eigent/backend/app/controller/model_controller.py

291 lines
11 KiB
Python

# ========= Copyright 2025-2026 @ Eigent.ai All Rights Reserved. =========
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ========= Copyright 2025-2026 @ Eigent.ai All Rights Reserved. =========
import logging
from fastapi import APIRouter, HTTPException
from pydantic import BaseModel, Field
from app.component.error_format import normalize_error_to_openai_format
from app.component.model_validation import (
ValidationErrorType,
ValidationStage,
validate_model_with_details,
)
from app.model.model_platform import NormalizedModelPlatform
logger = logging.getLogger("model_controller")
router = APIRouter()
class ValidateModelRequest(BaseModel):
model_platform: NormalizedModelPlatform = Field(
"OPENAI", description="Model platform"
)
model_type: str = Field("GPT_4O_MINI", description="Model type")
api_key: str | None = Field(None, description="API key")
url: str | None = Field(None, description="Model URL")
model_config_dict: dict | None = Field(
None, description="Model config dict"
)
extra_params: dict | None = Field(
None, description="Extra model parameters"
)
include_diagnostics: bool = Field(
False, description="Include detailed diagnostic information"
)
class ValidateModelResponse(BaseModel):
is_valid: bool = Field(..., description="Is valid")
is_tool_calls: bool = Field(..., description="Is tool call used")
error_code: str | None = Field(None, description="Error code")
error: dict | None = Field(None, description="OpenAI-style error object")
message: str = Field(..., description="Message")
error_type: str | None = Field(None, description="Detailed error type")
failed_stage: str | None = Field(
None, description="Stage where validation failed"
)
successful_stages: list[str] | None = Field(
None, description="Stages that succeeded"
)
diagnostic_info: dict | None = Field(
None, description="Diagnostic information"
)
model_response_info: dict | None = Field(
None, description="Model response information"
)
tool_call_info: dict | None = Field(
None, description="Tool call information"
)
validation_stages: dict[str, bool] | None = Field(
None, description="Validation stages status"
)
@router.post("/model/validate")
async def validate_model(request: ValidateModelRequest):
"""Validate model configuration and tool call support with detailed error messages.
This endpoint validates a model configuration and provides detailed error messages
to help users understand the root cause of validation failures. It checks:
1. Initialization (model type and platform)
2. Model creation (authentication, network, model availability)
3. Agent creation
4. Model call execution
5. Tool call execution
Returns detailed diagnostic information if include_diagnostics is True.
"""
platform = request.model_platform
model_type = request.model_type
has_custom_url = request.url is not None
has_config = request.model_config_dict is not None
logger.info(
"Model validation started",
extra={
"platform": platform,
"model_type": model_type,
"has_url": has_custom_url,
"has_config": has_config,
"include_diagnostics": request.include_diagnostics,
},
)
# API key validation
if request.api_key is not None and str(request.api_key).strip() == "":
logger.warning(
"Model validation failed: empty API key",
extra={"platform": platform, "model_type": model_type},
)
raise HTTPException(
status_code=400,
detail={
"message": "Invalid key. Validation failed. Please provide a valid API key.",
"error_code": "invalid_api_key",
"error_type": ValidationErrorType.AUTHENTICATION_ERROR.value,
"failed_stage": ValidationStage.INITIALIZATION.value,
"error": {
"type": "invalid_request_error",
"param": "api_key",
"code": "invalid_api_key",
"message": "API key cannot be empty. Please provide a valid API key.",
},
},
)
try:
extra = request.extra_params or {}
logger.debug(
"Starting detailed model validation",
extra={"platform": platform, "model_type": model_type},
)
validation_result = validate_model_with_details(
platform,
model_type,
api_key=request.api_key,
url=request.url,
model_config_dict=request.model_config_dict,
**extra,
)
# Build response message based on validation result
# Prefer raw error messages from providers as they are usually clear and informative
if validation_result.is_tool_calls:
message = "Validation successful. Model supports tool calling and tool execution completed successfully."
elif validation_result.is_valid:
if (
validation_result.error_type
== ValidationErrorType.TOOL_CALL_NOT_SUPPORTED
):
message = "Model call succeeded, but this model does not support tool calling functionality. Please try with another model that supports tool calls."
elif (
validation_result.error_type
== ValidationErrorType.TOOL_CALL_EXECUTION_FAILED
):
# Use raw error message if available, otherwise use the formatted one
message = (
validation_result.raw_error_message
or validation_result.error_message
or "Tool call execution failed."
)
else:
message = (
validation_result.raw_error_message
or validation_result.error_message
or "Model call succeeded, but tool call validation failed. Please check the model configuration."
)
else:
# Use raw error message as primary message - provider errors are usually clear
# Only add context for specific cases where it's helpful
if validation_result.raw_error_message:
message = validation_result.raw_error_message
elif validation_result.error_message:
message = validation_result.error_message
else:
message = "Model validation failed. Please check your configuration and try again."
# Convert error type to error code for backward compatibility
error_code = None
error_obj = None
if validation_result.error_type:
error_code = validation_result.error_type.value
# Create OpenAI-style error object
error_obj = {
"type": "invalid_request_error",
"param": None,
"code": validation_result.error_type.value,
"message": validation_result.error_message or message,
}
# Add specific error details if available
if validation_result.error_details:
error_obj["details"] = validation_result.error_details
# Build response
response_data = {
"is_valid": validation_result.is_valid,
"is_tool_calls": validation_result.is_tool_calls,
"error_code": error_code,
"error": error_obj,
"message": message,
}
# Include detailed diagnostic information if requested
if request.include_diagnostics:
response_data["error_type"] = (
validation_result.error_type.value
if validation_result.error_type
else None
)
response_data["failed_stage"] = (
validation_result.failed_stage.value
if validation_result.failed_stage
else None
)
response_data["successful_stages"] = [
stage.value for stage in validation_result.successful_stages
]
response_data["diagnostic_info"] = (
validation_result.diagnostic_info
)
response_data["model_response_info"] = (
validation_result.model_response_info
)
response_data["tool_call_info"] = validation_result.tool_call_info
response_data["validation_stages"] = {
stage.value: success
for stage, success in validation_result.validation_stages.items()
}
result = ValidateModelResponse(**response_data)
# Use error or warning log level if there's an issue
log_extra = {
"platform": platform,
"model_type": model_type,
"is_valid": validation_result.is_valid,
"is_tool_calls": validation_result.is_tool_calls,
"error_type": validation_result.error_type.value
if validation_result.error_type
else None,
"failed_stage": validation_result.failed_stage.value
if validation_result.failed_stage
else None,
}
if not validation_result.is_valid:
logger.error("Model validation completed", extra=log_extra)
elif validation_result.error_type:
logger.warning("Model validation completed", extra=log_extra)
else:
logger.info("Model validation completed", extra=log_extra)
return result
except HTTPException:
# Re-raise HTTP exceptions as-is
raise
except Exception as e:
# Fallback error handling for unexpected errors
logger.error(
"Unexpected error during model validation",
extra={
"platform": platform,
"model_type": model_type,
"error": str(e),
},
exc_info=True,
)
message, error_code, error_obj = normalize_error_to_openai_format(e)
raise HTTPException(
status_code=500,
detail={
"message": f"Unexpected error during validation: {message}",
"error_code": error_code or "internal_error",
"error": error_obj
or {
"type": "internal_error",
"message": str(e),
},
},
)