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

Quick test mode for simulation tests

Browse Source 
Fixed o4-mini name, OpenAI removed o4-mini-high
Add max_output_tokens property to ModelCapabilities
This commit is contained in:
Fahad
2025-06-23 18:33:47 +04:00
parent 8c1814d4eb
commit ce6c1fd7ea
35 changed files with 137 additions and 110 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
providers/base.py
View File

@@ -132,6 +132,7 @@ class ModelCapabilities:
model_name: str
friendly_name: str # Human-friendly name like "Gemini" or "OpenAI"
context_window: int # Total context window size in tokens
max_output_tokens: int # Maximum output tokens per request
supports_extended_thinking: bool = False
supports_system_prompts: bool = True
supports_streaming: bool = True

7
providers/custom.py
View File

@@ -158,6 +158,7 @@ class CustomProvider(OpenAICompatibleProvider):
model_name=resolved_name,
friendly_name=f"{self.FRIENDLY_NAME} ({resolved_name})",
context_window=32_768, # Conservative default
max_output_tokens=32_768, # Conservative default max output
supports_extended_thinking=False, # Most custom models don't support this
supports_system_prompts=True,
supports_streaming=True,
@@ -187,7 +188,7 @@ class CustomProvider(OpenAICompatibleProvider):
Returns:
True if model is intended for custom/local endpoint
"""
logging.debug(f"Custom provider validating model: '{model_name}'")
# logging.debug(f"Custom provider validating model: '{model_name}'")
# Try to resolve through registry first
config = self._registry.resolve(model_name)
@@ -195,12 +196,12 @@ class CustomProvider(OpenAICompatibleProvider):
model_id = config.model_name
# Use explicit is_custom flag for clean validation
if config.is_custom:
logging.debug(f"Model '{model_name}' -> '{model_id}' validated via registry (custom model)")
logging.debug(f"... [Custom] Model '{model_name}' -> '{model_id}' validated via registry")
return True
else:
# This is a cloud/OpenRouter model - CustomProvider should NOT handle these
# Let OpenRouter provider handle them instead
logging.debug(f"Model '{model_name}' -> '{model_id}' rejected (cloud model, defer to OpenRouter)")
# logging.debug(f"... [Custom] Model '{model_name}' -> '{model_id}' not custom (defer to OpenRouter)")
return False
# Handle version tags for unknown models (e.g., "my-model:latest")

9
providers/dial.py
View File

@@ -37,6 +37,7 @@ class DIALModelProvider(OpenAICompatibleProvider):
model_name="o3-2025-04-16",
friendly_name="DIAL (O3)",
context_window=200_000,
max_output_tokens=100_000,
supports_extended_thinking=False,
supports_system_prompts=True,
supports_streaming=True,
@@ -54,6 +55,7 @@ class DIALModelProvider(OpenAICompatibleProvider):
model_name="o4-mini-2025-04-16",
friendly_name="DIAL (O4-mini)",
context_window=200_000,
max_output_tokens=100_000,
supports_extended_thinking=False,
supports_system_prompts=True,
supports_streaming=True,
@@ -71,6 +73,7 @@ class DIALModelProvider(OpenAICompatibleProvider):
model_name="anthropic.claude-sonnet-4-20250514-v1:0",
friendly_name="DIAL (Sonnet 4)",
context_window=200_000,
max_output_tokens=64_000,
supports_extended_thinking=False,
supports_system_prompts=True,
supports_streaming=True,
@@ -88,6 +91,7 @@ class DIALModelProvider(OpenAICompatibleProvider):
model_name="anthropic.claude-sonnet-4-20250514-v1:0-with-thinking",
friendly_name="DIAL (Sonnet 4 Thinking)",
context_window=200_000,
max_output_tokens=64_000,
supports_extended_thinking=True, # Thinking mode variant
supports_system_prompts=True,
supports_streaming=True,
@@ -105,6 +109,7 @@ class DIALModelProvider(OpenAICompatibleProvider):
model_name="anthropic.claude-opus-4-20250514-v1:0",
friendly_name="DIAL (Opus 4)",
context_window=200_000,
max_output_tokens=64_000,
supports_extended_thinking=False,
supports_system_prompts=True,
supports_streaming=True,
@@ -122,6 +127,7 @@ class DIALModelProvider(OpenAICompatibleProvider):
model_name="anthropic.claude-opus-4-20250514-v1:0-with-thinking",
friendly_name="DIAL (Opus 4 Thinking)",
context_window=200_000,
max_output_tokens=64_000,
supports_extended_thinking=True, # Thinking mode variant
supports_system_prompts=True,
supports_streaming=True,
@@ -139,6 +145,7 @@ class DIALModelProvider(OpenAICompatibleProvider):
model_name="gemini-2.5-pro-preview-03-25-google-search",
friendly_name="DIAL (Gemini 2.5 Pro Search)",
context_window=1_000_000,
max_output_tokens=65_536,
supports_extended_thinking=False, # DIAL doesn't expose thinking mode
supports_system_prompts=True,
supports_streaming=True,
@@ -156,6 +163,7 @@ class DIALModelProvider(OpenAICompatibleProvider):
model_name="gemini-2.5-pro-preview-05-06",
friendly_name="DIAL (Gemini 2.5 Pro)",
context_window=1_000_000,
max_output_tokens=65_536,
supports_extended_thinking=False,
supports_system_prompts=True,
supports_streaming=True,
@@ -173,6 +181,7 @@ class DIALModelProvider(OpenAICompatibleProvider):
model_name="gemini-2.5-flash-preview-05-20",
friendly_name="DIAL (Gemini Flash 2.5)",
context_window=1_000_000,
max_output_tokens=65_536,
supports_extended_thinking=False,
supports_system_prompts=True,
supports_streaming=True,

4
providers/gemini.py
View File

@@ -24,6 +24,7 @@ class GeminiModelProvider(ModelProvider):
model_name="gemini-2.0-flash",
friendly_name="Gemini (Flash 2.0)",
context_window=1_048_576, # 1M tokens
max_output_tokens=65_536,
supports_extended_thinking=True, # Experimental thinking mode
supports_system_prompts=True,
supports_streaming=True,
@@ -42,6 +43,7 @@ class GeminiModelProvider(ModelProvider):
model_name="gemini-2.0-flash-lite",
friendly_name="Gemin (Flash Lite 2.0)",
context_window=1_048_576, # 1M tokens
max_output_tokens=65_536,
supports_extended_thinking=False, # Not supported per user request
supports_system_prompts=True,
supports_streaming=True,
@@ -59,6 +61,7 @@ class GeminiModelProvider(ModelProvider):
model_name="gemini-2.5-flash",
friendly_name="Gemini (Flash 2.5)",
context_window=1_048_576, # 1M tokens
max_output_tokens=65_536,
supports_extended_thinking=True,
supports_system_prompts=True,
supports_streaming=True,
@@ -77,6 +80,7 @@ class GeminiModelProvider(ModelProvider):
model_name="gemini-2.5-pro",
friendly_name="Gemini (Pro 2.5)",
context_window=1_048_576, # 1M tokens
max_output_tokens=65_536,
supports_extended_thinking=True,
supports_system_prompts=True,
supports_streaming=True,

1
providers/openai_compatible.py
View File

@@ -687,7 +687,6 @@ class OpenAICompatibleProvider(ModelProvider):
"o3-mini",
"o3-pro",
"o4-mini",
"o4-mini-high",
# Note: Claude models would be handled by a separate provider
}
supports = model_name.lower() in vision_models

24
providers/openai_provider.py
View File

@@ -24,6 +24,7 @@ class OpenAIModelProvider(OpenAICompatibleProvider):
model_name="o3",
friendly_name="OpenAI (O3)",
context_window=200_000, # 200K tokens
max_output_tokens=65536, # 64K max output tokens
supports_extended_thinking=False,
supports_system_prompts=True,
supports_streaming=True,
@@ -41,6 +42,7 @@ class OpenAIModelProvider(OpenAICompatibleProvider):
model_name="o3-mini",
friendly_name="OpenAI (O3-mini)",
context_window=200_000, # 200K tokens
max_output_tokens=65536, # 64K max output tokens
supports_extended_thinking=False,
supports_system_prompts=True,
supports_streaming=True,
@@ -58,6 +60,7 @@ class OpenAIModelProvider(OpenAICompatibleProvider):
model_name="o3-pro-2025-06-10",
friendly_name="OpenAI (O3-Pro)",
context_window=200_000, # 200K tokens
max_output_tokens=65536, # 64K max output tokens
supports_extended_thinking=False,
supports_system_prompts=True,
supports_streaming=True,
@@ -75,6 +78,7 @@ class OpenAIModelProvider(OpenAICompatibleProvider):
model_name="o4-mini",
friendly_name="OpenAI (O4-mini)",
context_window=200_000, # 200K tokens
max_output_tokens=65536, # 64K max output tokens
supports_extended_thinking=False,
supports_system_prompts=True,
supports_streaming=True,
@@ -85,30 +89,14 @@ class OpenAIModelProvider(OpenAICompatibleProvider):
supports_temperature=False, # O4 models don't accept temperature parameter
temperature_constraint=create_temperature_constraint("fixed"),
description="Latest reasoning model (200K context) - Optimized for shorter contexts, rapid reasoning",
aliases=["mini", "o4mini"],
),
"o4-mini-high": ModelCapabilities(
provider=ProviderType.OPENAI,
model_name="o4-mini-high",
friendly_name="OpenAI (O4-mini-high)",
context_window=200_000, # 200K tokens
supports_extended_thinking=False,
supports_system_prompts=True,
supports_streaming=True,
supports_function_calling=True,
supports_json_mode=True,
supports_images=True, # O4 models support vision
max_image_size_mb=20.0, # 20MB per OpenAI docs
supports_temperature=False, # O4 models don't accept temperature parameter
temperature_constraint=create_temperature_constraint("fixed"),
description="Enhanced O4 mini (200K context) - Higher reasoning effort for complex tasks",
aliases=["o4minihigh", "o4minihi", "mini-high"],
aliases=["mini", "o4mini", "o4-mini"],
),
"gpt-4.1-2025-04-14": ModelCapabilities(
provider=ProviderType.OPENAI,
model_name="gpt-4.1-2025-04-14",
friendly_name="OpenAI (GPT 4.1)",
context_window=1_000_000, # 1M tokens
max_output_tokens=32_768,
supports_extended_thinking=False,
supports_system_prompts=True,
supports_streaming=True,

1
providers/openrouter.py
View File

@@ -101,6 +101,7 @@ class OpenRouterProvider(OpenAICompatibleProvider):
model_name=resolved_name,
friendly_name=self.FRIENDLY_NAME,
context_window=32_768, # Conservative default context window
max_output_tokens=32_768,
supports_extended_thinking=False,
supports_system_prompts=True,
supports_streaming=True,

2
providers/registry.py
View File

@@ -24,8 +24,6 @@ class ModelProviderRegistry:
cls._instance._providers = {}
cls._instance._initialized_providers = {}
logging.debug(f"REGISTRY: Created instance {cls._instance}")
else:
logging.debug(f"REGISTRY: Returning existing instance {cls._instance}")
return cls._instance
@classmethod

2
providers/xai.py
View File

@@ -26,6 +26,7 @@ class XAIModelProvider(OpenAICompatibleProvider):
model_name="grok-3",
friendly_name="X.AI (Grok 3)",
context_window=131_072, # 131K tokens
max_output_tokens=131072,
supports_extended_thinking=False,
supports_system_prompts=True,
supports_streaming=True,
@@ -43,6 +44,7 @@ class XAIModelProvider(OpenAICompatibleProvider):
model_name="grok-3-fast",
friendly_name="X.AI (Grok 3 Fast)",
context_window=131_072, # 131K tokens
max_output_tokens=131072,
supports_extended_thinking=False,
supports_system_prompts=True,
supports_streaming=True,