fix(engine): prompt plaza edits not taking effect in generation (#146)#170
Open
liaoyl830 wants to merge 1 commit into
Open
Conversation
…minglinyi#146) Root cause: PromptRegistry caches prompt templates with a 300s TTL. After a user edits a prompt in the plaza, the API layer cache is cleared but the PromptRegistry cache is not invalidated, so the generation pipeline still serves stale templates for up to 5 minutes. Changes: 1. interfaces/api/v1/workbench/llm_control.py - _invalidate_plaza_cache() now also calls PromptRegistry.invalidate_cache() so all editing endpoints (update_node, rollback_node, create_node, delete_node, import_prompts, create_template) trigger full cache invalidation 2. infrastructure/ai/prompt_manager.py - update_node() now calls _invalidate_registry_cache() after commit - New _invalidate_registry_cache() static helper method 3. application/workflows/auto_novel_generation_workflow.py - Upgrade fallback logging from DEBUG to WARNING for both _get_workflow_system_template and _get_workflow_user_template - Makes it easier to diagnose when the pipeline is falling back to hardcoded templates instead of reading from the DB
Contributor
📝 WalkthroughWalkthroughThis PR adds PromptRegistry cache invalidation to keep node updates immediately visible across infrastructure and API layers. Logging for template fallbacks escalates to WARNING level with explicit fallback messages. A cache invalidation helper is introduced and integrated into node update and API cache operations. ChangesCache invalidation and logging improvements
Estimated code review effort🎯 2 (Simple) | ⏱️ ~12 minutes Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
🧹 Nitpick comments (2)
infrastructure/ai/prompt_manager.py (1)
1091-1100: ⚡ Quick winMake registry invalidation explicit instead of heuristic key-shape detection.
Line 1097 relies on string shape to infer key type. UUID-style IDs with hyphens can be treated as node keys, which risks missing the real cache entry. Prefer explicit invalidation by resolved
node_key(or an explicitby_keyflag) to avoid stale cache edge cases.Suggested change
- def _invalidate_registry_cache(node_key_or_id: str) -> None: - """使 PromptRegistry 缓存失效（node_key 或 node_id 均可）。""" + def _invalidate_registry_cache(node_key: Optional[str] = None) -> None: + """使 PromptRegistry 缓存失效。传 node_key 时做定向失效，否则全量失效。""" try: from infrastructure.ai.prompt_registry import get_prompt_registry registry = get_prompt_registry() - # 尝试按 node_key 失效；如果是 node_id 则清除全部缓存（更安全） - if '-' in node_key_or_id and len(node_key_or_id) < 50: - registry.invalidate_cache(node_key_or_id) + if node_key: + registry.invalidate_cache(node_key) else: registry.invalidate_cache() except Exception: pass # Registry 不可用时静默跳过- self._invalidate_registry_cache(node_id) + updated_node = self.get_node(node_id, by_key=False) + self._invalidate_registry_cache(updated_node.node_key if updated_node else None) - return self.get_node(node_id, by_key=False) + return updated_node🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@infrastructure/ai/prompt_manager.py` around lines 1091 - 1100, The _invalidate_registry_cache function currently heuristically treats node_key_or_id as a key when it contains a '-' and is short, which can misclassify UUID-style IDs; change it to resolve the canonical node_key (or accept an explicit by_key boolean) and call get_prompt_registry() and registry.invalidate_cache with the resolved node_key only when you truly have a key, otherwise call registry.invalidate_cache() to clear all; update the function signature (e.g., _invalidate_registry_cache(node_key_or_id: str, by_key: Optional[bool]=None)) or perform a lookup to convert a node_id to node_key before calling registry.invalidate_cache to ensure deterministic invalidation.interfaces/api/v1/workbench/llm_control.py (1)
275-276: ⚡ Quick winKeep non-blocking behavior, but emit a debug log on invalidation failure.
Line 275-276 fully suppresses failures, which makes stale-cache incidents harder to diagnose in production.
Suggested change
- except Exception: - pass # Registry 不可用时静默跳过，不影响 API 层功能 + except Exception as exc: + logger.debug("PromptRegistry cache invalidation skipped: %s", exc, exc_info=True)🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@interfaces/api/v1/workbench/llm_control.py` around lines 275 - 276, The except Exception: pass in interfaces/api/v1/workbench/llm_control.py should remain non-blocking but emit a debug-level log when registry invalidation fails; replace the pass with a debug log call (e.g., logger.debug("Registry invalidation failed, continuing without aborting", exc_info=True)) using the existing logger variable (or process_logger) so the exception and stack trace are recorded for diagnosis while keeping the API behavior unchanged.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Nitpick comments:
In `@infrastructure/ai/prompt_manager.py`:
- Around line 1091-1100: The _invalidate_registry_cache function currently
heuristically treats node_key_or_id as a key when it contains a '-' and is
short, which can misclassify UUID-style IDs; change it to resolve the canonical
node_key (or accept an explicit by_key boolean) and call get_prompt_registry()
and registry.invalidate_cache with the resolved node_key only when you truly
have a key, otherwise call registry.invalidate_cache() to clear all; update the
function signature (e.g., _invalidate_registry_cache(node_key_or_id: str,
by_key: Optional[bool]=None)) or perform a lookup to convert a node_id to
node_key before calling registry.invalidate_cache to ensure deterministic
invalidation.
In `@interfaces/api/v1/workbench/llm_control.py`:
- Around line 275-276: The except Exception: pass in
interfaces/api/v1/workbench/llm_control.py should remain non-blocking but emit a
debug-level log when registry invalidation fails; replace the pass with a debug
log call (e.g., logger.debug("Registry invalidation failed, continuing without
aborting", exc_info=True)) using the existing logger variable (or
process_logger) so the exception and stack trace are recorded for diagnosis
while keeping the API behavior unchanged.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: 098e8fd3-c16d-484b-8cf8-f091b94c2f15
📒 Files selected for processing (3)
application/workflows/auto_novel_generation_workflow.pyinfrastructure/ai/prompt_manager.pyinterfaces/api/v1/workbench/llm_control.py
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
变更类型
fixBug 修复变更说明
修复提示词广场编辑后不生效的问题。用户在提示词广场修改提示词并保存后，章节生成仍使用旧版本提示词。
根因分析
PromptRegistry对提示词节点有 300 秒（5 分钟）的内存缓存。用户编辑提示词后，API 层的_plaza_cache会被清除，但PromptRegistry的缓存未同步失效，导致生成管线在 5 分钟内仍读取旧版 DB 快照。修复内容
_invalidate_plaza_cache()扩展 — 所有提示词写操作（编辑、回滚、新建、删除、导入）现在同时清除PromptRegistry缓存，确保生成管线立即读取到最新版本。PromptManager.update_node()补充失效 — 数据库层更新节点后主动调用_invalidate_registry_cache()，作为 API 层失效的兜底。降级日志级别提升 —
_get_workflow_system_template()和_get_workflow_user_template()的硬编码回退日志从DEBUG提升至WARNING，便于排查管道是否意外降级到硬编码模板。架构影响
interfaces/infrastructure/application测试
_invalidate_registry_cache()包含异常捕获，PromptRegistry不可用时静默跳过Summary by CodeRabbit
Bug Fixes
Chores