llmjs2 1.0.6 → 1.0.8

package/README.md

@@ -1,6 +1,6 @@
 # llmjs2
 
-`llmjs2` is a zero-dependency Node.js library that provides a small, robust interface for calling Ollama and Ollama Cloud from Node 18+.
+`llmjs2` is a lightweight llm wrapper for building simple / personal AI applications.
 
 ## Features
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "llmjs2",
-  "version": "1.0.6",
+  "version": "1.0.8",
   "description": "Minimal zero-dependency Node.js client for Ollama and Ollama Cloud.",
   "type": "module",
   "main": "index.js",
@@ -9,6 +9,11 @@
       "import": "./index.js"
     }
   },
+  "files": [
+    "index.js",
+    "index.d.ts",
+    "README.md"
+  ],
   "types": "index.d.ts",
   "keywords": [
     "llm",

package/spec.txt

@@ -1,73 +0,0 @@
-This specification defines **llmjs2**, a minimalist Node.js library designed to provide a robust, standardized interface for interacting with LLMs. It focuses on a "zero-config" developer experience, initially targeting **Ollama** with a fallback to **Ollama Cloud**.
-
----
-
-## 1. Project Identity
-* **Name:** `llmjs2`
-* **Mission:** To be the most lightweight, concise, and robust bridge between Node.js applications and AI models.
-* **Core Principle:** Favor convention over configuration. Use OpenAI-compatible schemas to minimize the learning curve.
-
----
-
-## 2. Technical Requirements
-* **Runtime:** Node.js 18.0.0+ (Required for native `fetch` and `web streams`).
-* **Module System:** ESM (EcmaScript Modules).
-* **Dependencies:** **Zero.** The library must not depend on external packages to ensure a small footprint and security.
-
----
-
-## 3. Configuration & Environment
-The library automatically resolves connection details using the following hierarchy:
-1.  **Explicit Config:** Passed during initialization (if implemented).
-2.  **Environment Variables:** * `OLLAMA_BASE_URL`: The target API host. 
-    * `OLLAMA_API_KEY`: The bearer token for authenticated proxies or Cloud access.
-3.  **Default Fallback:** `https://api.ollama.com` (Ollama Cloud).
-
----
-
-## 4. API Specification
-
-### The `completion` Function
-The library exports a single overloaded function: `completion`.
-
-**Signatures:**
-* `completion(model: string, prompt: string): Promise<string>`
-* `completion(options: CompletionOptions): Promise<string>`
-
-**Input Object (`CompletionOptions`):**
-| Property | Type | Description |
-| :--- | :--- | :--- |
-| **model** | `string` | Format: `provider/model-name` (e.g., `ollama/llama3`). |
-| **messages** | `Array` | Standard OpenAI `role`/`content` objects. |
-
-**Output:** * Returns a **Promise** that resolves to a **string** containing the assistant's response.
-
----
-
-## 5. Internal Architecture
-
-### A. Provider Routing
-The library utilizes a "prefix-router." It splits the `model` string at the first `/`.
-* If prefix is `ollama`, the request is routed to the `OllamaProvider`.
-* The prefix is stripped before the request is sent to the provider's API.
-
-### B. Request Normalization
-To maintain simplicity, **llmjs2** ignores hyper-parameters like `temperature` or `max_tokens` in the high-level API, allowing the model's internal defaults to govern the output. This ensures the library remains "future-proof" against changing API parameters.
-
-### C. Error Handling
-The library must catch and wrap low-level network errors into high-level, actionable messages:
-* **Connection Error:** "llmjs2: Could not connect to [URL]. Check your OLLAMA_BASE_URL."
-* **Model Error:** "llmjs2: Model [name] not found on provider [provider]."
-
----
-
-## 6. Implementation Checklist
-
-* [ ] **Env Loader:** Logic to check `process.env` and apply fallbacks.
-* [ ] **URL Parser:** Logic to ensure the base URL and `/api/chat` path are joined correctly without double slashes.
-* [ ] **Fetch Wrapper:** A standard `POST` implementation using `Headers` and `body`.
-* [ ] **Response Extractor:** Logic to navigate the JSON response (e.g., `json.message.content`) and return the raw string.
-
----
-
-**Would you like me to generate the actual `index.ts` file that implements this full specification?**

package/test-generate-tools-suite.js

@@ -1,100 +0,0 @@
-import { generate } from './index.js';
-
-const MODEL = 'ollama/qwen3.5:397b-cloud';
-
-const tools = [
-  {
-    name: 'get_weather',
-    description: 'Get the current weather for a location',
-    parameters: {
-      location: {
-        type: 'string',
-        required: true,
-        description: 'The city and state, e.g. San Francisco, CA',
-      },
-      unit: {
-        type: 'string',
-        enum: ['celsius', 'fahrenheit'],
-        description: 'Temperature unit',
-      },
-    },
-    handler: ({ location, unit = 'fahrenheit' }) => {
-      const weatherData = {
-        'San Francisco, CA': { temp: 72, condition: 'Sunny' },
-        'New York, NY': { temp: 45, condition: 'Cloudy' },
-        'London, UK': { temp: 48, condition: 'Rainy' },
-      };
-
-      const data = weatherData[location] || { temp: 70, condition: 'Unknown' };
-      const temp = unit === 'celsius' ? Math.round((data.temp - 32) * (5 / 9)) : data.temp;
-      return `Weather in ${location}: ${temp}°${unit === 'celsius' ? 'C' : 'F'}, ${data.condition}`;
-    },
-  },
-];
-
-const tests = [
-  {
-    name: 'Tool call with default unit',
-    input: {
-      model: MODEL,
-      userPrompt: 'Please use get_weather to fetch the weather for San Francisco, CA.',
-      tools,
-    },
-    expected: /San Francisco, CA/i,
-  },
-  {
-    name: 'Tool call with explicit celsius',
-    input: {
-      model: MODEL,
-      userPrompt: 'Please use get_weather to fetch the weather for London, UK in celsius.',
-      tools,
-    },
-    expected: /London, UK/i,
-  },
-  {
-    name: 'No tool needed direct answer',
-    input: {
-      model: MODEL,
-      userPrompt: 'What is 2 + 2?',
-      tools,
-    },
-    expected: /4|four/i,
-  },
-  {
-    name: 'Explicit messages payload with tool definitions',
-    input: {
-      model: MODEL,
-      messages: [
-        { role: 'system', content: 'You are a tool-aware assistant.' },
-        { role: 'user', content: 'Use get_weather for New York, NY.' },
-      ],
-      tools,
-    },
-    expected: /Weather in New York, NY/i,
-  },
-];
-
-const runTest = async (test) => {
-  console.log(`\n=== ${test.name} ===`);
-  try {
-    const result = await generate(test.input);
-    console.log('Result:');
-    console.log(result);
-    if (test.expected && !test.expected.test(result)) {
-      console.warn('Warning: result did not match expected pattern.');
-    }
-  } catch (error) {
-    console.error('Error:', error?.message ?? error);
-  }
-};
-
-const runAll = async () => {
-  for (const test of tests) {
-    await runTest(test);
-  }
-};
-
-runAll().catch((error) => {
-  console.error('Unexpected failure:', error);
-  process.exit(1);
-});

package/test-generate-tools.js

@@ -1,57 +0,0 @@
-import { generate } from './index.js';
-
-const MODEL = 'ollama/qwen3.5:397b-cloud';
-
-const tools = [
-  {
-    name: 'get_weather',
-    description: 'Get the current weather for a location',
-    parameters: {
-      location: {
-        type: 'string',
-        required: true,
-        description: 'The city and state, e.g. San Francisco, CA',
-      },
-      unit: {
-        type: 'string',
-        enum: ['celsius', 'fahrenheit'],
-        description: 'Temperature unit',
-      },
-    },
-    handler: ({ location, unit = 'fahrenheit' }) => {
-      const weatherData = {
-        'San Francisco, CA': { temp: 72, condition: 'Sunny' },
-        'New York, NY': { temp: 45, condition: 'Cloudy' },
-        'London, UK': { temp: 48, condition: 'Rainy' },
-      };
-
-      const data = weatherData[location] || { temp: 70, condition: 'Unknown' };
-      const temp =
-        unit === 'celsius' ? Math.round((data.temp - 32) * (5 / 9)) : data.temp;
-      return `Weather in ${location}: ${temp}°${unit === 'celsius' ? 'C' : 'F'}, ${data.condition}`;
-    },
-  },
-];
-
-async function run() {
-  const input = {
-    model: MODEL,
-    userPrompt: 'Call get_weather for San Francisco, CA and return the result directly.',
-    images: [],
-    references: ['Use the weather tool if the user asks for weather.'],
-    systemPrompt: 'You are a tool-aware assistant. Use available tools when appropriate.',
-    tools,
-  };
-
-  try {
-    console.log('Running generate() with tool support...');
-    const response = await generate(input);
-    console.log('\nFinal response:');
-    console.log(response);
-  } catch (error) {
-    console.error('generate() failed:', error?.message ?? error);
-    process.exit(1);
-  }
-}
-
-run();

package/test-generate.js

@@ -1,31 +0,0 @@
-import { generate } from './index.js';
-
-const MODEL = 'ollama/qwen3.5:397b-cloud';
-const PROMPT = 'what is in the image? make it concise';
-
-const IMAGES = [
-  'https://thumbs.dreamstime.com/b/sample-jpeg-heartwarming-close-up-endearing-small-mouse-happily-munching-piece-cheese-357411130.jpg',
- // './grapes.jpg',
-];
-
-const REFERENCES = [
- // 'llmjs2 is a zero-dependency Node.js library for Ollama.',
-  //'https://en.wikipedia.org/wiki/History_of_Hong_Kong',
-];
-
-async function run() {
-  console.log('Testing generate() with prompt, images, and references...');
-
-  try {
-    const result = await generate(MODEL, PROMPT, IMAGES, REFERENCES);
-    console.log('generate() result:');
-    console.log(result);
-  } catch (error) {
-    console.error('generate() failed:', error?.message ?? error);
-  }
-}
-
-run().catch((error) => {
-  console.error('Unexpected failure:', error);
-  process.exit(1);
-});

package/test.js

@@ -1,33 +0,0 @@
-import { completion } from './index.js';
-
-const MODEL = process.env.TEST_MODEL || 'ollama/qwen3.5:397b-cloud';
-const PROMPT = 'Hello~';
-
-async function run() {
-  try {
-    console.log('Testing direct completion call...');
-    const result = await completion(MODEL, PROMPT);
-    console.log('Result:', result);
-  } catch (error) {
-    console.error('Direct call failed:', error.message);
-  }
-
-  try {
-    console.log('\nTesting options object call...');
-    const result = await completion({
-      model: MODEL,
-      messages: [
-        { role: 'system', content: 'You are a concise assistant.' },
-        { role: 'user', content: PROMPT },
-      ],
-    });
-    console.log('Result:', result);
-  } catch (error) {
-    console.error('Options call failed:', error.message);
-  }
-}
-
-run().catch((error) => {
-  console.error('Unexpected error:', error);
-  process.exit(1);
-});