refactor(api): TOML 配置 SSOT、统一错误契约、Auth/事务加固与可观测性 (#33)

配置 SSOT（TOML + .env）
统一错误契约
Auth 与事务边界
Redis / Celery 可靠性:业务 Redis（DB/0）与 Celery broker/backend（DB/1）显式拆分；连接池、sync client
可观测性（OpenTelemetry + LGTM）

api/app/agents/memoir/batch_phase1_prep.py

@@ -15,6 +15,7 @@ from app.core.config import settings
 from app.core.llm_call import LLMCallError, llm_json_call
 from app.core.logging import get_logger
 from app.features.conversation.models import Segment
+from app.features.memoir.constants import memoir
 
 logger = get_logger(__name__)
 
@@ -68,7 +69,7 @@ def run_batch_phase1_prep(
             llm,
             prompt,
             BatchPhase1LLMOutput,
-            max_tokens=int(settings.memoir_phase1_batch_llm_max_tokens),
+            max_tokens=int(memoir.phase1_batch_llm_max_tokens),
             agent="BatchPhase1Prep.run",
         )
     except LLMCallError as e:

api/app/agents/memoir/classification_agent.py

@@ -26,6 +26,7 @@ from app.core.config import settings
 from app.core.json_utils import extract_json_payload
 from app.core.llm_call import LLMCallError, llm_json_call
 from app.core.logging import get_logger
+from app.features.memoir.constants import memoir
 
 logger = get_logger(__name__)
 
@@ -144,7 +145,7 @@ class ClassificationAgent:
                     llm,
                     prompt,
                     ClassificationOutput,
-                    max_tokens=settings.memoir_classification_max_tokens,
+                    max_tokens=memoir.classification_max_tokens,
                     agent="ClassificationAgent.classify",
                 )
                 category = _normalize_llm_category(out.category)

api/app/agents/memoir/extraction_agent.py

@@ -14,6 +14,7 @@ from app.agents.stage_constants import normalize_chat_stage
 from app.core.config import settings
 from app.core.llm_call import LLMCallError, llm_json_call
 from app.core.logging import get_logger
+from app.features.memoir.constants import memoir
 
 logger = get_logger(__name__)
 
@@ -64,7 +65,7 @@ class ExtractionAgent:
                 llm,
                 prompt,
                 StateExtractionOutput,
-                max_tokens=settings.memoir_extraction_max_tokens,
+                max_tokens=memoir.extraction_max_tokens,
                 agent="ExtractionAgent.extract",
             )
             raw_slots = parsed.slots or {}

api/app/agents/memoir/fidelity_check_agent.py

@@ -13,6 +13,7 @@ from app.agents.memoir.schemas import FidelityOutput
 from app.core.config import settings
 from app.core.llm_call import LLMCallError, llm_json_call
 from app.core.logging import get_logger
+from app.features.memoir.constants import memoir
 
 logger = get_logger(__name__)
 
@@ -46,7 +47,7 @@ class FidelityCheckAgent:
         existing_canonical_markdown: str | None = None,
         is_append: bool = False,
     ) -> bool:
-        if not llm or not settings.memoir_fidelity_check_enabled:
+        if not llm or not memoir.fidelity_check_enabled:
             return True
         oral = (oral_text or "").strip()
         gen = (narrative_json or "").strip()
@@ -108,7 +109,7 @@ class FidelityCheckAgent:
                 llm,
                 prompt,
                 FidelityOutput,
-                max_tokens=settings.memoir_fidelity_check_max_tokens,
+                max_tokens=memoir.fidelity_check_max_tokens,
                 agent="FidelityCheckAgent.passes",
             )
             ok = bool(out.pass_)
@@ -120,7 +121,7 @@ class FidelityCheckAgent:
             return ok
         except LLMCallError as e:
             logger.warning("FidelityCheckAgent 解析失败: {}", e)
-            if is_append or settings.memoir_fidelity_fail_open_on_parse_error:
+            if is_append or memoir.fidelity_fail_open_on_parse_error:
                 logger.info("event=fidelity_parse_fail_open is_append={}", is_append)
                 return True
             logger.warning("event=fidelity_parse_fail_closed")

api/app/agents/memoir/narrative_agent.py

@@ -18,6 +18,7 @@ from app.core.config import settings
 from app.core.langchain_llm import invoke_json_object
 from app.core.llm_call import llm_json_call
 from app.core.logging import get_logger
+from app.features.memoir.constants import memoir
 
 logger = get_logger(__name__)
 
@@ -63,7 +64,7 @@ class NarrativeAgent:
                 llm,
                 prompt,
                 MemoirTitleOutput,
-                max_tokens=settings.memoir_title_max_tokens,
+                max_tokens=memoir.title_max_tokens,
                 agent="NarrativeAgent.generate_title",
                 fallback_factory=_title_fallback,
             )
@@ -118,7 +119,7 @@ class NarrativeAgent:
                     occupation=occupation,
                     language=language,
                 )
-                max_tokens = int(settings.memoir_narrative_merge_max_tokens)
+                max_tokens = int(memoir.narrative_merge_max_tokens)
                 agent_name = "NarrativeAgent.generate_narrative_merge"
             else:
                 prompt = get_narrative_json_prompt(
@@ -132,7 +133,7 @@ class NarrativeAgent:
                     occupation=occupation,
                     language=language,
                 )
-                max_tokens = int(settings.memoir_narrative_max_tokens)
+                max_tokens = int(memoir.narrative_max_tokens)
                 agent_name = "NarrativeAgent.generate_narrative"
             return invoke_json_object(
                 llm,

api/app/agents/memoir/orchestrator.py

@@ -29,6 +29,7 @@ from app.core.agent_logging import agent_span, agent_summary_enabled, log_agent_
 from app.core.config import settings
 from app.core.logging import get_logger
 from app.features.conversation.models import Segment
+from app.features.memoir.constants import memoir
 
 logger = get_logger(__name__)
 
@@ -90,7 +91,7 @@ class MemoirOrchestrator:
         use_batch = (
             bool(segments)
             and classify_extract_llm is not None
-            and settings.memoir_phase1_batch_llm_enabled
+            and memoir.phase1_batch_llm_enabled
         )
         if use_batch:
             try:
@@ -204,7 +205,7 @@ class MemoirOrchestrator:
             segments,
             state,
             classify_extract_llm,
-            chunk_size=int(settings.memoir_phase1_batch_llm_chunk_size),
+            chunk_size=int(memoir.phase1_batch_llm_chunk_size),
             on_chunk=on_phase1_chunk,
             language=language,
         )

api/app/agents/memoir/story_route_agent.py

@@ -21,6 +21,8 @@ from app.core.config import settings
 from app.core.llm_call import LLMCallError, llm_json_call
 from app.core.logging import get_logger
 from app.features.story.models import Story
+from app.features.memoir.constants import memoir
+from app.features.story.constants import story
 
 logger = get_logger(__name__)
 
@@ -47,7 +49,7 @@ def default_append_target_story_id(
     ordered = sort_stories_for_route(
         candidate_stories,
         meta,
-        summary_min_chars=int(settings.story_route_summary_min_chars),
+        summary_min_chars=int(story.route_summary_min_chars),
     )
     if not ordered:
         return None
@@ -247,7 +249,7 @@ class StoryRouteAgent:
             llm,
             prompt,
             StoryRouteDecision,
-            max_tokens=settings.memoir_story_route_max_tokens,
+            max_tokens=memoir.story_route_max_tokens,
             agent="StoryRouteAgent.decide",
             fallback_factory=_decide_fallback,
         )
@@ -295,7 +297,7 @@ class StoryRouteAgent:
                 llm,
                 prompt,
                 StoryBatchPlan,
-                max_tokens=settings.memoir_story_batch_plan_max_tokens,
+                max_tokens=memoir.story_batch_plan_max_tokens,
                 agent="StoryRouteAgent.plan_batch",
             )
         except LLMCallError as e:

api/app/agents/memoir/story_route_payload.py

@@ -15,6 +15,7 @@ if TYPE_CHECKING:
     from app.core.config import Settings
 
 from app.features.story.models import Story
+from app.features.story.constants import story
 
 _PLAIN_SNIPPET_NOISE = re.compile(r"[`*_#]+")
 
@@ -213,11 +214,11 @@ def build_route_candidate_rows(
 ) -> list[dict[str, Any]]:
     """排序 + 完整候选行（尚未做总预算降级）。"""
     meta = story_meta or {}
-    summary_min = int(settings.story_route_summary_min_chars)
+    summary_min = int(story.route_summary_min_chars)
     ordered = sort_stories_for_route(stories, meta, summary_min_chars=summary_min)
-    body_max = int(settings.story_route_candidate_body_max_chars)
-    head_c = int(settings.story_route_long_body_head_chars)
-    tail_c = int(settings.story_route_long_body_tail_chars)
+    body_max = int(story.route_candidate_body_max_chars)
+    head_c = int(story.route_long_body_head_chars)
+    tail_c = int(story.route_long_body_tail_chars)
     rows: list[dict[str, Any]] = []
     for s in ordered:
         rows.append(
@@ -231,8 +232,8 @@ def build_route_candidate_rows(
             )
         )
     by_id = {str(s.id): s for s in ordered}
-    total_max = int(settings.story_route_candidate_total_max_chars)
-    index_prev = int(settings.story_route_index_preview_chars)
+    total_max = int(story.route_candidate_total_max_chars)
+    index_prev = int(story.route_index_preview_chars)
     return apply_total_budget_downgrade(
         rows,
         stories_by_id=by_id,