GuangYu/QuestionProject
4
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Show retrieved knowledge chunks in audit logs

Browse Source
This commit is contained in:
Nelson
2026-07-09 10:29:58 +08:00
parent d6b7ec1e52
commit f39e51a4d4
9 changed files with 172 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
ai_knowledge_base_v2/apps/backend/alembic/versions/0003_ai_log_retrieved_chunks.py Normal file
View File

@@ -0,0 +1,17 @@
from __future__ import annotations
from alembic import op
import sqlalchemy as sa
revision = "0003_ai_log_retrieved_chunks"
down_revision = "0002_expand_model_config"
branch_labels = None
depends_on = None
def upgrade() -> None:
op.add_column("sys_ai_request_log", sa.Column("retrieved_chunks", sa.Text(), nullable=True))
def downgrade() -> None:
op.drop_column("sys_ai_request_log", "retrieved_chunks")

29
ai_knowledge_base_v2/apps/backend/app/api/admin_records.py
View File

@@ -1,8 +1,10 @@
from __future__ import annotations
import csv
import json
from datetime import datetime
from io import StringIO
from typing import Any
from fastapi import APIRouter, Depends, Query
from fastapi.responses import Response
@@ -236,6 +238,7 @@ def _ai_log_dict(log: AiRequestLog, *, include_prompt: bool = False) -> dict:
"modelName": log.model_name,
"knowledgeIds": log.knowledge_ids,
"retrieveCount": log.retrieve_count,
"retrievedChunks": _parse_retrieved_chunks(log.retrieved_chunks),
"inputToken": log.input_token,
"outputToken": log.output_token,
"totalToken": log.total_token,
@@ -247,3 +250,29 @@ def _ai_log_dict(log: AiRequestLog, *, include_prompt: bool = False) -> dict:
if include_prompt:
data["prompt"] = log.prompt
return data
def _parse_retrieved_chunks(raw: str | None) -> list[dict[str, Any]]:
if not raw:
return []
try:
chunks = json.loads(raw)
except json.JSONDecodeError:
return []
if not isinstance(chunks, list):
return []
normalized: list[dict[str, Any]] = []
for index, item in enumerate(chunks, start=1):
if not isinstance(item, dict):
continue
normalized.append(
{
"index": item.get("index") or index,
"knowledgeId": item.get("knowledgeId"),
"knowledgeName": item.get("knowledgeName") or "",
"title": item.get("title") or "",
"content": item.get("content") or "",
"sourceUrl": item.get("sourceUrl") or None,
}
)
return normalized

1
ai_knowledge_base_v2/apps/backend/app/models/logs.py
View File

@@ -19,6 +19,7 @@ class AiRequestLog(Base):
prompt: Mapped[str | None] = mapped_column(Text, nullable=True)
knowledge_ids: Mapped[str | None] = mapped_column(String(500), nullable=True)
retrieve_count: Mapped[int] = mapped_column(Integer, default=0, nullable=False)
retrieved_chunks: Mapped[str | None] = mapped_column(Text, nullable=True)
input_token: Mapped[int | None] = mapped_column(Integer, nullable=True)
output_token: Mapped[int | None] = mapped_column(Integer, nullable=True)
total_token: Mapped[int | None] = mapped_column(Integer, nullable=True)

25
ai_knowledge_base_v2/apps/backend/app/services/ai_request_log_service.py
View File

@@ -1,8 +1,12 @@
from __future__ import annotations
import json
from collections.abc import Sequence
from sqlalchemy.orm import Session
from app.models.logs import AiRequestLog
from app.services.rag_service import RetrievedChunk
class AiRequestLogService:
@@ -20,6 +24,7 @@ class AiRequestLogService:
input_token: int,
output_token: int,
cost_ms: int,
retrieved_chunks: Sequence[RetrievedChunk] | None = None,
) -> None:
db.add(
AiRequestLog(
@@ -30,6 +35,7 @@ class AiRequestLogService:
prompt=prompt,
knowledge_ids=knowledge_ids,
retrieve_count=retrieve_count,
retrieved_chunks=_dump_retrieved_chunks(retrieved_chunks),
input_token=input_token,
output_token=output_token,
total_token=input_token + output_token,
@@ -51,6 +57,7 @@ class AiRequestLogService:
retrieve_count: int,
cost_ms: int,
error_message: str,
retrieved_chunks: Sequence[RetrievedChunk] | None = None,
) -> None:
db.add(
AiRequestLog(
@@ -61,8 +68,26 @@ class AiRequestLogService:
prompt=prompt,
knowledge_ids=knowledge_ids,
retrieve_count=retrieve_count,
retrieved_chunks=_dump_retrieved_chunks(retrieved_chunks),
cost_ms=cost_ms,
status="FAILED",
error_message=error_message,
)
)
def _dump_retrieved_chunks(chunks: Sequence[RetrievedChunk] | None) -> str | None:
if not chunks:
return None
payload = [
{
"index": index,
"knowledgeId": chunk.knowledge_id,
"knowledgeName": chunk.knowledge_name,
"title": chunk.title,
"content": chunk.content,
"sourceUrl": chunk.source_url,
}
for index, chunk in enumerate(chunks, start=1)
]
return json.dumps(payload, ensure_ascii=False)

9
ai_knowledge_base_v2/apps/backend/app/services/chat_service.py
View File

@@ -87,6 +87,7 @@ class ChatService:
db.flush()
started_at = perf_counter()
rag_result = None
try:
# 获取历史消息（不含刚插入的 user_message，它还没 flush id）
history = list(
@@ -121,11 +122,12 @@ class ChatService:
message_id=user_message.id,
user_id=user.id,
model_name=None,
prompt=normalized_question,
knowledge_ids=None,
retrieve_count=0,
prompt=rag_result.prompt if rag_result is not None else normalized_question,
knowledge_ids=rag_result.knowledge_ids if rag_result is not None else None,
retrieve_count=len(rag_result.chunks) if rag_result is not None else 0,
cost_ms=cost_ms,
error_message=str(exc),
retrieved_chunks=rag_result.chunks if rag_result is not None else None,
)
db.commit()
raise HTTPException(
@@ -168,6 +170,7 @@ class ChatService:
input_token=completion.input_token,
output_token=completion.output_token,
cost_ms=cost_ms,
retrieved_chunks=rag_result.chunks,
)
db.commit()
return completion.answer

6
ai_knowledge_base_v2/apps/backend/app/services/chat_stream_service.py
View File

@@ -80,6 +80,7 @@ class ChatStreamService:
retrieve_count=len(rag_result.chunks) if rag_result is not None else 0,
cost_ms=cost_ms,
error_message=str(exc),
retrieved_chunks=rag_result.chunks if rag_result is not None else None,
)
db.commit()
raise HTTPException(status_code=status.HTTP_502_BAD_GATEWAY, detail=str(exc)) from exc
@@ -98,6 +99,7 @@ class ChatStreamService:
retrieve_count=len(rag_result.chunks) if rag_result is not None else 0,
cost_ms=cost_ms,
error_message="模型未返回有效内容",
retrieved_chunks=rag_result.chunks if rag_result is not None else None,
)
db.commit()
raise HTTPException(status_code=status.HTTP_502_BAD_GATEWAY, detail="模型未返回有效内容")
@@ -137,6 +139,7 @@ class ChatStreamService:
input_token=model_response.input_token if model_response is not None else 0,
output_token=_rough_token_count(answer),
cost_ms=cost_ms,
retrieved_chunks=rag_result.chunks if rag_result is not None else None,
)
db.commit()
@@ -205,6 +208,7 @@ class ChatStreamService:
retrieve_count=len(rag_result.chunks) if rag_result is not None else 0,
cost_ms=cost_ms,
error_message=str(exc),
retrieved_chunks=rag_result.chunks if rag_result is not None else None,
)
db.commit()
raise HTTPException(status_code=status.HTTP_502_BAD_GATEWAY, detail=str(exc)) from exc
@@ -223,6 +227,7 @@ class ChatStreamService:
retrieve_count=len(rag_result.chunks) if rag_result is not None else 0,
cost_ms=cost_ms,
error_message="模型未返回有效内容",
retrieved_chunks=rag_result.chunks if rag_result is not None else None,
)
db.commit()
raise HTTPException(status_code=status.HTTP_502_BAD_GATEWAY, detail="模型未返回有效内容")
@@ -295,6 +300,7 @@ def _write_success(
input_token=model_response.input_token if model_response is not None else 0,
output_token=_rough_token_count(answer),
cost_ms=cost_ms,
retrieved_chunks=rag_result.chunks if rag_result is not None else None,
)
db.commit()