# Testing Setup
Step-by-step guide to get the project running against live LLM APIs.
## 1. Install Rust
Requires **Rust 1.88+** (edition 2024).
```bash
rustc --version # should print 1.88.0 or later
```
## 2. Build the Workspace
```bash
cargo build --workspace
```
This compiles all workspace crates: core library, adapters, local-llm, memory, eval, and TUI. First build pulls dependencies and takes a few minutes.
## 3. Run Unit Tests
Verify everything compiles and passes before connecting to live APIs:
```bash
cargo test --workspace
```
## 4. Get API Keys
You need at least one provider key. Set up whichever providers you want to test.
### Anthropic (Claude)
1. Go to https://console.anthropic.com/
2. Sign up or log in
3. Navigate to **API Keys** in the left sidebar
4. Click **Create Key**, give it a name, copy the key (`sk-ant-...`)
5. Anthropic offers a free trial tier; paid plans at https://console.anthropic.com/settings/plans
### OpenAI (GPT-4o)
1. Go to https://platform.openai.com/
2. Sign up or log in
3. Navigate to **API Keys** (https://platform.openai.com/api-keys)
4. Click **Create new secret key**, copy the key (`sk-...`)
5. Requires a payment method; see pricing at https://openai.com/api/pricing/
### Ollama (Local, Free)
No API key needed. Install and pull a model:
```bash
# macOS
brew install ollama
# Or download from https://ollama.ai/download
# Start the server
ollama serve
# Pull a model (default: llama3.2, ~2GB)
ollama pull llama3.2
```
Ollama runs entirely on your machine — no account, no API key, no cost.
### Local On-Device (swink-agent-local-llm)
No API key needed. Models are lazily downloaded from HuggingFace on first use and cached in `~/.cache/huggingface/hub/`.
- **SmolLM3-3B** (GGUF Q4_K_M, ~1.92 GB) — text generation, tool use, reasoning
- **EmbeddingGemma-300M** (<200 MB) — text vectorization/embeddings
Context is capped at 8192 tokens by default; override with the `LOCAL_CONTEXT_LENGTH` env var. First run downloads ~2.1 GB of model weights.
```rust
use swink_agent_local_llm::default_local_connection;
let local_connection = default_local_connection()?;
```
### Google Gemini
1. Go to https://aistudio.google.com/
2. Sign up or log in
3. Navigate to **API Keys**
4. Create a key, copy it
### Azure OpenAI
1. Go to https://portal.azure.com/
2. Create an Azure OpenAI resource
3. Deploy a model and note the endpoint URL
4. Copy the API key from the resource's **Keys and Endpoint** section
### xAI (Grok)
1. Go to https://console.x.ai/
2. Sign up or log in
3. Navigate to **API Keys**, create and copy a key
### Mistral
1. Go to https://console.mistral.ai/
2. Sign up or log in
3. Navigate to **API Keys**, create and copy a key
### AWS Bedrock
1. Configure AWS credentials (`AWS_ACCESS_KEY_ID`, `AWS_SECRET_ACCESS_KEY`, and optionally `AWS_SESSION_TOKEN`)
2. Set `AWS_REGION` to a region with Bedrock access (e.g. `us-east-1`)
3. Ensure your IAM role has `bedrock:InvokeModelWithResponseStream` permission
## 5. Configure Environment
```bash
cp .env.example .env
```
Edit `.env` and uncomment/fill in the keys for the providers you set up:
```bash
# For Anthropic testing:
ANTHROPIC_API_KEY=sk-ant-your-key-here
# For OpenAI testing:
OPENAI_API_KEY=sk-your-key-here
# For Ollama (usually no changes needed):
# OLLAMA_HOST=http://localhost:11434
# OLLAMA_MODEL=llama3.2
```
Leave variables commented out to skip that provider. The TUI auto-selects by priority: Proxy > OpenAI > Anthropic > Ollama.
## 6. Launch the TUI
```bash
cargo run -p swink-agent-tui
```
The TUI auto-loads `.env` via dotenvy — no need to source it manually. If no API keys are found (env or keychain), the first-run wizard prompts for provider selection and key entry.
## 7. Verify Each Provider
Once the TUI is running, test each configured provider:
**Check current provider:**
```
#info
```
**Switch model within the current provider:**
```
/model claude-sonnet-4-6
```
Note: `/model` changes the model ID on the active provider. To test a different provider, update `.env` and restart the TUI.
**Test basic conversation:**
```
What is 2 + 2?
```
**Test tool execution (built-in bash tool):**
```
List the files in the current directory
```
**Test thinking mode (Anthropic, Google Gemini, Ollama with supported models):**
```
/thinking medium
Explain why the sky is blue
```
Thinking mode availability by provider:
| Anthropic | Full (`/thinking` levels: minimal, low, medium, high, extra-high) |
| Google Gemini | Supported on deep-think models |
| Ollama | Supported on models that emit `<think>` tags (e.g. DeepSeek) |
| Local (SmolLM3) | Emits `<think>` tags, parsed into thinking events |
| OpenAI, Azure, xAI, Mistral, Bedrock | Not supported |
## 8. Test Scenarios Checklist
| Basic chat | Assistant responds, tokens/cost update in status bar |
| Multi-turn | Context is preserved across messages |
| Tool use | Agent calls bash/read_file/write_file, tool panel shows spinner then result |
| Streaming | Text appears incrementally, not all at once |
| Thinking | Dimmed thinking section appears (Anthropic with `/thinking` enabled) |
| Abort | Press `Esc` during generation — stops cleanly |
| Long output | Conversation scrolls, manual scroll with arrow keys works |
| Session save/load | `#save` then `#load <id>` restores conversation |
| Error recovery | Send a prompt that exceeds context window — agent should recover |
| Model switching | `/model <id>` changes model within current provider, `#info` confirms |
## 9. Logs
If something goes wrong, check the log file:
```bash
cat ~/.config/swink-agent/logs/swink-agent.log
```
Logs are daily rolling files with tracing output. Look for `ERROR` or `WARN` entries.
## 10. Cleanup
API keys stored in the OS keychain persist across sessions. To remove them:
```bash
# macOS — delete from Keychain Access app, search "swink-agent"
# Or from inside the TUI:
#keys # see what's stored
```
To remove local config and sessions:
```bash
rm -rf ~/.config/swink-agent/
```
