Implement unified story generation flow

backend/app/services/adapters/__init__.py

@@ -9,6 +9,7 @@ from app.services.adapters.asr import openai as _asr_openai_adapter  # noqa: F40
 from app.services.adapters.base import AdapterConfig, BaseAdapter
 
 # Image adapters
+from app.services.adapters.image import antigravity as _image_antigravity_adapter  # noqa: F401
 from app.services.adapters.image import cqtai as _image_cqtai_adapter  # noqa: F401
 from app.services.adapters.registry import AdapterRegistry
 

backend/app/services/adapters/asr/openai.py

@@ -2,10 +2,11 @@
 
 from __future__ import annotations
 
+import re
 from io import BytesIO
 
 from fastapi import HTTPException
-from openai import AsyncOpenAI
+from openai import APIConnectionError, APIStatusError, APITimeoutError, AsyncOpenAI
 
 from app.core.logging import get_logger
 from app.services.adapters.asr.models import TranscriptionOutput
@@ -15,6 +16,14 @@ from app.services.adapters.registry import AdapterRegistry
 logger = get_logger(__name__)
 
 
+def _mask_openai_error(message: str) -> str:
+    """Avoid leaking bearer tokens while keeping ASR smoke failures actionable."""
+
+    sanitized = message.replace("\n", " ").strip()
+    sanitized = re.sub(r"Bearer\s+[A-Za-z0-9._-]+", "Bearer ***", sanitized)
+    return re.sub(r"sk-[A-Za-z0-9_-]+", "sk-***", sanitized)
+
+
 @AdapterRegistry.register("asr", "openai_asr")
 class OpenAIASRAdapter(BaseAdapter[TranscriptionOutput]):
     """Transcribe uploaded voice turn audio with OpenAI audio transcription."""
@@ -37,7 +46,11 @@ class OpenAIASRAdapter(BaseAdapter[TranscriptionOutput]):
                 detail="OPENAI_API_KEY 未配置，无法使用 OpenAI 语音转写。",
             )
 
-        client = AsyncOpenAI(api_key=self.config.api_key)
+        client = AsyncOpenAI(
+            api_key=self.config.api_key,
+            base_url=self.config.api_base or None,
+            timeout=self.config.timeout_ms / 1000,
+        )
         audio_file = BytesIO(audio_bytes)
         audio_file.name = file_name or "voice-turn.webm"
 
@@ -51,11 +64,29 @@ class OpenAIASRAdapter(BaseAdapter[TranscriptionOutput]):
                 language=language,
                 prompt=prompt,
             )
+        except APIStatusError as exc:
+            detail = _mask_openai_error(getattr(exc, "message", str(exc)))
+            logger.warning(
+                "openai_asr_failed",
+                status_code=exc.status_code,
+                error=detail,
+            )
+            raise HTTPException(
+                status_code=503,
+                detail=f"OpenAI ASR 调用失败（HTTP {exc.status_code}）：{detail}",
+            ) from exc
+        except (APITimeoutError, APIConnectionError) as exc:
+            detail = _mask_openai_error(str(exc))
+            logger.warning("openai_asr_failed", error=detail)
+            raise HTTPException(
+                status_code=503,
+                detail=f"OpenAI ASR 网络连接失败：{detail}",
+            ) from exc
         except Exception as exc:
             logger.warning("openai_asr_failed", error=str(exc))
             raise HTTPException(
                 status_code=503,
-                detail="语音转写服务暂时不可用，请稍后重试。",
+                detail=f"OpenAI ASR 调用异常：{_mask_openai_error(str(exc))}",
             ) from exc
 
         transcript_text = (getattr(response, "text", "") or "").strip()

backend/app/services/adapters/tts/minimax.py

@@ -126,6 +126,11 @@ class MiniMaxTTSAdapter(BaseAdapter[bytes]):
         except Exception:
             return False
 
+    @property
+    def estimated_cost(self) -> float:
+        """预估每次短文本语音合成成本 (USD)。"""
+        return 0.01
+
     @retry(
         stop=stop_after_attempt(3),
         wait=wait_exponential(multiplier=1, min=1, max=10),