feat(llm): full multi-tier S/A/B/C/D with routing + pricing

Estende ModelTier a 5 livelli (S/A/B/C/D) con routing automatico: S/A/B via Anthropic SDK, C/D via OpenRouter (OpenAI SDK). Aggiunge prezzi per tier S (Opus), A (Sonnet placeholder) e D (Llama). Refactor LLMClient.complete con dispatch tramite tier_models map e helper _call_anthropic / _call_openrouter. Settings esposte per tutti e 5 i modelli env-configurabili. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-10 09:18:57 +02:00
parent 7482600146
commit 33d8e275e7
10 changed files with 241 additions and 36 deletions
@@ -121,6 +121,94 @@ def test_completion_uses_custom_model_tier_b(mocker):
    assert out.model == "claude-opus-4-7"


+def test_completion_tier_s_uses_anthropic_with_opus(mocker):
+    fake_anthropic = mocker.MagicMock()
+    fake_msg = mocker.MagicMock()
+    fake_msg.content = [mocker.MagicMock(text="(strategy s)")]
+    fake_msg.usage = mocker.MagicMock(input_tokens=50, output_tokens=100)
+    fake_anthropic.messages.create.return_value = fake_msg
+    mocker.patch("multi_swarm.llm.client.Anthropic", return_value=fake_anthropic)
+
+    client = LLMClient(openrouter_api_key="or-x", anthropic_api_key="an-x")
+    g = make_genome(ModelTier.S)
+    out = client.complete(g, system="sys", user="usr")
+
+    fake_anthropic.messages.create.assert_called_once()
+    call_kwargs = fake_anthropic.messages.create.call_args.kwargs
+    assert call_kwargs["model"] == "claude-opus-4-7"
+    assert out.tier == ModelTier.S
+    assert out.model == "claude-opus-4-7"
+
+
+def test_completion_tier_a_uses_anthropic_with_sonnet(mocker):
+    fake_anthropic = mocker.MagicMock()
+    fake_msg = mocker.MagicMock()
+    fake_msg.content = [mocker.MagicMock(text="(strategy a)")]
+    fake_msg.usage = mocker.MagicMock(input_tokens=40, output_tokens=80)
+    fake_anthropic.messages.create.return_value = fake_msg
+    mocker.patch("multi_swarm.llm.client.Anthropic", return_value=fake_anthropic)
+
+    client = LLMClient(openrouter_api_key="or-x", anthropic_api_key="an-x")
+    g = make_genome(ModelTier.A)
+    out = client.complete(g, system="sys", user="usr")
+
+    fake_anthropic.messages.create.assert_called_once()
+    call_kwargs = fake_anthropic.messages.create.call_args.kwargs
+    assert call_kwargs["model"] == "claude-sonnet-4-6"
+    assert out.tier == ModelTier.A
+    assert out.model == "claude-sonnet-4-6"
+
+
+def test_completion_tier_d_uses_openrouter_with_llama(mocker):
+    fake_openai = mocker.MagicMock()
+    fake_response = mocker.MagicMock()
+    fake_response.choices = [
+        mocker.MagicMock(message=mocker.MagicMock(content="(strategy d)"))
+    ]
+    fake_response.usage = mocker.MagicMock(prompt_tokens=30, completion_tokens=70)
+    fake_openai.chat.completions.create.return_value = fake_response
+    mocker.patch("multi_swarm.llm.client.OpenAI", return_value=fake_openai)
+
+    client = LLMClient(openrouter_api_key="or-x", anthropic_api_key=None)
+    g = make_genome(ModelTier.D)
+    out = client.complete(g, system="sys", user="usr")
+
+    fake_openai.chat.completions.create.assert_called_once()
+    call_kwargs = fake_openai.chat.completions.create.call_args.kwargs
+    assert call_kwargs["model"] == "meta-llama/llama-3.3-70b-instruct"
+    assert out.tier == ModelTier.D
+    assert out.model == "meta-llama/llama-3.3-70b-instruct"
+
+
+def test_completion_uses_custom_model_tier_s(mocker):
+    fake_anthropic = mocker.MagicMock()
+    fake_msg = mocker.MagicMock()
+    fake_msg.content = [mocker.MagicMock(text="(strategy custom-s)")]
+    fake_msg.usage = mocker.MagicMock(input_tokens=10, output_tokens=20)
+    fake_anthropic.messages.create.return_value = fake_msg
+    mocker.patch("multi_swarm.llm.client.Anthropic", return_value=fake_anthropic)
+
+    client = LLMClient(
+        openrouter_api_key="or-x",
+        anthropic_api_key="an-x",
+        model_tier_s="claude-future-mega",
+    )
+    g = make_genome(ModelTier.S)
+    out = client.complete(g, system="sys", user="usr")
+
+    call_kwargs = fake_anthropic.messages.create.call_args.kwargs
+    assert call_kwargs["model"] == "claude-future-mega"
+    assert out.model == "claude-future-mega"
+
+
+def test_completion_tier_s_without_anthropic_key_raises(mocker):
+    mocker.patch("multi_swarm.llm.client.OpenAI", return_value=mocker.MagicMock())
+    client = LLMClient(openrouter_api_key="or-x", anthropic_api_key=None)
+    g = make_genome(ModelTier.S)
+    with pytest.raises(RuntimeError, match="tier S"):
+        client.complete(g, system="sys", user="usr")
+
+
@pytest.mark.slow
 def test_completion_succeeds_after_one_retry(mocker):
    """Dopo 1 fallimento transient, il retry riesce al 2 tentativo."""