torbjorn/my-pal-mcp-server
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: GPT-5.2 support

Browse Source
This commit is contained in:
Fahad
2025-12-11 19:11:50 +00:00
parent 5c3dd75ca6
commit 8b16405f06
33 changed files with 216 additions and 119 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
tests/CASSETTE_MAINTENANCE.md
View File

@@ -228,10 +228,10 @@ Some integration tests maintain cassettes for multiple model variants to ensure
### Consensus Tool Cassettes
The `test_consensus_integration.py` test uses parameterized fixtures to test both `gpt-5` and `gpt-5.1` models:
The `test_consensus_integration.py` test uses parameterized fixtures to test both `gpt-5` and `gpt-5.2` models:
- `tests/openai_cassettes/consensus_step1_gpt5_for.json` - Cassette for gpt-5 model
- `tests/openai_cassettes/consensus_step1_gpt51_for.json` - Cassette for gpt-5.1 model
- `tests/openai_cassettes/consensus_step1_gpt52_for.json` - Cassette for gpt-5.2 model
**When updating consensus cassettes:**
@@ -249,9 +249,9 @@ rm tests/openai_cassettes/consensus_step1_gpt5_for.json
# Run the test with real API key (it will record for gpt-5)
OPENAI_API_KEY="your-real-key" python -m pytest tests/test_consensus_integration.py::test_consensus_multi_model_consultations[gpt-5] -v
# Or for gpt-5.1
rm tests/openai_cassettes/consensus_step1_gpt51_for.json
OPENAI_API_KEY="your-real-key" python -m pytest tests/test_consensus_integration.py::test_consensus_multi_model_consultations[gpt-5.1] -v
# Or for gpt-5.2
rm tests/openai_cassettes/consensus_step1_gpt52_for.json
OPENAI_API_KEY="your-real-key" python -m pytest tests/test_consensus_integration.py::test_consensus_multi_model_consultations[gpt-5.2] -v
```
This dual-coverage approach ensures that both model families continue to work correctly as the codebase evolves.

28
tests/gemini_cassettes/consensus/step2_gemini25_flash_against/mldev.json
View File

File diff suppressed because one or more lines are too long

82
tests/openai_cassettes/consensus_step1_gpt52_for.json Normal file
View File

File diff suppressed because one or more lines are too long

4
tests/test_auto_mode_comprehensive.py
View File

@@ -95,8 +95,8 @@ class TestAutoModeComprehensive:
},
{
"EXTENDED_REASONING": "gpt-5.1-codex", # GPT-5.1 Codex prioritized for coding tasks
"FAST_RESPONSE": "gpt-5.1", # Prefer gpt-5.1 for speed
"BALANCED": "gpt-5.1", # Prefer gpt-5.1 for balanced
"FAST_RESPONSE": "gpt-5.2", # Prefer gpt-5.2 for speed
"BALANCED": "gpt-5.2", # Prefer gpt-5.2 for balanced
},
),
# Only X.AI API available

8
tests/test_auto_mode_model_listing.py
View File

@@ -83,7 +83,7 @@ def test_error_listing_respects_env_restrictions(monkeypatch, reset_registry):
pass
monkeypatch.setenv("GOOGLE_ALLOWED_MODELS", "gemini-2.5-pro")
monkeypatch.setenv("OPENAI_ALLOWED_MODELS", "gpt-5.1")
monkeypatch.setenv("OPENAI_ALLOWED_MODELS", "gpt-5.2")
monkeypatch.setenv("OPENROUTER_ALLOWED_MODELS", "gpt5nano")
monkeypatch.setenv("XAI_ALLOWED_MODELS", "")
@@ -104,7 +104,7 @@ def test_error_listing_respects_env_restrictions(monkeypatch, reset_registry):
("OPENAI_API_KEY", "test-openai"),
("OPENROUTER_API_KEY", "test-openrouter"),
("GOOGLE_ALLOWED_MODELS", "gemini-2.5-pro"),
("OPENAI_ALLOWED_MODELS", "gpt-5.1"),
("OPENAI_ALLOWED_MODELS", "gpt-5.2"),
("OPENROUTER_ALLOWED_MODELS", "gpt5nano"),
("XAI_ALLOWED_MODELS", ""),
):
@@ -139,7 +139,7 @@ def test_error_listing_respects_env_restrictions(monkeypatch, reset_registry):
assert payload["status"] == "error"
available_models = _extract_available_models(payload["content"])
assert set(available_models) == {"gemini-2.5-pro", "gpt-5.1", "gpt5nano", "openai/gpt-5-nano"}
assert set(available_models) == {"gemini-2.5-pro", "gpt-5.2", "gpt5nano", "openai/gpt-5-nano"}
@pytest.mark.no_mock_provider
@@ -225,6 +225,6 @@ def test_error_listing_without_restrictions_shows_full_catalog(monkeypatch, rese
available_models = _extract_available_models(payload["content"])
assert "gemini-2.5-pro" in available_models
assert any(model in available_models for model in {"gpt-5.1", "gpt-5"})
assert any(model in available_models for model in {"gpt-5.2", "gpt-5"})
assert "grok-4" in available_models
assert len(available_models) >= 5

4
tests/test_auto_mode_provider_selection.py
View File

@@ -99,8 +99,8 @@ class TestAutoModeProviderSelection:
# Should select appropriate OpenAI models based on new preference order
assert extended_reasoning == "gpt-5.1-codex" # GPT-5.1 Codex prioritized for extended reasoning
assert fast_response == "gpt-5.1" # gpt-5.1 comes first in fast response preference
assert balanced == "gpt-5.1" # gpt-5.1 for balanced
assert fast_response == "gpt-5.2" # gpt-5.2 comes first in fast response preference
assert balanced == "gpt-5.2" # gpt-5.2 for balanced
finally:
# Restore original environment

6
tests/test_consensus_integration.py
View File

@@ -20,7 +20,7 @@ CASSETTE_DIR.mkdir(exist_ok=True)
# Mapping of OpenAI model names to their cassette files
CONSENSUS_CASSETTES = {
"gpt-5": CASSETTE_DIR / "consensus_step1_gpt5_for.json",
"gpt-5.1": CASSETTE_DIR / "consensus_step1_gpt51_for.json",
"gpt-5.2": CASSETTE_DIR / "consensus_step1_gpt52_for.json",
}
GEMINI_REPLAY_DIR = Path(__file__).parent / "gemini_cassettes"
@@ -32,11 +32,11 @@ GEMINI_REPLAY_PATH = GEMINI_REPLAY_DIR / "consensus" / "step2_gemini25_flash_aga
@pytest.mark.integration
@pytest.mark.asyncio
@pytest.mark.no_mock_provider
@pytest.mark.parametrize("openai_model", ["gpt-5", "gpt-5.1"])
@pytest.mark.parametrize("openai_model", ["gpt-5", "gpt-5.2"])
async def test_consensus_multi_model_consultations(monkeypatch, openai_model):
"""Exercise ConsensusTool against OpenAI model (supporting) and gemini-2.5-flash (critical).
Tests both gpt-5 and gpt-5.1 to ensure regression coverage for both model families.
Tests both gpt-5 and gpt-5.2 to ensure regression coverage for both model families.
"""
# Get the cassette path for this model

8
tests/test_intelligent_fallback.py
View File

@@ -37,14 +37,14 @@ class TestIntelligentFallback:
@patch.dict(os.environ, {"OPENAI_API_KEY": "sk-test-key", "GEMINI_API_KEY": ""}, clear=False)
def test_prefers_openai_o3_mini_when_available(self):
"""Test that gpt-5.1 is preferred when OpenAI API key is available (based on new preference order)"""
"""Test that gpt-5.2 is preferred when OpenAI API key is available (based on new preference order)"""
# Register only OpenAI provider for this test
from providers.openai import OpenAIModelProvider
ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)
fallback_model = ModelProviderRegistry.get_preferred_fallback_model()
assert fallback_model == "gpt-5.1" # Based on new preference order: gpt-5.1 before o4-mini
assert fallback_model == "gpt-5.2" # Based on new preference order: gpt-5.2 before o4-mini
@patch.dict(os.environ, {"OPENAI_API_KEY": "", "GEMINI_API_KEY": "test-gemini-key"}, clear=False)
def test_prefers_gemini_flash_when_openai_unavailable(self):
@@ -147,8 +147,8 @@ class TestIntelligentFallback:
history, tokens = build_conversation_history(context, model_context=None)
# Verify that ModelContext was called with gpt-5.1 (the intelligent fallback based on new preference order)
mock_context_class.assert_called_once_with("gpt-5.1")
# Verify that ModelContext was called with gpt-5.2 (the intelligent fallback based on new preference order)
mock_context_class.assert_called_once_with("gpt-5.2")
def test_auto_mode_with_gemini_only(self):
"""Test auto mode behavior when only Gemini API key is available"""

17
tests/test_openai_provider.py
View File

@@ -50,7 +50,7 @@ class TestOpenAIProvider:
assert provider.validate_model_name("o4-mini") is True
assert provider.validate_model_name("gpt-5") is True
assert provider.validate_model_name("gpt-5-mini") is True
assert provider.validate_model_name("gpt-5.1") is True
assert provider.validate_model_name("gpt-5.2") is True
assert provider.validate_model_name("gpt-5.1-codex") is True
assert provider.validate_model_name("gpt-5.1-codex-mini") is True
@@ -62,6 +62,7 @@ class TestOpenAIProvider:
assert provider.validate_model_name("gpt5") is True
assert provider.validate_model_name("gpt5-mini") is True
assert provider.validate_model_name("gpt5mini") is True
assert provider.validate_model_name("gpt5.2") is True
assert provider.validate_model_name("gpt5.1") is True
assert provider.validate_model_name("gpt5.1-codex") is True
assert provider.validate_model_name("codex-mini") is True
@@ -83,7 +84,8 @@ class TestOpenAIProvider:
assert provider._resolve_model_name("gpt5") == "gpt-5"
assert provider._resolve_model_name("gpt5-mini") == "gpt-5-mini"
assert provider._resolve_model_name("gpt5mini") == "gpt-5-mini"
assert provider._resolve_model_name("gpt5.1") == "gpt-5.1"
assert provider._resolve_model_name("gpt5.2") == "gpt-5.2"
assert provider._resolve_model_name("gpt5.1") == "gpt-5.2"
assert provider._resolve_model_name("gpt5.1-codex") == "gpt-5.1-codex"
assert provider._resolve_model_name("codex-mini") == "gpt-5.1-codex-mini"
@@ -95,7 +97,8 @@ class TestOpenAIProvider:
assert provider._resolve_model_name("o4-mini") == "o4-mini"
assert provider._resolve_model_name("gpt-5") == "gpt-5"
assert provider._resolve_model_name("gpt-5-mini") == "gpt-5-mini"
assert provider._resolve_model_name("gpt-5.1") == "gpt-5.1"
assert provider._resolve_model_name("gpt-5.2") == "gpt-5.2"
assert provider._resolve_model_name("gpt-5.1") == "gpt-5.2"
assert provider._resolve_model_name("gpt-5.1-codex") == "gpt-5.1-codex"
assert provider._resolve_model_name("gpt-5.1-codex-mini") == "gpt-5.1-codex-mini"
@@ -158,12 +161,12 @@ class TestOpenAIProvider:
assert capabilities.supports_function_calling is True
assert capabilities.supports_temperature is True
def test_get_capabilities_gpt51(self):
"""Test GPT-5.1 capabilities reflect new metadata."""
def test_get_capabilities_gpt52(self):
"""Test GPT-5.2 capabilities reflect new metadata."""
provider = OpenAIModelProvider("test-key")
capabilities = provider.get_capabilities("gpt-5.1")
assert capabilities.model_name == "gpt-5.1"
capabilities = provider.get_capabilities("gpt-5.2")
assert capabilities.model_name == "gpt-5.2"
assert capabilities.supports_streaming is True
assert capabilities.supports_function_calling is True
assert capabilities.supports_json_mode is True

10
tests/test_per_tool_model_defaults.py
View File

@@ -133,8 +133,8 @@ class TestModelSelection:
ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)
model = ModelProviderRegistry.get_preferred_fallback_model(ToolModelCategory.FAST_RESPONSE)
# OpenAI now prefers gpt-5.1 for fast response (based on our new preference order)
assert model == "gpt-5.1"
# OpenAI now prefers gpt-5.2 for fast response (based on our new preference order)
assert model == "gpt-5.2"
def test_fast_response_with_gemini_only(self):
"""Test FAST_RESPONSE prefers flash when only Gemini is available."""
@@ -167,8 +167,8 @@ class TestModelSelection:
ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)
model = ModelProviderRegistry.get_preferred_fallback_model(ToolModelCategory.BALANCED)
# OpenAI prefers gpt-5.1 for balanced (based on our new preference order)
assert model == "gpt-5.1"
# OpenAI prefers gpt-5.2 for balanced (based on our new preference order)
assert model == "gpt-5.2"
def test_no_category_uses_balanced_logic(self):
"""Test that no category specified uses balanced logic."""
@@ -209,7 +209,7 @@ class TestFlexibleModelSelection:
"env": {"OPENAI_API_KEY": "test-key"},
"provider_type": ProviderType.OPENAI,
"category": ToolModelCategory.FAST_RESPONSE,
"expected": "gpt-5.1", # Based on new preference order
"expected": "gpt-5.2", # Based on new preference order
},
]

8
tests/test_providers.py
View File

@@ -209,7 +209,7 @@ class TestOpenAIProvider:
assert provider.validate_model_name("o4-mini")
assert provider.validate_model_name("o4mini")
assert provider.validate_model_name("o4-mini")
assert provider.validate_model_name("gpt-5.1")
assert provider.validate_model_name("gpt-5.2")
assert provider.validate_model_name("gpt-5.1-codex")
assert provider.validate_model_name("gpt-5.1-codex-mini")
assert not provider.validate_model_name("gpt-4o")
@@ -223,11 +223,11 @@ class TestOpenAIProvider:
for alias in aliases:
assert not provider.get_capabilities(alias).supports_extended_thinking
def test_gpt51_family_capabilities(self):
"""Ensure GPT-5.1 family exposes correct capability flags."""
def test_gpt52_family_capabilities(self):
"""Ensure GPT-5.2 base model exposes correct capability flags."""
provider = OpenAIModelProvider(api_key="test-key")
base = provider.get_capabilities("gpt-5.1")
base = provider.get_capabilities("gpt-5.2")
assert base.supports_streaming
assert base.allow_code_generation

7
tests/test_supported_models_aliases.py
View File

@@ -54,7 +54,7 @@ class TestSupportedModelsAliases:
assert "o3mini" in provider.MODEL_CAPABILITIES["o3-mini"].aliases
assert "o3pro" in provider.MODEL_CAPABILITIES["o3-pro"].aliases
assert "gpt4.1" in provider.MODEL_CAPABILITIES["gpt-4.1"].aliases
assert "gpt5.1" in provider.MODEL_CAPABILITIES["gpt-5.1"].aliases
assert "gpt5.2" in provider.MODEL_CAPABILITIES["gpt-5.2"].aliases
assert "gpt5.1-codex" in provider.MODEL_CAPABILITIES["gpt-5.1-codex"].aliases
assert "codex-mini" in provider.MODEL_CAPABILITIES["gpt-5.1-codex-mini"].aliases
@@ -64,14 +64,15 @@ class TestSupportedModelsAliases:
assert provider._resolve_model_name("o3pro") == "o3-pro" # o3pro resolves to o3-pro
assert provider._resolve_model_name("o4mini") == "o4-mini"
assert provider._resolve_model_name("gpt4.1") == "gpt-4.1" # gpt4.1 resolves to gpt-4.1
assert provider._resolve_model_name("gpt5.1") == "gpt-5.1"
assert provider._resolve_model_name("gpt5.2") == "gpt-5.2"
assert provider._resolve_model_name("gpt5.1") == "gpt-5.2"
assert provider._resolve_model_name("gpt5.1-codex") == "gpt-5.1-codex"
assert provider._resolve_model_name("codex-mini") == "gpt-5.1-codex-mini"
# Test case insensitive resolution
assert provider._resolve_model_name("Mini") == "gpt-5-mini" # mini -> gpt-5-mini now
assert provider._resolve_model_name("O3MINI") == "o3-mini"
assert provider._resolve_model_name("Gpt5.1") == "gpt-5.1"
assert provider._resolve_model_name("Gpt5.1") == "gpt-5.2"
def test_xai_provider_aliases(self):
"""Test XAI provider's alias structure."""