genai-lite 0.1.0 → 0.1.1

package/README.md

@@ -215,6 +215,101 @@ import type {
 } from 'genai-lite';
 ```
 
+## Utilities
+
+genai-lite includes useful utilities for working with LLMs, available through the `genai-lite/utils` subpath:
+
+### Token Counting
+
+Count the number of tokens in a string using OpenAI's tiktoken library:
+
+```typescript
+import { countTokens } from 'genai-lite/utils';
+
+const text = 'Hello, this is a sample text for token counting.';
+const tokenCount = countTokens(text); // Uses gpt-4 tokenizer by default
+console.log(`Token count: ${tokenCount}`);
+
+// Specify a different model's tokenizer
+const gpt35Tokens = countTokens(text, 'gpt-3.5-turbo');
+```
+
+**Note:** The `countTokens` function uses the `js-tiktoken` library and supports all models that have tiktoken encodings.
+
+### Smart Text Preview
+
+Generate intelligent previews of large text blocks that preserve context:
+
+```typescript
+import { getSmartPreview } from 'genai-lite/utils';
+
+const largeCodeFile = `
+function calculateTotal(items) {
+  let total = 0;
+  
+  for (const item of items) {
+    total += item.price * item.quantity;
+  }
+  
+  return total;
+}
+
+function applyDiscount(total, discountPercent) {
+  return total * (1 - discountPercent / 100);
+}
+
+// ... many more lines of code ...
+`;
+
+// Get a preview that shows at least 5 lines but extends to a logical break point
+const preview = getSmartPreview(largeCodeFile, { 
+  minLines: 5, 
+  maxLines: 10 
+});
+```
+
+The `getSmartPreview` function intelligently truncates text:
+- Returns the full content if it's shorter than `maxLines`
+- Shows at least `minLines` of content
+- Extends to the next blank line (up to `maxLines`) to avoid cutting off in the middle of a code block or paragraph
+- Adds `... (content truncated)` when content is truncated
+
+### Example: Building Token-Aware Prompts
+
+Combine these utilities to build prompts that fit within model context windows:
+
+```typescript
+import { LLMService, fromEnvironment } from 'genai-lite';
+import { countTokens, getSmartPreview } from 'genai-lite/utils';
+
+const llm = new LLMService(fromEnvironment);
+
+// Large source file
+const sourceCode = await fs.readFile('large-file.js', 'utf-8');
+
+// Get a smart preview that fits within token budget
+let preview = getSmartPreview(sourceCode, { minLines: 20, maxLines: 50 });
+let tokenCount = countTokens(preview, 'gpt-4.1-mini');
+
+// Adjust preview if needed to fit token budget
+const maxTokens = 4000;
+if (tokenCount > maxTokens) {
+  preview = getSmartPreview(sourceCode, { minLines: 10, maxLines: 30 });
+}
+
+// Send to LLM
+const response = await llm.sendMessage({
+  providerId: 'openai',
+  modelId: 'gpt-4.1-mini',
+  messages: [
+    { 
+      role: 'user', 
+      content: `Analyze this code:\n\n${preview}` 
+    }
+  ]
+});
+```
+
 ## Contributing
 
 Contributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.

package/dist/utils/index.d.ts

@@ -0,0 +1 @@
+export * from './prompt';

package/dist/utils/index.js

@@ -0,0 +1,17 @@
+"use strict";
+var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    var desc = Object.getOwnPropertyDescriptor(m, k);
+    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
+      desc = { enumerable: true, get: function() { return m[k]; } };
+    }
+    Object.defineProperty(o, k2, desc);
+}) : (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    o[k2] = m[k];
+}));
+var __exportStar = (this && this.__exportStar) || function(m, exports) {
+    for (var p in m) if (p !== "default" && !Object.prototype.hasOwnProperty.call(exports, p)) __createBinding(exports, m, p);
+};
+Object.defineProperty(exports, "__esModule", { value: true });
+__exportStar(require("./prompt"), exports);

package/dist/utils/prompt.d.ts

@@ -0,0 +1,6 @@
+import { TiktokenModel } from 'js-tiktoken';
+export declare function countTokens(text: string, model?: TiktokenModel): number;
+export declare function getSmartPreview(content: string, config: {
+    minLines: number;
+    maxLines: number;
+}): string;

package/dist/utils/prompt.js

@@ -0,0 +1,55 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.countTokens = countTokens;
+exports.getSmartPreview = getSmartPreview;
+const js_tiktoken_1 = require("js-tiktoken");
+const tokenizerCache = new Map();
+function getTokenizer(model) {
+    if (tokenizerCache.has(model)) {
+        return tokenizerCache.get(model);
+    }
+    try {
+        const tokenizer = (0, js_tiktoken_1.encodingForModel)(model);
+        tokenizerCache.set(model, tokenizer);
+        return tokenizer;
+    }
+    catch (error) {
+        console.error(`Failed to initialize tokenizer for model ${model}:`, error);
+        throw error;
+    }
+}
+function countTokens(text, model = 'gpt-4') {
+    if (!text)
+        return 0;
+    try {
+        const tokenizer = getTokenizer(model);
+        return tokenizer.encode(text).length;
+    }
+    catch (error) {
+        // Fallback to a rough estimate if tokenizer fails for any reason
+        return Math.ceil(text.length / 4);
+    }
+}
+function getSmartPreview(content, config) {
+    const lines = content.split('\n');
+    // If the file is not longer than maxLines, return it in full
+    if (lines.length <= config.maxLines) {
+        return content;
+    }
+    // Always show at least minLines
+    let endLine = config.minLines;
+    let emptyLinesCount = lines
+        .slice(0, config.minLines)
+        .filter((line) => line.trim() === '').length;
+    // If we haven't found at least two empty lines, keep looking up to maxLines
+    if (emptyLinesCount < 2 && lines.length > config.minLines) {
+        for (let i = config.minLines; i < Math.min(lines.length, config.maxLines); i++) {
+            if (lines[i].trim() === '') {
+                endLine = i + 1; // Include the empty line
+                break;
+            }
+            endLine = i + 1;
+        }
+    }
+    return lines.slice(0, endLine).join('\n') + '\n... (content truncated)';
+}

package/package.json

@@ -1,9 +1,21 @@
 {
   "name": "genai-lite",
-  "version": "0.1.0",
+  "version": "0.1.1",
   "description": "A lightweight, portable toolkit for interacting with various Generative AI APIs.",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",
+  "exports": {
+    ".": {
+      "import": "./dist/index.js",
+      "require": "./dist/index.js",
+      "types": "./dist/index.d.ts"
+    },
+    "./utils": {
+      "import": "./dist/utils/index.js",
+      "require": "./dist/utils/index.js",
+      "types": "./dist/utils/index.d.ts"
+    }
+  },
   "author": "Luigi Acerbi <luigi.acerbi@gmail.com>",
   "license": "MIT",
   "funding": {
@@ -26,6 +38,7 @@
   "dependencies": {
     "@anthropic-ai/sdk": "^0.52.0",
     "@google/genai": "^1.0.1",
+    "js-tiktoken": "^1.0.20",
     "openai": "^4.103.0"
   },
   "devDependencies": {