# TypeDialog Agent Tests

This directory contains test agent files for validating LLM provider integrations and streaming functionality.

## Test Files

- **`test-streaming.agent.mdx`** - Tests streaming LLM responses with Claude
- **`test-openai.agent.mdx`** - Tests OpenAI provider integration
- **`test-gemini.agent.mdx`** - Tests Gemini provider integration

## Running Tests

### Individual Tests

From the project root directory:

```bash
# Test streaming with Claude
cargo run --package typedialog-ag -- tests/agent/test-streaming.agent.mdx

# Test OpenAI provider
cargo run --package typedialog-ag -- tests/agent/test-openai.agent.mdx

# Test Gemini provider
cargo run --package typedialog-ag -- tests/agent/test-gemini.agent.mdx
```

### Requirements

Each test requires the appropriate API key:

```bash
# For Claude/streaming test
export ANTHROPIC_API_KEY=sk-ant-...

# For OpenAI test
export OPENAI_API_KEY=sk-...

# For Gemini test
export GEMINI_API_KEY=...
```

## Unit Tests

For automated testing without API calls, use the mock tests in the core library:

```bash
# Run all core library tests (includes mock streaming tests)
cargo test --package typedialog-ag-core

# Run only LLM provider tests
cargo test --package typedialog-ag-core llm::

# Run with ignored tests (requires API keys)
cargo test --package typedialog-ag-core -- --ignored
```

## Integration Tests

Full integration tests that exercise the complete pipeline:

```bash
# Run integration tests
cargo test --package typedialog-ag-core --test integration_test

# Run simple integration test
cargo test --package typedialog-ag-core --test simple_integration_test
```

## Test Coverage

The test suite covers:

### Mock Tests (No API Keys Required)
- SSE parsing for Claude and OpenAI
- JSON parsing for Gemini and Ollama
- Streaming chunk handling
- Token usage extraction
- Error handling
- Empty input edge cases

### Integration Tests (API Keys Required)
- Real LLM provider calls
- Template rendering
- Agent execution pipeline
- Streaming callbacks
- Validation rules

## Provider-Specific Tests

### Claude (test-streaming.agent.mdx)
Tests:
- SSE streaming format
- Token usage in stream
- Real-time output rendering

### OpenAI (test-openai.agent.mdx)
Tests:
- GPT model integration
- SSE streaming (note: no token usage in stream)
- API error handling

### Gemini (test-gemini.agent.mdx)
Tests:
- JSON streaming format
- Role mapping (assistant → model)
- Token usage in stream

## Adding New Tests

To add a new test agent:

1. Create a `.agent.mdx` file following the existing format
2. Set required configuration in frontmatter:
   ```yaml
   @agent {
     role: test role,
     llm: model-name
   }
   ```
3. Add input directives if needed:
   ```yaml
   @input test_input: String
   ```
4. Document in this README

## Troubleshooting

### API Key Not Set
```
Error: ANTHROPIC_API_KEY environment variable not set
```
**Solution**: Set the appropriate environment variable

### Quota Exceeded
```
Error: 429 Too Many Requests
```
**Solution**: Wait for quota reset or upgrade API plan

### Ollama Not Running
```
Error: Failed to call Ollama API - is Ollama running?
```
**Solution**: Start Ollama server with `ollama serve`

## See Also

- [Demos](../../demos/agent/README.md) - Full-featured demos
- [LLM Integration Docs](../../crates/typedialog-agent/typedialog-ag-core/LLM_INTEGRATION.md) - Complete LLM documentation