Added Claude Support and Unit Tests

README.md

@@ -51,7 +51,7 @@ This guide will walk you through setting up your API key, downloading a dataset,
 
 ### Step 1: Configure Your API Key
 
-First, tell CARIBOU about your OpenAI or DeepSeek API key. This is a one-time setup.
+First, tell CARIBOU about your OpenAI, Anthropic (Claude), or DeepSeek API key. This is a one-time setup.
 
 ```bash
 caribou config set-openai-key "sk-YourSecretKeyGoesHere"
@@ -63,6 +63,12 @@ or
 caribou config set-deepseek-key "sk-YourSecretKeyGoesHere"
 ```
 
+or
+
+```bash
+caribou config set-anthropic-key "sk-ant-YourSecretKeyGoesHere"
+```
+
 
 Your key will be stored securely in a local `.env` file within the CARIBOU configuration directory.
 
@@ -91,7 +97,7 @@ This will trigger a series of prompts:
 2.  **Select a driver agent:** Choose which agent in the system will receive the first instruction.
 3.  **Select Dataset:** Pick the dataset you downloaded in Step 2.
 4.  **Choose a sandbox backend:** Select `docker` or `singularity`.
-5.  **Choose an LLM backend:** Select `chatgpt` or `ollama`.
+5.  **Choose an LLM backend:** Select `chatgpt`, `claude`, `deepseek`, or `ollama`.
 
 After configuration, the session will begin, and you can start giving instructions to your agent team\!
 
@@ -161,6 +167,10 @@ Manage your CARIBOU configuration.
     ```bash
     caribou config set-deepseek-key "sk-..."
     ```
+  * **Set your Anthropic API key:**
+    ```bash
+    caribou config set-anthropic-key "sk-ant-..."
+    ```
 
 -----
 

caribou/README.md

@@ -51,7 +51,7 @@ This guide will walk you through setting up your API key, downloading a dataset,
 
 ### Step 1: Configure Your API Key
 
-First, tell CARIBOU about your OpenAI or DeepSeek API key. This is a one-time setup.
+First, tell CARIBOU about your OpenAI, Anthropic (Claude), or DeepSeek API key. This is a one-time setup.
 
 ```bash
 caribou config set-openai-key "sk-YourSecretKeyGoesHere"
@@ -63,6 +63,12 @@ or
 caribou config set-deepseek-key "sk-YourSecretKeyGoesHere"
 ```
 
+or
+
+```bash
+caribou config set-anthropic-key "sk-ant-YourSecretKeyGoesHere"
+```
+
 
 Your key will be stored securely in a local `.env` file within the CARIBOU configuration directory.
 
@@ -91,7 +97,7 @@ This will trigger a series of prompts:
 2.  **Select a driver agent:** Choose which agent in the system will receive the first instruction.
 3.  **Select Dataset:** Pick the dataset you downloaded in Step 2.
 4.  **Choose a sandbox backend:** Select `docker` or `singularity`.
-5.  **Choose an LLM backend:** Select `chatgpt` or `ollama`.
+5.  **Choose an LLM backend:** Select `chatgpt`, `claude`, `deepseek`, or `ollama`.
 
 After configuration, the session will begin, and you can start giving instructions to your agent team\!
 
@@ -161,6 +167,10 @@ Manage your CARIBOU configuration.
     ```bash
     caribou config set-deepseek-key "sk-..."
     ```
+  * **Set your Anthropic API key:**
+    ```bash
+    caribou config set-anthropic-key "sk-ant-..."
+    ```
 
 -----
 

caribou/src/caribou/cli/config_cli.py

@@ -71,4 +71,36 @@ def set_deepseek_key(
         new_content = content.strip() + f"\n{key_to_set}\n"
 
     ENV_FILE.write_text(new_content)
-    console.print(f"[bold green]✅ DeepSeek API key has been set successfully in:[/bold green] {ENV_FILE}")
+    console.print(f"[bold green]✅ DeepSeek API key has been set successfully in:[/bold green] {ENV_FILE}")
+
+@config_app.command("set-anthropic-key")
+def set_anthropic_key(
+    ctx: typer.Context,
+    api_key: Optional[str] = typer.Argument(None, help="Your Anthropic API key (e.g., 'sk-ant-...')"),
+):
+    """
+    Saves your Anthropic API key to the Caribou environment file.
+    """
+    if api_key is None:
+        console.print("[bold red]Error:[/bold red] You must provide an API key.\n")
+        typer.echo(ctx.parent.get_help())
+        raise typer.Exit()
+
+    if not api_key.startswith("sk-"):
+        console.print(
+            "[yellow]Warning: Key does not look like a standard Anthropic API key (should start with 'sk-').[/yellow]"
+        )
+
+    if not ENV_FILE.exists():
+        ENV_FILE.touch()
+
+    content = ENV_FILE.read_text()
+    key_to_set = f'ANTHROPIC_API_KEY="{api_key}"'
+
+    if re.search(r"^ANTHROPIC_API_KEY=.*$", content, flags=re.MULTILINE):
+        new_content = re.sub(r"^ANTHROPIC_API_KEY=.*$", key_to_set, content, flags=re.MULTILINE)
+    else:
+        new_content = content.strip() + f"\n{key_to_set}\n"
+
+    ENV_FILE.write_text(new_content.strip())
+    console.print(f"[bold green]✅ Anthropic API key has been set successfully in:[/bold green] {ENV_FILE}")

caribou/src/caribou/cli/run_cli.py

@@ -370,7 +370,7 @@ def initialize_context(
 
     # ---- LLM Backend ----
     if llm_backend is None:
-        llm_backend = Prompt.ask("Choose an LLM backend", choices=["chatgpt", "ollama", "deepseek"], default="chatgpt")
+        llm_backend = Prompt.ask("Choose an LLM backend", choices=["chatgpt", "claude", "ollama", "deepseek"], default="chatgpt")
 
     console.print(f"[cyan]Initializing LLM backend: {llm_backend}[/cyan]")
 
@@ -380,7 +380,14 @@ def initialize_context(
             raise typer.Exit(1)
         context.llm_client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
         context.model_name = "gpt-5.2"
-
+    elif llm_backend == "claude":
+        anthropic_key = os.getenv("ANTHROPIC_API_KEY")
+        if not anthropic_key:
+            console.print("[bold red]Error: ANTHROPIC_API_KEY not set. Use 'caribou config set-anthropic-key'.[/bold red]")
+            raise typer.Exit(1)
+        from caribou.core.anthropic_wrapper import AnthropicClient
+        context.llm_client = AnthropicClient(api_key=anthropic_key)
+        context.model_name = "claude-sonnet-4-5-20250929"
     elif llm_backend == "deepseek":
         if not os.getenv("DEEPSEEK_API_KEY"):
             console.print("[bold red]Error: DEEPSEEK_API_KEY not set. Use 'caribou config set-deepseek-key'.[/bold red]")
@@ -459,7 +466,7 @@ def main_run_callback(
     dataset: Optional[Path] = typer.Option(None, "--dataset", "-ds", help="Path to the primary dataset file (.h5ad).", readable=True),
     reference_dataset: Optional[Path] = typer.Option(None, "--reference-dataset", "-ref", help="Path to an optional reference dataset file (.h5ad).", readable=True),
     resources_dir: Optional[Path] = typer.Option(None, "--resources", help="Path to a directory of resource files to mount.", exists=True, file_okay=False),
-    llm_backend: Optional[str] = typer.Option(None, "--llm", help="LLM backend: 'chatgpt', 'ollama', or 'deepseek'."),
+    llm_backend: Optional[str] = typer.Option(None, "--llm", help="LLM backend: 'chatgpt', 'claude', 'ollama', or 'deepseek'."),
     ollama_host: str = typer.Option("http://localhost:11434", "--ollama-host", help="Base URL for Ollama backend."),
     sandbox: Optional[str] = typer.Option(None, "--sandbox", help="Sandbox backend: 'docker' or 'singularity'."),
     force_refresh: bool = typer.Option(False, "--force-refresh", help="Force refresh/rebuild of the sandbox environment."),
@@ -515,7 +522,7 @@ def run_interactive(
     dataset: Path = typer.Option(None, "--dataset", "-ds", help="Path to the primary dataset file (.h5ad).", readable=True),
     reference_dataset: Path = typer.Option(None, "--reference-dataset", "-ref", help="Path to an optional reference dataset file (.h5ad).", readable=True),
     resources_dir: Path = typer.Option(None, "--resources", help="Path to a directory of resource files to mount.", exists=True, file_okay=False),
-    llm_backend: str = typer.Option(None, "--llm", help="LLM backend to use: 'chatgpt', 'ollama', or 'deepseek'."),
+    llm_backend: str = typer.Option(None, "--llm", help="LLM backend to use: 'chatgpt', 'claude', 'ollama', or 'deepseek'."),
     ollama_host: str = typer.Option("http://localhost:11434", "--ollama-host", help="Base URL for Ollama backend."),
     sandbox: str = typer.Option(None, "--sandbox", help="Sandbox backend to use: 'docker' or 'singularity'."),
     force_refresh: bool = typer.Option(False, "--force-refresh", help="Force refresh/rebuild of the sandbox environment."),
@@ -560,7 +567,7 @@ def run_auto(
     dataset: Path = typer.Option(None, "--dataset", "-ds", help="Path to the primary dataset file (.h5ad).", readable=True),
     reference_dataset: Path = typer.Option(None, "--reference-dataset", "-ref", help="Path to an optional reference dataset file (.h5ad).", readable=True),
     resources_dir: Path = typer.Option(None, "--resources", help="Path to a directory of resource files to mount.", exists=True, file_okay=False),
-    llm_backend: str = typer.Option(None, "--llm", help="LLM backend to use: 'chatgpt', 'ollama', or 'deepseek'."),
+    llm_backend: str = typer.Option(None, "--llm", help="LLM backend to use: 'chatgpt', 'claude', 'ollama', or 'deepseek'."),
     ollama_host: str = typer.Option("http://localhost:11434", "--ollama-host", help="Base URL for Ollama backend."),
     sandbox: str = typer.Option(None, "--sandbox", help="Sandbox backend to use: 'docker' or 'singularity'."),
     force_refresh: bool = typer.Option(False, "--force-refresh", help="Force refresh/rebuild of the sandbox environment."),

caribou/src/caribou/core/anthropic_wrapper.py

@@ -0,0 +1,75 @@
+"""
+Lightweight Anthropic client wrapper that mimics the subset of the OpenAI
+chat API used by CARIBOU. Exposes a `.chat.completions.create(...)` method
+that returns an object shaped like the OpenAI response:
+
+    resp.choices[0].message.content
+"""
+
+from __future__ import annotations
+
+from types import SimpleNamespace
+from typing import Any, Dict, List, Optional
+
+import anthropic
+
+
+class AnthropicClient:
+    """
+    Example:
+        client = AnthropicClient(api_key="sk-ant-...", model="claude-sonnet-4-5-20250929")
+        resp = client.chat.completions.create(model="claude-sonnet-4-5-20250929", messages=[...])
+        print(resp.choices[0].message.content)
+    """
+
+    def __init__(
+        self,
+        *,
+        api_key: str,
+        model: str = "claude-sonnet-4-5-20250929",
+        max_output_tokens: int = 1024,
+        base_url: Optional[str] = None,
+    ):
+        client_kwargs: Dict[str, Any] = {"api_key": api_key}
+        if base_url:
+            client_kwargs["base_url"] = base_url
+        self._client = anthropic.Anthropic(**client_kwargs)
+        self._default_model = model
+        self._max_output_tokens = max_output_tokens
+        self.chat = SimpleNamespace(completions=SimpleNamespace(create=self._chat_create))
+
+    def _chat_create(
+        self,
+        *,
+        messages: List[Dict[str, str]],
+        model: Optional[str] = None,
+        temperature: Optional[float] = None,
+        max_output_tokens: Optional[int] = None,
+        **_: Any,
+    ):
+        system_parts: List[str] = []
+        converted: List[Dict[str, str]] = []
+        for msg in messages:
+            role = msg.get("role", "user")
+            content = msg.get("content", "")
+            if role == "system":
+                system_parts.append(str(content))
+                continue
+            converted.append({"role": role if role in ("assistant", "user") else "user", "content": content})
+
+        system_prompt = "\n\n".join(system_parts) if system_parts else None
+
+        response = self._client.messages.create(
+            model=model or self._default_model,
+            system=system_prompt,
+            messages=converted,
+            temperature=temperature,
+            max_tokens=max_output_tokens or self._max_output_tokens,
+        )
+
+        text_chunks = [block.text for block in response.content if getattr(block, "type", None) == "text"]
+        content = "".join(text_chunks)
+
+        message = SimpleNamespace(content=content, role="assistant")
+        choice = SimpleNamespace(message=message, index=0, finish_reason=getattr(response, "stop_reason", "stop"))
+        return SimpleNamespace(choices=[choice])

caribou/tests/QUICKSTART.md

@@ -0,0 +1,127 @@
+# CARIBOU Test Suite - Quick Start Guide
+
+## Installation
+
+1. **Activate your conda environment (if using conda):**
+
+```bash
+conda activate olaf  # or your environment name
+```
+
+2. **Install test dependencies:**
+
+```bash
+cd /data1/peerd/riffled/riffled/Olaf_project/CARIBOU
+python -m pip install pytest pytest-cov
+```
+
+Or install all requirements:
+
+```bash
+python -m pip install -r requirements.txt
+```
+
+**Important:** Make sure pytest is installed in the same Python environment where `anthropic`, `openai`, and other CARIBOU dependencies are installed.
+
+## Running Tests
+
+### Option 1: Using the Test Runner Script (Recommended)
+
+```bash
+cd caribou/tests
+./run_tests.sh
+```
+
+Available options:
+- `./run_tests.sh --unit` - Run only unit tests
+- `./run_tests.sh --integration` - Run only integration tests
+- `./run_tests.sh --verbose` - Verbose output
+- `./run_tests.sh --coverage` - Generate coverage report
+
+### Option 2: Using pytest Directly
+
+From the project root directory:
+
+```bash
+# Run all tests
+pytest caribou/tests/
+
+# Run specific test categories
+pytest caribou/tests/unit/
+pytest caribou/tests/integration/
+
+# Run specific test file
+pytest caribou/tests/unit/test_message_utils.py
+
+# Run with verbose output
+pytest caribou/tests/ -v
+
+# Run with coverage
+pytest caribou/tests/ --cov=caribou --cov-report=html
+```
+
+## What Gets Tested
+
+✅ **LLM API Wrappers**
+- AnthropicClient (OpenAI compatibility)
+- OllamaClient (local models)
+
+✅ **Message Routing**
+- Delegation detection (`delegate_to_agent`)
+- RAG query detection (`query_rag_<topic>`)
+- Artifact extraction (notes, TODOs)
+
+✅ **History Management**
+- MemoryManager with episodic summarization
+- Context assembly and compression
+
+✅ **Agent System**
+- Multi-agent configuration
+- Prompt generation
+- Agent switching
+
+✅ **End-to-End Integration**
+- Complete message flows
+- Multi-agent conversations
+- Error handling
+
+## Verifying the Setup
+
+Run a quick smoke test:
+
+```bash
+pytest caribou/tests/unit/test_message_utils.py::TestDelegationDetection::test_detect_simple_delegation -v
+```
+
+Expected output:
+```
+test_detect_simple_delegation PASSED
+```
+
+## Troubleshooting
+
+**Import errors?**
+The tests automatically add `caribou/src` to the Python path via `conftest.py`. If you still get import errors, verify the directory structure:
+
+```
+CARIBOU/
+└── caribou/
+    ├── src/
+    │   └── caribou/
+    │       ├── core/
+    │       ├── execution/
+    │       └── agents/
+    └── tests/
+        ├── conftest.py  # ← Should add src/ to path
+        ├── unit/
+        └── integration/
+```
+
+**Tests hanging?**
+All external API calls are mocked - tests should run quickly (< 10 seconds total).
+
+## Next Steps
+
+- See [README.md](README.md) for detailed documentation
+- Run with coverage to see what's tested: `./run_tests.sh --coverage`
+- Open `htmlcov/index.html` to view the coverage report