agent-platform/services/model-gateway-service/app/infrastructure/provider.py

import httpx

from core_domain import ChatCompletionRequestContract, ChatCompletionResponseContract
from core_shared import JSONValue

from app.bootstrap.settings import ModelGatewayServiceSettings


class ModelProviderClientError(Exception):
    pass


class ModelProviderClient:
    def __init__(self, *, settings: ModelGatewayServiceSettings) -> None:
        self.settings = settings

    def create_chat_completion(
        self,
        payload: ChatCompletionRequestContract,
    ) -> ChatCompletionResponseContract:
        if payload.model is None:
            raise ModelProviderClientError("model is required for chat completion")

        request_payload = {
            "model": payload.model,
            "messages": [item.model_dump(mode="json") for item in payload.messages],
        }
        if payload.temperature is not None:
            request_payload["temperature"] = payload.temperature
        if payload.max_tokens is not None:
            request_payload["max_tokens"] = payload.max_tokens

        request_headers: dict[str, str] = {"content-type": "application/json"}
        if self.settings.provider_api_key:
            request_headers["authorization"] = f"Bearer {self.settings.provider_api_key}"

        try:
            with httpx.Client(timeout=60.0) as client:
                response = client.post(
                    f"{self.settings.provider_base_url.rstrip('/')}/chat/completions",
                    json=request_payload,
                    headers=request_headers,
                )
                response.raise_for_status()
        except httpx.HTTPError as exc:
            raise ModelProviderClientError(f"model provider request failed: {exc}") from exc

        response_json = _coerce_json_dict(response.json())
        content = _extract_response_content(response_json)
        finish_reason = _extract_finish_reason(response_json)
        usage_json = _extract_usage_json(response_json)
        return ChatCompletionResponseContract(
            model=payload.model,
            content=content,
            finish_reason=finish_reason,
            usage_json=usage_json,
            raw_response_json=response_json,
        )


def _coerce_json_dict(payload: JSONValue) -> dict[str, JSONValue]:
    if isinstance(payload, dict):
        return {str(key): value for key, value in payload.items()}
    return {}


def _extract_response_content(payload: dict[str, JSONValue]) -> str:
    choices = payload.get("choices")
    if isinstance(choices, list) and choices:
        first_choice = choices[0]
        if isinstance(first_choice, dict):
            message = first_choice.get("message")
            if isinstance(message, dict):
                content = message.get("content")
                if isinstance(content, str):
                    return content
            text = first_choice.get("text")
            if isinstance(text, str):
                return text
    return ""


def _extract_finish_reason(payload: dict[str, JSONValue]) -> str | None:
    choices = payload.get("choices")
    if isinstance(choices, list) and choices:
        first_choice = choices[0]
        if isinstance(first_choice, dict):
            finish_reason = first_choice.get("finish_reason")
            if isinstance(finish_reason, str):
                return finish_reason
    return None


def _extract_usage_json(payload: dict[str, JSONValue]) -> dict[str, JSONValue]:
    usage = payload.get("usage")
    if isinstance(usage, dict):
        return {str(key): value for key, value in usage.items()}
    return {}