ Deleting non-existent cache key: "non-existent"`);
-    const deleteResult3 = await deleteCache({ kvCacheKey: "non-existent" });
+    const deleteResult3 = await deleteCache({ kvCacheKey: 'non-existent' });
     if (deleteResult3.success) {
         console.log("▸ Delete succeeded (cache key didn't exist)");
     }
-    console.log("\n▸ " + "=".repeat(70));
-    console.log("\n▸ CLEANUP OPTIONS:");
-    console.log("▸ " + "-".repeat(70));
-    console.log("\n▸ 1. Delete entire cache key (all models):");
+    console.log('\n▸ ' + '='.repeat(70));
+    console.log('\n▸ CLEANUP OPTIONS:');
+    console.log('▸ ' + '-'.repeat(70));
+    console.log('\n▸ 1. Delete entire cache key (all models):');
     console.log('▸    await deleteCache({ kvCacheKey: "my-session" })');
-    console.log("▸    └─ Removes: my-session/ (entire directory)");
-    console.log("\n▸ 2. Delete specific model within cache key:");
+    console.log('▸    └─ Removes: my-session/ (entire directory)');
+    console.log('\n▸ 2. Delete specific model within cache key:');
     console.log('▸    await deleteCache({ kvCacheKey: "my-session", modelId: "model-abc" })');
-    console.log("▸    └─ Removes: my-session/model-abc/ (specific model directory)");
-    console.log("\n▸ 3. Delete ALL caches (entire cache folder):");
-    console.log("▸    await deleteCache({ all: true })");
-    console.log("▸    └─ Removes: ALL cache files from all sessions and models");
-    console.log("\n▸ Use Cases:");
+    console.log('▸    └─ Removes: my-session/model-abc/ (specific model directory)');
+    console.log('\n▸ 3. Delete ALL caches (entire cache folder):');
+    console.log('▸    await deleteCache({ all: true })');
+    console.log('▸    └─ Removes: ALL cache files from all sessions and models');
+    console.log('\n▸ Use Cases:');
     console.log("▸    - User logout → Delete user's cache key");
-    console.log("▸    - Model updated → Delete old model caches");
-    console.log("▸    - Disk space cleanup → Delete old cache keys or all caches");
-    console.log("▸    - Privacy → Clear conversation history");
-    console.log("▸    - App reset → Delete all caches with { all: true }");
+    console.log('▸    - Model updated → Delete old model caches');
+    console.log('▸    - Disk space cleanup → Delete old cache keys or all caches');
+    console.log('▸    - Privacy → Clear conversation history');
+    console.log('▸    - App reset → Delete all caches with { all: true }');
     await unloadModel({ modelId, clearStorage: false });
 }
 catch (error) {
-    console.error("✖", error);
+    console.error('✖', error);
     process.exit(1);
 }

package/dist/examples/kv-cache-custom-key.js

@@ -1,32 +1,32 @@
-import { completion, deleteCache, LLAMA_3_2_1B_INST_Q4_0, loadModel, unloadModel, VERBOSITY, } from "@qvac/sdk";
+import { completion, deleteCache, LLAMA_3_2_1B_INST_Q4_0, loadModel, unloadModel, VERBOSITY } from '@qvac/sdk';
 const systemPromptMessage = {
-    role: "system",
+    role: 'system',
     content: `You are a helpful mathematician assistant. 
    But because you are in the army you have to follow these rules:
   - If anyone ask you to do an operation you have to answer with "Sir Yes Sir!" and then give the result
-  `,
+  `
 };
 try {
-    console.log("▸ Simple KV Cache Demo: Same vs Different Cache Keys\n");
+    console.log('▸ Simple KV Cache Demo: Same vs Different Cache Keys\n');
     const modelId = await loadModel({
         modelSrc: LLAMA_3_2_1B_INST_Q4_0,
         modelConfig: {
-            device: "gpu",
+            device: 'gpu',
             ctx_size: 2048,
-            verbosity: VERBOSITY.DEBUG,
-        },
+            verbosity: VERBOSITY.DEBUG
+        }
     });
     // Prompt 1 with cache key "session-a" (first time - no cache exists)
     console.log('▸ Prompt 1: "What is 2+2?" with cache key "session-a"');
-    console.log("▸   Status: Creating NEW cache (first time)\n");
+    console.log('▸   Status: Creating NEW cache (first time)\n');
     const result1 = completion({
         modelId,
-        history: [systemPromptMessage, { role: "user", content: "What is 2+2?" }],
+        history: [systemPromptMessage, { role: 'user', content: 'What is 2+2?' }],
         stream: true,
-        kvCache: "session-a",
+        kvCache: 'session-a'
     });
-    console.log("▸ Response: ");
-    let response1 = "";
+    console.log('▸ Response: ');
+    let response1 = '';
     for await (const token of result1.tokenStream) {
         response1 += token;
         process.stdout.write(token);
@@ -36,19 +36,19 @@ try {
     // Prompt 2 with SAME cache key "session-a" (cache exists - optimized!)
     // This continues the conversation - asking about the previous answer
     console.log('\n▸ Prompt 2: "Now multiply that result by 3" with cache key "session-a" (SAME)');
-    console.log("▸   Status: REUSING existing cache (only last message sent, but context remembered!)\n");
+    console.log('▸   Status: REUSING existing cache (only last message sent, but context remembered!)\n');
     const result2 = completion({
         modelId,
         history: [
             systemPromptMessage,
-            { role: "user", content: "What is 2+2?" },
-            { role: "assistant", content: response1 },
-            { role: "user", content: "Now multiply that result by 3" }, // Refers to previous answer!
+            { role: 'user', content: 'What is 2+2?' },
+            { role: 'assistant', content: response1 },
+            { role: 'user', content: 'Now multiply that result by 3' } // Refers to previous answer!
         ],
         stream: true,
-        kvCache: "session-a", // Same cache key!
+        kvCache: 'session-a' // Same cache key!
     });
-    console.log("▸ Response: ");
+    console.log('▸ Response: ');
     for await (const token of result2.tokenStream) {
         process.stdout.write(token);
     }
@@ -56,14 +56,14 @@ try {
     console.log(`\n▸ Stats: TTFT=${stats2?.timeToFirstToken}ms, CacheTokens=${stats2?.cacheTokens}\n`);
     // Prompt 3 with DIFFERENT cache key "session-b" (new cache)
     console.log('\n▸ Prompt 3: "What is 10+10?" with cache key "session-b" (DIFFERENT)');
-    console.log("▸   Status: Creating NEW cache (different key = isolated cache)\n");
+    console.log('▸   Status: Creating NEW cache (different key = isolated cache)\n');
     const result3 = completion({
         modelId,
-        history: [systemPromptMessage, { role: "user", content: "What is 10+10?" }],
+        history: [systemPromptMessage, { role: 'user', content: 'What is 10+10?' }],
         stream: true,
-        kvCache: "session-b", // Different cache key!
+        kvCache: 'session-b' // Different cache key!
     });
-    console.log("▸ Response: ");
+    console.log('▸ Response: ');
     for await (const token of result3.tokenStream) {
         process.stdout.write(token);
     }
@@ -71,30 +71,30 @@ try {
     console.log(`\n▸ Stats: TTFT=${stats3?.timeToFirstToken}ms, CacheTokens=${stats3?.cacheTokens}\n`);
     // Prompt 4: Go back to session-a to trigger flush of session-b
     console.log('\n▸ Prompt 4: "What is 5+5?" with cache key "session-a" (BACK TO FIRST)');
-    console.log("▸   Status: This should trigger session-b to be flushed to disk\n");
+    console.log('▸   Status: This should trigger session-b to be flushed to disk\n');
     const result4 = completion({
         modelId,
-        history: [systemPromptMessage, { role: "user", content: "What is 5+5?" }],
+        history: [systemPromptMessage, { role: 'user', content: 'What is 5+5?' }],
         stream: true,
-        kvCache: "session-a", // Back to session-a!
+        kvCache: 'session-a' // Back to session-a!
     });
-    console.log("▸ Response: ");
+    console.log('▸ Response: ');
     for await (const token of result4.tokenStream) {
         process.stdout.write(token);
     }
     const stats4 = await result4.stats;
     console.log(`\n▸ Stats: TTFT=${stats4?.timeToFirstToken}ms, CacheTokens=${stats4?.cacheTokens}\n`);
     // Summary
-    console.log("\n▸ Summary:");
+    console.log('\n▸ Summary:');
     console.log('▸   - Prompts 1 & 2 share cache key "session-a" → same cache file');
     console.log("▸   - Prompt 2 built on Prompt 1's context (model remembered the answer!)");
-    console.log("▸   - Only the last message was sent, but full conversation context maintained");
+    console.log('▸   - Only the last message was sent, but full conversation context maintained');
     console.log('▸   - Prompt 3 uses cache key "session-b" → separate cache file (isolated)');
-    console.log("▸   - Different cache keys provide complete isolation between sessions");
+    console.log('▸   - Different cache keys provide complete isolation between sessions');
     await deleteCache({ all: true });
     await unloadModel({ modelId, clearStorage: false });
 }
 catch (error) {
-    console.error("✖", error);
+    console.error('✖', error);
     process.exit(1);
 }

package/dist/examples/kv-cache-example.js

@@ -1,25 +1,23 @@
-import { completion, LLAMA_3_2_1B_INST_Q4_0, loadModel, unloadModel, VERBOSITY, } from "@qvac/sdk";
+import { completion, LLAMA_3_2_1B_INST_Q4_0, loadModel, unloadModel, VERBOSITY } from '@qvac/sdk';
 try {
     // Load the model
     const modelId = await loadModel({
         modelSrc: LLAMA_3_2_1B_INST_Q4_0,
         modelConfig: {
-            device: "gpu",
+            device: 'gpu',
             ctx_size: 2048,
-            verbosity: VERBOSITY.ERROR,
-        },
+            verbosity: VERBOSITY.ERROR
+        }
     });
-    console.log("▸ Testing KV Cache functionality...\n");
+    console.log('▸ Testing KV Cache functionality...\n');
     // First conversation with auto-keyed cache enabled
-    console.log("▸ First conversation (building cache for the next turn):");
-    const history1 = [
-        { role: "user", content: "What is the capital of France?" },
-    ];
+    console.log('▸ First conversation (building cache for the next turn):');
+    const history1 = [{ role: 'user', content: 'What is the capital of France?' }];
     const result1 = completion({
         modelId,
         history: history1,
         stream: true,
-        kvCache: true,
+        kvCache: true
     }); // kvCache = true
     for await (const token of result1.tokenStream) {
         process.stdout.write(token);
@@ -28,14 +26,14 @@ try {
     const stats1 = final1.stats;
     console.log(`\n▸ First completion stats: ${JSON.stringify(stats1)}\n`);
     // Continue conversation (should reuse the completed first-turn cache).
-    console.log("▸ Continuing conversation (reusing previous turn cache):");
+    console.log('▸ Continuing conversation (reusing previous turn cache):');
     const history2 = [
-        { role: "user", content: "What is the capital of France?" },
+        { role: 'user', content: 'What is the capital of France?' },
         {
-            role: "assistant",
-            content: final1.cacheableAssistantContent ?? final1.contentText,
+            role: 'assistant',
+            content: final1.cacheableAssistantContent ?? final1.contentText
         },
-        { role: "user", content: "What about Germany?" },
+        { role: 'user', content: 'What about Germany?' }
     ];
     // Auto-keyed caching should:
     // 1. Find the cache saved after turn 1 under [user, assistant]
@@ -45,7 +43,7 @@ try {
         modelId,
         history: history2,
         stream: true,
-        kvCache: true,
+        kvCache: true
     }); // kvCache = true
     for await (const token of result2.tokenStream) {
         process.stdout.write(token);
@@ -53,22 +51,22 @@ try {
     const stats2 = await result2.stats;
     console.log(`\n▸ Second completion stats: ${JSON.stringify(stats2)}\n`);
     // Compare with non-cached version
-    console.log("▸ Same conversation without cache:");
+    console.log('▸ Same conversation without cache:');
     const result3 = completion({
         modelId,
         history: history2,
         stream: true,
-        kvCache: false,
+        kvCache: false
     }); // kvCache = false
     for await (const token of result3.tokenStream) {
         process.stdout.write(token);
     }
     const stats3 = await result3.stats;
     console.log(`\n▸ Non-cached completion stats: ${JSON.stringify(stats3)}\n`);
-    console.log("▸ KV Cache test completed!");
+    console.log('▸ KV Cache test completed!');
     await unloadModel({ modelId, clearStorage: false });
 }
 catch (error) {
-    console.error("✖", error);
+    console.error('✖', error);
     process.exit(1);
 }

package/dist/examples/llamacpp-cache.js

@@ -1,28 +1,28 @@
-import { completion, deleteCache, LLAMA_3_2_1B_INST_Q4_0, loadModel, unloadModel, VERBOSITY, } from "@qvac/sdk";
-const sharedCacheKey = "llamacpp-cache-demo-shared";
-const isolatedCacheKey = "llamacpp-cache-demo-isolated";
+import { completion, deleteCache, LLAMA_3_2_1B_INST_Q4_0, loadModel, unloadModel, VERBOSITY } from '@qvac/sdk';
+const sharedCacheKey = 'llamacpp-cache-demo-shared';
+const isolatedCacheKey = 'llamacpp-cache-demo-isolated';
 const systemPrompt = {
-    role: "system",
-    content: "You are a concise travel assistant. Keep track of the user's preferences across turns.",
+    role: 'system',
+    content: "You are a concise travel assistant. Keep track of the user's preferences across turns."
 };
 function printStats(stats) {
-    console.log("\n▸ Performance Stats:", {
+    console.log('\n▸ Performance Stats:', {
         timeToFirstToken: stats?.timeToFirstToken,
         tokensPerSecond: stats?.tokensPerSecond,
-        cacheTokens: stats?.cacheTokens,
+        cacheTokens: stats?.cacheTokens
     });
 }
 async function runCachedCompletion(params) {
     console.log(`\n▸ ${params.label}`);
-    console.log(`▸ Cache key: ${params.kvCache || "disabled"}`);
-    console.log("▸ AI Response:");
+    console.log(`▸ Cache key: ${params.kvCache || 'disabled'}`);
+    console.log('▸ AI Response:');
     const result = completion({
         modelId: params.modelId,
         history: params.history,
         stream: true,
-        kvCache: params.kvCache,
+        kvCache: params.kvCache
     });
-    let text = "";
+    let text = '';
     for await (const token of result.tokenStream) {
         text += token;
         process.stdout.write(token);
@@ -33,61 +33,61 @@ async function runCachedCompletion(params) {
 }
 let modelId;
 try {
-    console.log("▸ Loading llama.cpp model with cache support...");
+    console.log('▸ Loading llama.cpp model with cache support...');
     modelId = await loadModel({
         modelSrc: LLAMA_3_2_1B_INST_Q4_0,
         modelConfig: {
             ctx_size: 4096,
-            verbosity: VERBOSITY.ERROR,
+            verbosity: VERBOSITY.ERROR
         },
         onProgress: (p) => {
             const mb = (n) => (n / 1e6).toFixed(1);
             const line = `▸ Downloading ${p.percentage.toFixed(0)}% (${mb(p.downloaded)}/${mb(p.total)} MB)`;
             process.stderr.write(process.stderr.isTTY ? `\r${line}` : `${line}\n`);
             if (p.percentage >= 100)
-                process.stderr.write("\n");
-        },
+                process.stderr.write('\n');
+        }
     });
     console.log(`▸ Model loaded successfully. Model ID: ${modelId}`);
     await deleteCache({ kvCacheKey: sharedCacheKey });
     await deleteCache({ kvCacheKey: isolatedCacheKey });
     const firstUserMessage = {
-        role: "user",
-        content: "I am planning a weekend in Lisbon. I prefer museums, seafood, and walking routes. Suggest a day plan in 3 bullets.",
+        role: 'user',
+        content: 'I am planning a weekend in Lisbon. I prefer museums, seafood, and walking routes. Suggest a day plan in 3 bullets.'
     };
     const firstTurn = await runCachedCompletion({
-        label: "Turn 1: create a named cache",
+        label: 'Turn 1: create a named cache',
         modelId,
         kvCache: sharedCacheKey,
-        history: [systemPrompt, firstUserMessage],
+        history: [systemPrompt, firstUserMessage]
     });
     await runCachedCompletion({
-        label: "Turn 2: reuse the same cache key",
+        label: 'Turn 2: reuse the same cache key',
         modelId,
         kvCache: sharedCacheKey,
         history: [
             systemPrompt,
             firstUserMessage,
-            { role: "assistant", content: firstTurn.text.trim() },
+            { role: 'assistant', content: firstTurn.text.trim() },
             {
-                role: "user",
-                content: "Now revise the plan for a rainy day, but keep the same preferences.",
-            },
-        ],
+                role: 'user',
+                content: 'Now revise the plan for a rainy day, but keep the same preferences.'
+            }
+        ]
     });
     await runCachedCompletion({
-        label: "Turn 3: use a different cache key for an isolated session",
+        label: 'Turn 3: use a different cache key for an isolated session',
         modelId,
         kvCache: isolatedCacheKey,
         history: [
             systemPrompt,
             {
-                role: "user",
-                content: "I am planning a weekend in Berlin. I prefer nightlife and late starts. Suggest a day plan in 3 bullets.",
-            },
-        ],
+                role: 'user',
+                content: 'I am planning a weekend in Berlin. I prefer nightlife and late starts. Suggest a day plan in 3 bullets.'
+            }
+        ]
     });
-    console.log("\n▸ Summary:");
+    console.log('\n▸ Summary:');
     console.log(`▸ - Reusing "${sharedCacheKey}" keeps conversation context across turns.`);
     console.log(`▸ - Switching to "${isolatedCacheKey}" starts a separate cache session.`);
     console.log('▸ - Use deleteCache({ kvCacheKey: "your-session" }) to clear a saved session.');
@@ -102,9 +102,9 @@ catch (error) {
             await unloadModel({ modelId, clearStorage: false });
         }
         catch (cleanupError) {
-            console.warn("✖ cleanup failed:", cleanupError);
+            console.warn('✖ cleanup failed:', cleanupError);
         }
     }
-    console.error("✖", error);
+    console.error('✖', error);
     process.exit(1);
 }

package/dist/examples/llamacpp-dynamic-tools.js

@@ -12,51 +12,51 @@
  *   bun run build
  *   bun run bare:example dist/examples/llamacpp-dynamic-tools.js
  */
-import { z } from "zod";
-import { completion, loadModel, unloadModel, QWEN3_1_7B_INST_Q4, } from "@qvac/sdk";
+import { z } from 'zod';
+import { completion, loadModel, unloadModel, QWEN3_1_7B_INST_Q4 } from '@qvac/sdk';
 const weatherSchema = z.object({
-    city: z.string().describe("City name"),
+    city: z.string().describe('City name')
 });
 const horoscopeSchema = z.object({
-    sign: z.string().describe("An astrological sign, e.g. Taurus or Aquarius"),
+    sign: z.string().describe('An astrological sign, e.g. Taurus or Aquarius')
 });
 const dateSchema = z.object({});
 const toolSchemas = {
     get_weather: weatherSchema,
     get_horoscope: horoscopeSchema,
-    get_date: dateSchema,
+    get_date: dateSchema
 };
 const weatherTools = [
     {
-        name: "get_weather",
-        description: "Get current weather for a city",
-        parameters: weatherSchema,
-    },
+        name: 'get_weather',
+        description: 'Get current weather for a city',
+        parameters: weatherSchema
+    }
 ];
 const horoscopeTools = [
     {
-        name: "get_horoscope",
+        name: 'get_horoscope',
         description: "Get today's horoscope for an astrological sign",
-        parameters: horoscopeSchema,
-    },
+        parameters: horoscopeSchema
+    }
 ];
 const dateTools = [
     {
-        name: "get_date",
+        name: 'get_date',
         description: "Get today's date",
-        parameters: dateSchema,
-    },
+        parameters: dateSchema
+    }
 ];
 function executeToolCall(call) {
-    if (call.name === "get_weather") {
+    if (call.name === 'get_weather') {
         const args = call.arguments;
         return `The weather in ${args.city} is rainy, 8°C with heavy clouds.`;
     }
-    if (call.name === "get_horoscope") {
+    if (call.name === 'get_horoscope') {
         const args = call.arguments;
         return `Horoscope for ${args.sign}: a great day for new beginnings.`;
     }
-    if (call.name === "get_date") {
+    if (call.name === 'get_date') {
         return new Date().toISOString().slice(0, 10);
     }
     return `Unknown tool: ${call.name}`;
@@ -67,7 +67,7 @@ async function chatTurn({ modelId, kvCache, history, tools }) {
         history,
         tools,
         kvCache,
-        stream: true,
+        stream: true
     });
     const tokensTask = (async () => {
         for await (const token of result.tokenStream) {
@@ -83,7 +83,7 @@ async function chatTurn({ modelId, kvCache, history, tools }) {
     const text = await result.text;
     const toolCalls = await result.toolCalls;
     if (toolCalls.length === 0) {
-        history.push({ role: "assistant", content: text });
+        history.push({ role: 'assistant', content: text });
         return;
     }
     for (const call of toolCalls) {
@@ -95,9 +95,9 @@ async function chatTurn({ modelId, kvCache, history, tools }) {
             }
         }
     }
-    history.push({ role: "assistant", content: text });
+    history.push({ role: 'assistant', content: text });
     for (const call of toolCalls) {
-        history.push({ role: "tool", content: executeToolCall(call) });
+        history.push({ role: 'tool', content: executeToolCall(call) });
     }
     // Follow-up turn so the model can incorporate the tool results.
     await chatTurn({ modelId, kvCache, history, tools });
@@ -108,45 +108,45 @@ async function main() {
         modelConfig: {
             ctx_size: 4096,
             tools: true,
-            toolsMode: "dynamic",
+            toolsMode: 'dynamic'
         },
         onProgress: (p) => {
             const mb = (n) => (n / 1e6).toFixed(1);
             const line = `▸ Downloading ${p.percentage.toFixed(0)}% (${mb(p.downloaded)}/${mb(p.total)} MB)`;
             process.stderr.write(process.stderr.isTTY ? `\r${line}` : `${line}\n`);
             if (p.percentage >= 100)
-                process.stderr.write("\n");
-        },
+                process.stderr.write('\n');
+        }
     });
     console.log(`▸ Model loaded: ${modelId}`);
     const kvCache = `dynamic-tools-${Date.now()}`;
     const history = [
         {
-            role: "system",
-            content: "You are a helpful assistant that uses tools when they are available. " +
-                "User's cat is named Windy and dog is named Butch.",
-        },
+            role: 'system',
+            content: 'You are a helpful assistant that uses tools when they are available. ' +
+                "User's cat is named Windy and dog is named Butch."
+        }
     ];
     // Turn 1 — only weather tools available.
-    history.push({ role: "user", content: "What's the weather in Tokyo?" });
-    console.log("\n▸ Turn 1 (tools=weather):\n");
+    history.push({ role: 'user', content: "What's the weather in Tokyo?" });
+    console.log('\n▸ Turn 1 (tools=weather):\n');
     await chatTurn({ modelId, kvCache, history, tools: weatherTools });
     // Turn 2 — same session, swap to horoscope tools. Dynamic mode lets the
     // model see a different tool set without invalidating the kv-cache.
     history.push({
-        role: "user",
-        content: "Now check my horoscope for Aquarius.",
+        role: 'user',
+        content: 'Now check my horoscope for Aquarius.'
     });
-    console.log("\n\n▸ Turn 2 (tools=horoscope):\n");
+    console.log('\n\n▸ Turn 2 (tools=horoscope):\n');
     await chatTurn({ modelId, kvCache, history, tools: horoscopeTools });
     // Turn 3 — swap to a parameterless tool to confirm empty-arg flows work.
-    history.push({ role: "user", content: "What's today's date?" });
-    console.log("\n\n▸ Turn 3 (tools=date):\n");
+    history.push({ role: 'user', content: "What's today's date?" });
+    console.log('\n\n▸ Turn 3 (tools=date):\n');
     await chatTurn({ modelId, kvCache, history, tools: dateTools });
-    console.log("\n\n▸ Done.");
+    console.log('\n\n▸ Done.');
     await unloadModel({ modelId, clearStorage: false });
 }
 main().catch((err) => {
-    console.error("✖", err);
+    console.error('✖', err);
     process.exit(1);
 });