minor update

Author: CodeForgeNet

README.md

@@ -8,7 +8,7 @@
 
 ## ⚡ The Problem
 
-LLMs are non-deterministic. Traditional "Exact Match" testing fails because a model might be 100% correct but use different wording.
+LLMs are non-deterministic. Traditional "Exact Match" testing fails because a model might be 100% correct but use different wording. This makes it difficult to maintain consistent quality and reliability in production applications that depend on LLM outputs.
 
 ## 🚀 The Solution
 
@@ -17,19 +17,28 @@ LLMs are non-deterministic. Traditional "Exact Match" testing fails because a mo
 ### Key Features
 
 * **Semantic Scoring (FREE):** Uses local embeddings to verify output meaning against expectations.
-* **CI/CD Native:** Drops into GitHub Actions or GitLab CI to block regressions.
+* **CI/CD Native:** Integrates seamlessly with GitHub Actions, GitLab CI, and other CI/CD platforms to block regressions.
 * **Watch Mode:** Iterate on prompts in real-time with instant feedback.
+* **Multi-Provider Support:** Works with OpenAI, Anthropic, OpenRouter, and custom LLM endpoints.
 * **Auto-Fix (Premium):** Don't just find errors—fix them. Our engine rewrites failing prompts automatically.
+* **Performance Metrics:** Track latency, token usage, and cost alongside semantic accuracy.
 
 ---
 
 ## 📦 Installation
 
 ```bash
 npm install -g tuneprompt
+```
+
+Or use npx without installation:
 
+```bash
+npx tuneprompt@latest run
 ```
 
+---
+
 ## 🛠️ Quick Start
 
 ### 1. Initialize your project
@@ -38,7 +47,6 @@ This creates a `tuneprompt.config.js` and a sample test directory.
 
 ```bash
 tuneprompt init
-
 ```
 
 ### 2. Define a Test Case (`tests/onboarding.json`)
@@ -53,18 +61,44 @@ tuneprompt init
     "method": "semantic"
   }
 }
-
 ```
 
 ### 3. Run the Suite
 
 ```bash
 tuneprompt run
+```
+
+### 4. Watch Mode for Development
 
+Iterate on prompts with live feedback:
+
+```bash
+tuneprompt watch
 ```
 
 ---
 
+## 🧪 Test Configuration
+
+Tests are defined as JSON files in your test directory. Each test includes:
+
+- `description`: Human-readable description of the test
+- `prompt`: The input prompt to test
+- `expect`: Expected output for semantic comparison
+- `config`: Test-specific configuration options
+
+### Configuration Options
+
+| Option | Type | Description |
+|--------|------|-------------|
+| `threshold` | Number | Semantic similarity threshold (0.0 - 1.0) |
+| `method` | String | Scoring method (`semantic`, `exact`, `regex`) |
+| `provider` | String | LLM provider to use for this test |
+| `timeout` | Number | Request timeout in milliseconds |
+
+---
+
 ## 🤖 Continuous Integration
 
 Ensure prompt integrity on every Pull Request. Add this to `.github/workflows/prompt-test.yml`:
@@ -77,11 +111,28 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v3
+      - name: Setup Node.js
+        uses: actions/setup-node@v3
+        with:
+          node-version: '18'
       - name: Install & Test
         run: |
           npm install -g tuneprompt
           tuneprompt run --ci
+        env:
+          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+```
+
+For GitLab CI, add this to `.gitlab-ci.yml`:
 
+```yaml
+prompt-tests:
+  stage: test
+  script:
+    - npm install -g tuneprompt
+    - tuneprompt run --ci
+  variables:
+    OPENAI_API_KEY: $CI_OPENAI_API_KEY
 ```
 
 ---
@@ -92,7 +143,6 @@ jobs:
 
 ```bash
 tuneprompt fix
-
 ```
 
 The engine analyzes the failure, extracts constraints, and uses iterative meta-prompting to **rewrite your prompt** until it passes the test.
@@ -101,24 +151,102 @@ The engine analyzes the failure, extracts constraints, and uses iterative meta-p
 
 ## 📜 Configuration
 
-`tuneprompt` supports OpenAI, Anthropic, and OpenRouter out of the box.
+Create a `tuneprompt.config.js` file in your project root:
 
 ```javascript
 // tuneprompt.config.js
 module.exports = {
   providers: {
-    openai: { apiKey: process.env.OPENAI_API_KEY, model: 'gpt-4o' },
-    anthropic: { apiKey: process.env.ANTHROPIC_API_KEY, model: 'claude-3-5-sonnet' }
+    openai: {
+      apiKey: process.env.OPENAI_API_KEY,
+      model: 'gpt-4o',
+      temperature: 0.7
+    },
+    anthropic: {
+      apiKey: process.env.ANTHROPIC_API_KEY,
+      model: 'claude-3-5-sonnet'
+    },
+    openrouter: {
+      apiKey: process.env.OPENROUTER_API_KEY,
+      model: 'openai/gpt-4o'
+    }
   },
-  threshold: 0.8 // Default semantic similarity
+  testDir: './tests',           // Directory containing test files
+  threshold: 0.8,              // Default semantic similarity
+  timeout: 30000,              // Default request timeout (ms)
+  concurrency: 5,              // Number of concurrent requests
+  verbose: false               // Enable detailed logging
 };
+```
+
+### Environment Variables
 
+Set these environment variables to configure your LLM providers:
+
+- `OPENAI_API_KEY` - OpenAI API key
+- `ANTHROPIC_API_KEY` - Anthropic API key
+- `OPENROUTER_API_KEY` - OpenRouter API key
+- `TUNEPROMPT_LICENSE_KEY` - Premium license key
+
+---
+
+## 🛠️ CLI Commands
+
+| Command | Description |
+|---------|-------------|
+| `tuneprompt init` | Initialize a new project with config and sample tests |
+| `tuneprompt run` | Run all tests once |
+| `tuneprompt watch` | Watch for changes and run tests automatically |
+| `tuneprompt fix` | Auto-fix failing prompts (premium) |
+| `tuneprompt report` | Generate detailed test reports |
+| `tuneprompt activate` | Activate premium features |
+
+### CLI Options
+
+- `--ci` - CI mode (exits with code 1 on test failure)
+- `--verbose` - Show detailed output
+- `--report` - Generate test reports in various formats
+- `--provider` - Override default provider for this run
+- `--threshold` - Override default threshold for this run
+
+---
+
+## 📊 Reporting
+
+Generate detailed reports in multiple formats:
+
+```bash
+# Generate HTML report
+tuneprompt run --report html
+
+# Generate JSON report
+tuneprompt run --report json
+
+# Generate JUnit XML for CI integration
+tuneprompt run --report junit
 ```
 
 ---
 
+## 🔐 Privacy & Security
+
+- All semantic comparisons happen locally using open-source embedding models
+- Your prompts and expected outputs never leave your machine (unless using premium cloud features)
+- API keys are only sent to the respective LLM providers
+- Premium features offer optional cloud processing for faster results
+
+---
+
+## 🤝 Contributing
+
+We welcome contributions! Please see our [Contributing Guide](CONTRIBUTING.md) for details on how to get started.
+
+---
+
 ## ⚖️ License
 
 MIT © Tuneprompt. Premium features require a license key via `tuneprompt activate`.
 
+For commercial use and enterprise support, contact us at [contact@tuneprompt.com](mailto:contact@tuneprompt.com).
+
 ---