test: Add extensive coverage for discussion metrics and compression

- Add tests/test_discussion_compression.py to verify AI sub-agent compression logic across Gemini, Anthropic, DeepSeek, and Gemini CLI providers.
- Add tests/test_discussion_metrics.py to verify AppController correctly extracts and accumulates token usage (input/output/cache) and logs token history.

tests/test_discussion_compression.py

@@ -0,0 +1,69 @@
+import pytest
+from unittest.mock import MagicMock, patch
+from src import ai_client
+
+@pytest.fixture(autouse=True)
+def reset_ai_client():
+    ai_client.reset_session()
+    yield
+    ai_client.reset_session()
+
+def test_discussion_compression_gemini():
+    ai_client.set_provider("gemini", "gemini-2.5-flash-lite")
+    
+    mock_gemini = MagicMock()
+    mock_response = MagicMock()
+    mock_response.text = "This is a compressed summary."
+    mock_gemini.models.generate_content.return_value = mock_response
+    
+    with patch("src.ai_client._gemini_client", mock_gemini), \
+         patch("src.ai_client._ensure_gemini_client"):
+        result = ai_client.run_discussion_compression("User: Hello\nAI: Hi there!")
+        
+        assert result == "This is a compressed summary."
+        mock_gemini.models.generate_content.assert_called_once()
+        args, kwargs = mock_gemini.models.generate_content.call_args
+        assert "[HISTORY]" in kwargs["contents"]
+        assert "User: Hello" in kwargs["contents"]
+
+def test_discussion_compression_anthropic():
+    ai_client.set_provider("anthropic", "claude-3-haiku")
+    
+    mock_anthropic = MagicMock()
+    mock_response = MagicMock()
+    mock_block = MagicMock()
+    mock_block.text = "Anthropic summary."
+    mock_response.content = [mock_block]
+    mock_anthropic.messages.create.return_value = mock_response
+    
+    with patch("src.ai_client._anthropic_client", mock_anthropic), \
+         patch("src.ai_client._ensure_anthropic_client"):
+        result = ai_client.run_discussion_compression("Some history")
+        
+        assert result == "Anthropic summary."
+        mock_anthropic.messages.create.assert_called_once()
+        kwargs = mock_anthropic.messages.create.call_args[1]
+        assert "Some history" in kwargs["messages"][0]["content"]
+
+def test_discussion_compression_deepseek():
+    ai_client.set_provider("deepseek", "deepseek-chat")
+    
+    mock_response = MagicMock()
+    mock_response.json.return_value = {"choices": [{"message": {"content": "DeepSeek summary."}}]}
+    
+    with patch("src.ai_client.requests.post", return_value=mock_response), \
+         patch("src.ai_client._load_credentials", return_value={"deepseek": {"api_key": "test"}}):
+        result = ai_client.run_discussion_compression("DeepSeek history")
+        
+        assert result == "DeepSeek summary."
+
+def test_discussion_compression_gemini_cli():
+    ai_client.set_provider("gemini_cli", "gemini-1.5-flash")
+    
+    mock_adapter = MagicMock()
+    mock_adapter.send.return_value = {"text": "CLI summary."}
+    
+    with patch("src.ai_client.GeminiCliAdapter", return_value=mock_adapter):
+        result = ai_client.run_discussion_compression("CLI history")
+        
+        assert result == "CLI summary."

tests/test_discussion_metrics.py

@@ -0,0 +1,63 @@
+import pytest
+from src.app_controller import AppController
+
+@pytest.fixture
+def controller():
+    c = AppController()
+    c.init_state()
+    return c
+
+def test_on_comms_entry_updates_metrics(controller: AppController):
+    # Initial state
+    assert controller.session_usage["input_tokens"] == 0
+    assert controller.session_usage["output_tokens"] == 0
+    assert len(controller._token_history) == 0
+
+    # Simulate an AI response with usage
+    payload = {
+        "text": "Hello world",
+        "usage": {
+            "input_tokens": 100,
+            "output_tokens": 50,
+            "cache_read_input_tokens": 10,
+            "cache_creation_input_tokens": 5,
+            "total_tokens": 165
+        },
+        "model": "test-model"
+    }
+    entry = {
+        "kind": "response",
+        "payload": payload,
+        "ts": "2024-01-01T00:00:00"
+    }
+
+    controller._on_comms_entry(entry)
+
+    # Verify metrics updated
+    assert controller.session_usage["input_tokens"] == 100
+    assert controller.session_usage["output_tokens"] == 50
+    assert controller.session_usage["cache_read_input_tokens"] == 10
+    assert controller.session_usage["cache_creation_input_tokens"] == 5
+    assert controller.session_usage["total_tokens"] == 165
+
+    # Verify token history recorded
+    assert len(controller._token_history) == 1
+    assert controller._token_history[0]["input"] == 100
+    assert controller._token_history[0]["output"] == 50
+    assert controller._token_history[0]["model"] == "test-model"
+
+def test_on_comms_entry_accumulates_metrics(controller: AppController):
+    # First response
+    controller._on_comms_entry({
+        "kind": "response",
+        "payload": {"usage": {"input_tokens": 100, "output_tokens": 50}}
+    })
+    # Second response
+    controller._on_comms_entry({
+        "kind": "response",
+        "payload": {"usage": {"input_tokens": 200, "output_tokens": 20}}
+    })
+
+    assert controller.session_usage["input_tokens"] == 300
+    assert controller.session_usage["output_tokens"] == 70
+    assert len(controller._token_history) == 2