Schema now lists all models including locally available models

New tool to list all models `listmodels`
Integration test to for all the different combinations of API keys
Tweaks to codereview prompt for a better quality input from Claude
Fixed missing 'low' severity in codereview
This commit is contained in:
Fahad
2025-06-16 19:07:35 +04:00
parent cb17582d8f
commit 70b64adff3
10 changed files with 822 additions and 24 deletions

View File

@@ -6,6 +6,7 @@ from .analyze import AnalyzeTool
from .chat import ChatTool
from .codereview import CodeReviewTool
from .debug import DebugIssueTool
from .listmodels import ListModelsTool
from .precommit import Precommit
from .refactor import RefactorTool
from .testgen import TestGenerationTool
@@ -18,6 +19,7 @@ __all__ = [
"DebugIssueTool",
"AnalyzeTool",
"ChatTool",
"ListModelsTool",
"Precommit",
"RefactorTool",
"TestGenerationTool",

View File

@@ -99,6 +99,9 @@ class ToolRequest(BaseModel):
class BaseTool(ABC):
# Class-level cache for OpenRouter registry to avoid multiple loads
_openrouter_registry_cache = None
"""
Abstract base class for all Gemini tools.
@@ -210,6 +213,20 @@ class BaseTool(ABC):
"""
pass
@classmethod
def _get_openrouter_registry(cls):
"""Get cached OpenRouter registry instance."""
if BaseTool._openrouter_registry_cache is None:
import logging
from providers.openrouter_registry import OpenRouterModelRegistry
logger = logging.getLogger(__name__)
logger.info("Loading OpenRouter registry for the first time (will be cached for all tools)")
BaseTool._openrouter_registry_cache = OpenRouterModelRegistry()
return BaseTool._openrouter_registry_cache
def is_effective_auto_mode(self) -> bool:
"""
Check if we're in effective auto mode for schema generation.

View File

@@ -39,21 +39,32 @@ class CodeReviewRequest(ToolRequest):
)
prompt: str = Field(
...,
description="User's summary of what the code does, expected behavior, constraints, and review objectives",
description=(
"User's summary of what the code does, expected behavior, constraints, and review objectives. "
"IMPORTANT: Before using this tool, Claude should first perform its own preliminary review - "
"examining the code structure, identifying potential issues, understanding the business logic, "
"and noting areas of concern. Include Claude's initial observations about code quality, potential "
"bugs, architectural patterns, and specific areas that need deeper scrutiny. This dual-perspective "
"approach (Claude's analysis + external model's review) provides more comprehensive feedback and "
"catches issues that either reviewer might miss alone."
),
)
images: Optional[list[str]] = Field(
None,
description="Optional images of architecture diagrams, UI mockups, design documents, or visual references for code review context",
description=(
"Optional images of architecture diagrams, UI mockups, design documents, or visual references "
"for code review context"
),
)
review_type: str = Field("full", description="Type of review: full|security|performance|quick")
focus_on: Optional[str] = Field(
None,
description="Specific aspects to focus on, or additional context that would help understand areas of concern",
description=("Specific aspects to focus on, or additional context that would help understand areas of concern"),
)
standards: Optional[str] = Field(None, description="Coding standards or guidelines to enforce")
severity_filter: str = Field(
"all",
description="Minimum severity to report: critical|high|medium|all",
description="Minimum severity to report: critical|high|medium|low|all",
)
@@ -81,7 +92,8 @@ class CodeReviewTool(BaseTool):
"Choose thinking_mode based on review scope: 'low' for small code snippets, "
"'medium' for standard files/modules (default), 'high' for complex systems/architectures, "
"'max' for critical security audits or large codebases requiring deepest analysis. "
"Note: If you're not currently using a top-tier model such as Opus 4 or above, these tools can provide enhanced capabilities."
"Note: If you're not currently using a top-tier model such as Opus 4 or above, these tools "
"can provide enhanced capabilities."
)
def get_input_schema(self) -> dict[str, Any]:
@@ -96,12 +108,24 @@ class CodeReviewTool(BaseTool):
"model": self.get_model_field_schema(),
"prompt": {
"type": "string",
"description": "User's summary of what the code does, expected behavior, constraints, and review objectives",
"description": (
"User's summary of what the code does, expected behavior, constraints, and review "
"objectives. IMPORTANT: Before using this tool, Claude should first perform its own "
"preliminary review - examining the code structure, identifying potential issues, "
"understanding the business logic, and noting areas of concern. Include Claude's initial "
"observations about code quality, potential bugs, architectural patterns, and specific "
"areas that need deeper scrutiny. This dual-perspective approach (Claude's analysis + "
"external model's review) provides more comprehensive feedback and catches issues that "
"either reviewer might miss alone."
),
},
"images": {
"type": "array",
"items": {"type": "string"},
"description": "Optional images of architecture diagrams, UI mockups, design documents, or visual references for code review context",
"description": (
"Optional images of architecture diagrams, UI mockups, design documents, or visual "
"references for code review context"
),
},
"review_type": {
"type": "string",
@@ -111,7 +135,10 @@ class CodeReviewTool(BaseTool):
},
"focus_on": {
"type": "string",
"description": "Specific aspects to focus on, or additional context that would help understand areas of concern",
"description": (
"Specific aspects to focus on, or additional context that would help understand "
"areas of concern"
),
},
"standards": {
"type": "string",
@@ -119,7 +146,7 @@ class CodeReviewTool(BaseTool):
},
"severity_filter": {
"type": "string",
"enum": ["critical", "high", "medium", "all"],
"enum": ["critical", "high", "medium", "low", "all"],
"default": "all",
"description": "Minimum severity level to report",
},
@@ -132,16 +159,29 @@ class CodeReviewTool(BaseTool):
"thinking_mode": {
"type": "string",
"enum": ["minimal", "low", "medium", "high", "max"],
"description": "Thinking depth: minimal (0.5% of model max), low (8%), medium (33%), high (67%), max (100% of model max)",
"description": (
"Thinking depth: minimal (0.5% of model max), low (8%), medium (33%), high (67%), "
"max (100% of model max)"
),
},
"use_websearch": {
"type": "boolean",
"description": "Enable web search for documentation, best practices, and current information. Particularly useful for: brainstorming sessions, architectural design discussions, exploring industry best practices, working with specific frameworks/technologies, researching solutions to complex problems, or when current documentation and community insights would enhance the analysis.",
"description": (
"Enable web search for documentation, best practices, and current information. "
"Particularly useful for: brainstorming sessions, architectural design discussions, "
"exploring industry best practices, working with specific frameworks/technologies, "
"researching solutions to complex problems, or when current documentation and community "
"insights would enhance the analysis."
),
"default": True,
},
"continuation_id": {
"type": "string",
"description": "Thread continuation ID for multi-turn conversations. Can be used to continue conversations across different tools. Only provide this if continuing a previous conversation thread.",
"description": (
"Thread continuation ID for multi-turn conversations. Can be used to continue "
"conversations across different tools. Only provide this if continuing a previous "
"conversation thread."
),
},
},
"required": ["files", "prompt"] + (["model"] if self.is_effective_auto_mode() else []),
@@ -263,7 +303,8 @@ class CodeReviewTool(BaseTool):
{file_content}
=== END CODE ===
Please provide a code review aligned with the user's context and expectations, following the format specified in the system prompt."""
Please provide a code review aligned with the user's context and expectations, following the format specified """
"in the system prompt." ""
return full_prompt
@@ -285,10 +326,14 @@ Please provide a code review aligned with the user's context and expectations, f
**Claude's Next Steps:**
1. **Understand the Context**: First examine the specific functions, files, and code sections mentioned in the review to understand each issue thoroughly.
1. **Understand the Context**: First examine the specific functions, files, and code sections mentioned in """
"""the review to understand each issue thoroughly.
2. **Present Options to User**: After understanding the issues, ask the user which specific improvements they would like to implement, presenting them as a clear list of options.
2. **Present Options to User**: After understanding the issues, ask the user which specific improvements """
"""they would like to implement, presenting them as a clear list of options.
3. **Implement Selected Fixes**: Only implement the fixes the user chooses, ensuring each change is made correctly and maintains code quality.
3. **Implement Selected Fixes**: Only implement the fixes the user chooses, ensuring each change is made """
"""correctly and maintains code quality.
Remember: Always understand the code context before suggesting fixes, and let the user decide which improvements to implement."""
Remember: Always understand the code context before suggesting fixes, and let the user decide which """
"""improvements to implement."""

279
tools/listmodels.py Normal file
View File

@@ -0,0 +1,279 @@
"""
List Models Tool - Display all available models organized by provider
This tool provides a comprehensive view of all AI models available in the system,
organized by their provider (Gemini, OpenAI, X.AI, OpenRouter, Custom).
It shows which providers are configured and what models can be used.
"""
import os
from typing import Any, Optional
from mcp.types import TextContent
from tools.base import BaseTool, ToolRequest
from tools.models import ToolModelCategory, ToolOutput
class ListModelsTool(BaseTool):
"""
Tool for listing all available AI models organized by provider.
This tool helps users understand:
- Which providers are configured (have API keys)
- What models are available from each provider
- Model aliases and their full names
- Context window sizes and capabilities
"""
def get_name(self) -> str:
return "listmodels"
def get_description(self) -> str:
return (
"LIST AVAILABLE MODELS - Display all AI models organized by provider. "
"Shows which providers are configured, available models, their aliases, "
"context windows, and capabilities. Useful for understanding what models "
"can be used and their characteristics."
)
def get_input_schema(self) -> dict[str, Any]:
"""Return the JSON schema for the tool's input"""
return {"type": "object", "properties": {}, "required": []}
def get_system_prompt(self) -> str:
"""No AI model needed for this tool"""
return ""
def get_request_model(self):
"""Return the Pydantic model for request validation."""
return ToolRequest
async def prepare_prompt(self, request: ToolRequest) -> str:
"""Not used for this utility tool"""
return ""
def format_response(self, response: str, request: ToolRequest, model_info: Optional[dict] = None) -> str:
"""Not used for this utility tool"""
return response
async def execute(self, arguments: dict[str, Any]) -> list[TextContent]:
"""
List all available models organized by provider.
This overrides the base class execute to provide direct output without AI model calls.
Args:
arguments: Standard tool arguments (none required)
Returns:
Formatted list of models by provider
"""
from config import MODEL_CAPABILITIES_DESC
from providers.openrouter_registry import OpenRouterModelRegistry
output_lines = ["# Available AI Models\n"]
# Check native providers
native_providers = {
"gemini": {
"name": "Google Gemini",
"env_key": "GEMINI_API_KEY",
"models": {
"flash": "gemini-2.5-flash-preview-05-20",
"pro": "gemini-2.5-pro-preview-06-05",
},
},
"openai": {
"name": "OpenAI",
"env_key": "OPENAI_API_KEY",
"models": {
"o3": "o3",
"o3-mini": "o3-mini",
"o3-pro": "o3-pro",
"o4-mini": "o4-mini",
"o4-mini-high": "o4-mini-high",
},
},
"xai": {
"name": "X.AI (Grok)",
"env_key": "XAI_API_KEY",
"models": {
"grok": "grok-3",
"grok-3": "grok-3",
"grok-3-fast": "grok-3-fast",
"grok3": "grok-3",
"grokfast": "grok-3-fast",
},
},
}
# Check each native provider
for provider_key, provider_info in native_providers.items():
api_key = os.getenv(provider_info["env_key"])
is_configured = api_key and api_key != f"your_{provider_key}_api_key_here"
output_lines.append(f"## {provider_info['name']} {'' if is_configured else ''}")
if is_configured:
output_lines.append("**Status**: Configured and available")
output_lines.append("\n**Models**:")
for alias, full_name in provider_info["models"].items():
# Get description from MODEL_CAPABILITIES_DESC
desc = MODEL_CAPABILITIES_DESC.get(alias, "")
if isinstance(desc, str):
# Extract context window from description
import re
context_match = re.search(r"\(([^)]+context)\)", desc)
context_info = context_match.group(1) if context_match else ""
output_lines.append(f"- `{alias}` → `{full_name}` - {context_info}")
# Extract key capability
if "Ultra-fast" in desc:
output_lines.append(" - Fast processing, quick iterations")
elif "Deep reasoning" in desc:
output_lines.append(" - Extended reasoning with thinking mode")
elif "Strong reasoning" in desc:
output_lines.append(" - Logical problems, systematic analysis")
elif "EXTREMELY EXPENSIVE" in desc:
output_lines.append(" - ⚠️ Professional grade (very expensive)")
else:
output_lines.append(f"**Status**: Not configured (set {provider_info['env_key']})")
output_lines.append("")
# Check OpenRouter
openrouter_key = os.getenv("OPENROUTER_API_KEY")
is_openrouter_configured = openrouter_key and openrouter_key != "your_openrouter_api_key_here"
output_lines.append(f"## OpenRouter {'' if is_openrouter_configured else ''}")
if is_openrouter_configured:
output_lines.append("**Status**: Configured and available")
output_lines.append("**Description**: Access to multiple cloud AI providers via unified API")
try:
registry = OpenRouterModelRegistry()
aliases = registry.list_aliases()
# Group by provider for better organization
providers_models = {}
for alias in aliases[:20]: # Limit to first 20 to avoid overwhelming output
config = registry.resolve(alias)
if config and not (hasattr(config, "is_custom") and config.is_custom):
# Extract provider from model_name
provider = config.model_name.split("/")[0] if "/" in config.model_name else "other"
if provider not in providers_models:
providers_models[provider] = []
providers_models[provider].append((alias, config))
output_lines.append("\n**Available Models** (showing top 20):")
for provider, models in sorted(providers_models.items()):
output_lines.append(f"\n*{provider.title()}:*")
for alias, config in models[:5]: # Limit each provider to 5 models
context_str = f"{config.context_window // 1000}K" if config.context_window else "?"
output_lines.append(f"- `{alias}` → `{config.model_name}` ({context_str} context)")
total_models = len(aliases)
output_lines.append(f"\n...and {total_models - 20} more models available")
except Exception as e:
output_lines.append(f"**Error loading models**: {str(e)}")
else:
output_lines.append("**Status**: Not configured (set OPENROUTER_API_KEY)")
output_lines.append("**Note**: Provides access to GPT-4, Claude, Mistral, and many more")
output_lines.append("")
# Check Custom API
custom_url = os.getenv("CUSTOM_API_URL")
output_lines.append(f"## Custom/Local API {'' if custom_url else ''}")
if custom_url:
output_lines.append("**Status**: Configured and available")
output_lines.append(f"**Endpoint**: {custom_url}")
output_lines.append("**Description**: Local models via Ollama, vLLM, LM Studio, etc.")
try:
registry = OpenRouterModelRegistry()
custom_models = []
for alias in registry.list_aliases():
config = registry.resolve(alias)
if config and hasattr(config, "is_custom") and config.is_custom:
custom_models.append((alias, config))
if custom_models:
output_lines.append("\n**Custom Models**:")
for alias, config in custom_models:
context_str = f"{config.context_window // 1000}K" if config.context_window else "?"
output_lines.append(f"- `{alias}` → `{config.model_name}` ({context_str} context)")
if config.description:
output_lines.append(f" - {config.description}")
except Exception as e:
output_lines.append(f"**Error loading custom models**: {str(e)}")
else:
output_lines.append("**Status**: Not configured (set CUSTOM_API_URL)")
output_lines.append("**Example**: CUSTOM_API_URL=http://localhost:11434 (for Ollama)")
output_lines.append("")
# Add summary
output_lines.append("## Summary")
# Count configured providers
configured_count = sum(
[
1
for p in native_providers.values()
if os.getenv(p["env_key"])
and os.getenv(p["env_key"]) != f"your_{p['env_key'].lower().replace('_api_key', '')}_api_key_here"
]
)
if is_openrouter_configured:
configured_count += 1
if custom_url:
configured_count += 1
output_lines.append(f"**Configured Providers**: {configured_count}")
# Get total available models
try:
from tools.analyze import AnalyzeTool
tool = AnalyzeTool()
total_models = len(tool._get_available_models())
output_lines.append(f"**Total Available Models**: {total_models}")
except Exception:
pass
# Add usage tips
output_lines.append("\n**Usage Tips**:")
output_lines.append("- Use model aliases (e.g., 'flash', 'o3', 'opus') for convenience")
output_lines.append("- In auto mode, Claude will select the best model for each task")
output_lines.append("- Custom models are only available when CUSTOM_API_URL is set")
output_lines.append("- OpenRouter provides access to many cloud models with one API key")
# Format output
content = "\n".join(output_lines)
tool_output = ToolOutput(
status="success",
content=content,
content_type="text",
metadata={
"tool_name": self.name,
"configured_providers": configured_count,
},
)
return [TextContent(type="text", text=tool_output.model_dump_json())]
def get_model_category(self) -> ToolModelCategory:
"""Return the model category for this tool."""
return ToolModelCategory.FAST_RESPONSE # Simple listing, no AI needed