Optimize LLM hook cache-friendly context injection#201
Merged
EterUltimate merged 1 commit intoJun 11, 2026
Conversation
Contributor
Reviewer's GuideRefactors LLM hook context injection to prefer cache-friendly temporary extra_user_content_parts while preserving configurable legacy fallbacks, updates configuration defaults/normalization and WebUI schema accordingly, and adds tests and docs to cover the new behavior. Sequence diagram for cache-friendly LLM hook context injection with legacy fallbacksequenceDiagram
participant H as LLMHookHandler
participant Req as req
participant TP as TextPart
H->>H: _inject(req, injections, hook_start)
H->>H: _append_extra_user_content(req, context_text)
alt [extra_user_content_parts supported]
H->>TP: TextPart(text=context_text)
H->>TP: mark_as_temp()
H->>Req: extra_user_content_parts.append(part)
else [not supported]
H->>H: _legacy_inject(req, injection_text, target)
alt [target == "prompt"]
H->>Req: set req.prompt
else [target == "system_prompt" or default]
H->>Req: set req.system_prompt
end
end
Flow diagram for LLM hook injection target normalization and usageflowchart TD
A[配置文件或WebUI输入
llm_hook_injection_target] --> B[PluginConfig._normalize_llm_hook_injection_target]
B --> C{在 LLM_HOOK_TARGET_ALIASES 中?}
C -- 是 --> D[标准化为
extra_user_content_parts
system_prompt
或 prompt]
C -- 否 --> E[记录 warning
并使用
CACHE_FRIENDLY_LLM_HOOK_TARGET]
D --> F[PluginConfig.llm_hook_injection_target]
E --> F
F --> G[LLMHookHandler._inject
读取 llm_hook_injection_target]
G --> H{支持 extra_user_content_parts?}
H -- 是 --> I[使用临时
extra_user_content_parts 注入]
H -- 否 --> J[按 legacy 目标
回退到 system_prompt
或 prompt]
File-Level Changes
Possibly linked issues
Tips and commandsInteracting with Sourcery
Customizing Your ExperienceAccess your dashboard to:
Getting Help
|
EterUltimate
force-pushed
the
codex/self-learning-cache-hit
branch
from
483e6f8 to
caaf3a6
Compare
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Hey - I've found 1 issue, and left some high level feedback:
- In
_append_extra_user_content, failures silently fall back to legacy injection; consider adding a debug log whenextra_user_content_parts/TextPartsupport is missing so it's easier to understand why the cache‑friendly path isn't being used at runtime. - The
<context>\n{injection_text}\n</context>wrapper string is duplicated between_injectand the tests; extracting this format into a small helper/constant would reduce the risk of future drift between behavior and assertions.
Prompt for AI Agents
Please address the comments from this code review:
## Overall Comments
- In `_append_extra_user_content`, failures silently fall back to legacy injection; consider adding a debug log when `extra_user_content_parts`/`TextPart` support is missing so it's easier to understand why the cache‑friendly path isn't being used at runtime.
- The `<context>\n{injection_text}\n</context>` wrapper string is duplicated between `_inject` and the tests; extracting this format into a small helper/constant would reduce the risk of future drift between behavior and assertions.
## Individual Comments
### Comment 1
<location path="config.py" line_range="344-353" />
<code_context>
def _normalize_target_blacklist(cls, value) -> List[str]:
return normalize_identifier_list(value)
+ @field_validator("llm_hook_injection_target", mode="before")
+ @classmethod
+ def _normalize_llm_hook_injection_target(cls, value) -> str:
+ target = str(value or CACHE_FRIENDLY_LLM_HOOK_TARGET).strip()
+ normalized = LLM_HOOK_TARGET_ALIASES.get(target)
+ if normalized:
+ return normalized
+ logger.warning(
+ f"未知 LLM Hook 注入目标 {value!r}，"
+ "已回退到 cache-friendly extra_user_content_parts"
+ )
+ return CACHE_FRIENDLY_LLM_HOOK_TARGET
+
def model_post_init(self, __context) -> None:
</code_context>
<issue_to_address>
**suggestion:** Normalize `llm_hook_injection_target` more robustly (e.g., case-insensitive, alias-friendly).
The validator currently does `str(value).strip()` then `LLM_HOOK_TARGET_ALIASES.get(target)`, so values are whitespace-tolerant but still case-sensitive (e.g., `"System_Prompt"` falls back to `extra_user_content_parts`). Normalizing case for `target` and alias keys (or otherwise canonicalizing the input) would make configuration more resilient to variations from different environments or UIs.
Suggested implementation:
```python
@field_validator("llm_hook_injection_target", mode="before")
@classmethod
def _normalize_llm_hook_injection_target(cls, value) -> str:
# 允许大小写和空白差异的健壮归一化
raw_target = str(value or CACHE_FRIENDLY_LLM_HOOK_TARGET).strip()
normalized_key = raw_target.lower()
# 优先使用小写键查找，其次回退到原始键，方便兼容旧配置
normalized = (
LLM_HOOK_TARGET_ALIASES.get(normalized_key)
or LLM_HOOK_TARGET_ALIASES.get(raw_target)
)
if normalized:
return normalized
logger.warning(
f"未知 LLM Hook 注入目标 {value!r}，"
"已回退到 cache-friendly extra_user_content_parts"
)
return CACHE_FRIENDLY_LLM_HOOK_TARGET
```
1. 为了让大小写无关的查找完全生效，建议在定义 `LLM_HOOK_TARGET_ALIASES` 的位置，将所有键统一为小写，例如：
`LLM_HOOK_TARGET_ALIASES = {k.lower(): v for k, v in _ORIGINAL_ALIASES.items()}` 或者在静态定义时就使用小写键。
2. 如果存在直接使用原来大小写敏感键访问 `LLM_HOOK_TARGET_ALIASES` 的代码，需要一并调整为使用小写键（例如：`LLM_HOOK_TARGET_ALIASES[some_key.lower()]`），以保持行为一致。
</issue_to_address>Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.
Comment on lines
+344
to
+353
| @field_validator("llm_hook_injection_target", mode="before") | ||
| @classmethod | ||
| def _normalize_llm_hook_injection_target(cls, value) -> str: | ||
| target = str(value or CACHE_FRIENDLY_LLM_HOOK_TARGET).strip() | ||
| normalized = LLM_HOOK_TARGET_ALIASES.get(target) | ||
| if normalized: | ||
| return normalized | ||
| logger.warning( | ||
| f"未知 LLM Hook 注入目标 {value!r}，" | ||
| "已回退到 cache-friendly extra_user_content_parts" |
Contributor
There was a problem hiding this comment.
suggestion: Normalize llm_hook_injection_target more robustly (e.g., case-insensitive, alias-friendly).
The validator currently does str(value).strip() then LLM_HOOK_TARGET_ALIASES.get(target), so values are whitespace-tolerant but still case-sensitive (e.g., "System_Prompt" falls back to extra_user_content_parts). Normalizing case for target and alias keys (or otherwise canonicalizing the input) would make configuration more resilient to variations from different environments or UIs.
Suggested implementation:
@field_validator("llm_hook_injection_target", mode="before")
@classmethod
def _normalize_llm_hook_injection_target(cls, value) -> str:
# 允许大小写和空白差异的健壮归一化
raw_target = str(value or CACHE_FRIENDLY_LLM_HOOK_TARGET).strip()
normalized_key = raw_target.lower()
# 优先使用小写键查找，其次回退到原始键，方便兼容旧配置
normalized = (
LLM_HOOK_TARGET_ALIASES.get(normalized_key)
or LLM_HOOK_TARGET_ALIASES.get(raw_target)
)
if normalized:
return normalized
logger.warning(
f"未知 LLM Hook 注入目标 {value!r}，"
"已回退到 cache-friendly extra_user_content_parts"
)
return CACHE_FRIENDLY_LLM_HOOK_TARGET- 为了让大小写无关的查找完全生效，建议在定义
LLM_HOOK_TARGET_ALIASES的位置，将所有键统一为小写，例如：
LLM_HOOK_TARGET_ALIASES = {k.lower(): v for k, v in _ORIGINAL_ALIASES.items()}或者在静态定义时就使用小写键。 - 如果存在直接使用原来大小写敏感键访问
LLM_HOOK_TARGET_ALIASES的代码，需要一并调整为使用小写键（例如：LLM_HOOK_TARGET_ALIASES[some_key.lower()]），以保持行为一致。
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.
Summary
extra_user_content_partsinstead of changing the stablesystem_promptprefixllm_hook_injection_targetto the cache-friendly AstrBot API while preserving legacy fallback targets for older AstrBot versionsTextPartinjection and fallback behaviorValidation
python -m pytest tests\unit\test_feature_delegation.py tests\unit\test_config.py tests\unit\test_config_service.py --basetemp .tmp\pytest-basetemp -p no:cacheproviderpython -m py_compile config.py services\hooks\llm_hook_handler.py webui\services\config_service.py tests\unit\test_feature_delegation.py tests\unit\test_config.py tests\unit\test_config_service.pygit diff --cached --checkTextPart.mark_as_temp()exists and hook injection keepsreq.system_promptunchanged while appending aTextPartSummary by Sourcery
Optimize LLM hook context injection to use cache-friendly temporary user content parts by default while maintaining backward-compatible fallbacks and updating configuration, tests, and documentation accordingly.
Enhancements:
Documentation:
Tests: