uvos/vibe
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

v2.13.0 (#733)

Browse source 
Co-authored-by: Maxime Dolores <maxime.dolores@ext.mistral.ai>
Co-authored-by: Michel Thomazo <51709227+michelTho@users.noreply.github.com>
Co-authored-by: Peter Evers <peter.evers@mistral.ai>
Co-authored-by: Pierre Rossinès <pierre.rossines@mistral.ai>
Co-authored-by: Quentin <quentin.torroba@mistral.ai>
Co-authored-by: Val <102326092+vdeva@users.noreply.github.com>
Co-authored-by: Vincent G <10739306+VinceOPS@users.noreply.github.com>
Co-authored-by: Mistral Vibe <vibe@mistral.ai>
This commit is contained in:
Clément Drouin 2026-05-29 16:17:54 +02:00 • committed by GitHub
parent 198277af3f
commit ad0d5c9520
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
61 changed files with 969 additions and 461 deletions
Download patch file Download diff file Expand all files Collapse all files

93
tests/backend/test_anthropic_adapter.py
View file

@ -315,7 +315,7 @@ class TestAdapterPrepareRequest:
payload = json.loads(req.body)
assert payload["model"] == "claude-sonnet-4-20250514"
assert payload["max_tokens"] == 1024
assert payload["temperature"] == 0.5
assert "temperature" not in payload
assert req.endpoint == "/v1/messages"
assert req.headers["anthropic-version"] == "2023-06-01"
@ -392,9 +392,10 @@ class TestAdapterPrepareRequest:
thinking="medium",
)
payload = json.loads(req.body)
assert payload["thinking"] == {"type": "enabled", "budget_tokens": 10000}
assert payload["thinking"] == {"type": "adaptive", "display": "summarized"}
assert payload["output_config"] == {"effort": "medium"}
assert payload["max_tokens"] == 1024
assert payload["temperature"] == 1
assert "temperature" not in payload
def test_system_cached(self, adapter, provider):
messages = [
@ -439,12 +440,8 @@ class TestAdapterPrepareRequest:
assert len(payload["tools"]) == 1
assert payload["tools"][0]["name"] == "test_tool"
@pytest.mark.parametrize(
"level,expected_budget", [("low", 1024), ("medium", 10_000), ("high", 32_000)]
)
def test_thinking_levels_budget_model(
self, adapter, provider, level, expected_budget
):
@pytest.mark.parametrize("level", ["low", "medium", "high", "max"])
def test_thinking_levels(self, adapter, provider, level):
messages = [LLMMessage(role=Role.user, content="Hello")]
req = adapter.prepare_request(
model_name="claude-sonnet-4-20250514",
@ -458,59 +455,12 @@ class TestAdapterPrepareRequest:
thinking=level,
)
payload = json.loads(req.body)
assert payload["thinking"] == {
"type": "enabled",
"budget_tokens": expected_budget,
}
assert payload["temperature"] == 1
assert payload["max_tokens"] == expected_budget + 8192
@pytest.mark.parametrize(
"model_name", ["claude-opus-4-6-20260101", "claude-opus-4-7-20260418"]
)
@pytest.mark.parametrize("level", ["low", "medium", "high"])
def test_thinking_levels_adaptive_model(self, adapter, provider, model_name, level):
messages = [LLMMessage(role=Role.user, content="Hello")]
req = adapter.prepare_request(
model_name=model_name,
messages=messages,
temperature=0.5,
tools=None,
max_tokens=None,
tool_choice=None,
enable_streaming=False,
provider=provider,
thinking=level,
)
payload = json.loads(req.body)
assert payload["thinking"] == {"type": "adaptive", "display": "summarized"}
assert payload["output_config"] == {"effort": level}
if "opus-4-7" in model_name:
assert "temperature" not in payload
else:
assert payload["temperature"] == 1
assert "temperature" not in payload
assert payload["max_tokens"] == 32_768
@pytest.mark.parametrize("thinking_level", ["off", "low", "medium", "high"])
def test_temperature_omitted_for_deprecated_model(
self, adapter, provider, thinking_level
):
messages = [LLMMessage(role=Role.user, content="Hello")]
req = adapter.prepare_request(
model_name="claude-opus-4-7-20260418",
messages=messages,
temperature=0.5,
tools=None,
max_tokens=None,
tool_choice=None,
enable_streaming=False,
provider=provider,
thinking=thinking_level,
)
payload = json.loads(req.body)
assert "temperature" not in payload
def test_history_forced_thinking_budget_model(self, adapter, provider):
def test_history_forced_thinking(self, adapter, provider):
messages = [
LLMMessage(role=Role.user, content="Hello"),
LLMMessage(
@ -532,34 +482,9 @@ class TestAdapterPrepareRequest:
provider=provider,
)
payload = json.loads(req.body)
assert payload["thinking"] == {"type": "enabled", "budget_tokens": 10_000}
assert payload["temperature"] == 1
assert payload["max_tokens"] == 18_192
def test_history_forced_thinking_adaptive_model(self, adapter, provider):
messages = [
LLMMessage(role=Role.user, content="Hello"),
LLMMessage(
role=Role.assistant,
content="Answer",
reasoning_content="thinking...",
reasoning_signature="sig",
),
LLMMessage(role=Role.user, content="Follow up"),
]
req = adapter.prepare_request(
model_name="claude-opus-4-6-20260101",
messages=messages,
temperature=0.5,
tools=None,
max_tokens=None,
tool_choice=None,
enable_streaming=False,
provider=provider,
)
payload = json.loads(req.body)
assert payload["thinking"] == {"type": "adaptive", "display": "summarized"}
assert payload["output_config"] == {"effort": "medium"}
assert "temperature" not in payload
assert payload["max_tokens"] == 32_768