sync: 鍚屾灞€鍩熺綉鏈€鏂颁唬鐮?(source master @ b69568e4)

This commit is contained in:
gozeros committed 2026-10-21 14:51:31 +08:00
1 parent 4f7f07504a
commit 8bc8b0e082
1651 files changed
+278599 -11416

No files matched your search

@@ -17,7 +17,7 @@ import logging
import os
import re
import uuid
from pathlib import Path
from pathlib import Path, PurePosixPath
from urllib.parse import quote
from fastapi import APIRouter, Depends, HTTPException, Query, Request
@@ -27,6 +27,7 @@ from pydantic import BaseModel, Field
from app.gateway.authz import require_permission
from app.gateway.deps import get_config, get_current_user
from app.gateway.document_rewrite_diagnostics import build_document_rewrite_diagnostic
from app.gateway.document_rewrite_pipeline import (
DOCUMENT_REWRITE_SYSTEM_PROMPT as _DOCUMENT_REWRITE_SYSTEM_PROMPT,
)
@@ -66,6 +67,7 @@ from app.gateway.word_export import (
build_markdown_docx,
)
from deerflow.config.app_config import AppConfig
from deerflow.config.system_settings import load_system_settings
from deerflow.models import create_chat_model
from deerflow.utils.stream_text import InlineThinkTagFilter
@@ -150,9 +152,8 @@ class DocumentRewriteRequest(BaseModel):
"""Request for a complete Markdown artifact rewrite.
This endpoint deliberately streams a *temporary* draft. The artifact is
written only after generation and Markdown validation succeed and the
original hash still matches. It is therefore safe for the sandbox to
visually clear and retype the document without risking a half-written file.
written to a separate sibling file only after generation and Markdown
validation succeed. The source artifact is never overwritten.
"""
thread_id: str = Field(..., alias="threadId")
@@ -189,6 +190,29 @@ def _document_rewrite_key(path: Path) -> str:
return os.path.normcase(str(path.resolve()))
def _next_document_rewrite_output_path(source_path: Path, virtual_path: str) -> tuple[Path, str]:
"""Reserve a readable, unused sibling name for one generated rewrite.
The actual and virtual paths must share the same filename. We make the
candidate deterministic and only advance the numeric suffix when a prior
rewrite already exists, so a conversation's new file card remains easy to
understand without ever replacing an earlier result.
"""
virtual = PurePosixPath(virtual_path)
suffix = source_path.suffix
stem = source_path.stem
for index in range(1, 10_000):
output_name = (
f"{stem}-改写版{suffix}"
if index == 1
else f"{stem}-改写版-{index}{suffix}"
)
candidate = source_path.with_name(output_name)
if not candidate.exists():
return candidate, str(virtual.with_name(output_name))
raise HTTPException(status_code=409, detail="改写版文件过多,请整理后再试。")
def _instruction_with_report_outline(instruction: str, report_outline: str | None) -> str:
"""Make an optional report template an explicit constraint for the rewrite model."""
normalized_instruction = instruction.strip()
@@ -235,13 +259,20 @@ def _artifact_job_matches(job: dict, *, thread_id: str, virtual_path: str) -> bo
def _artifact_job_response(job: dict) -> dict:
"""Limit durable job data to the rewrite fields the browser can hydrate."""
snapshot = job.get("input_snapshot") or {}
result = job.get("result_snapshot") or {}
output_path = snapshot.get("output_path") or result.get("outputPath")
diagnostic = result.get("diagnostic") if isinstance(result, dict) else None
return {
"id": job.get("id"),
"status": job.get("status"),
"phase": job.get("phase"),
"progress": job.get("progress"),
"createdAt": job.get("created_at"),
"result": job.get("result_snapshot") or {},
"result": result,
"errorCode": job.get("error_code"),
"errorMessage": job.get("error_message"),
"diagnostic": diagnostic if isinstance(diagnostic, dict) else None,
"outputPath": output_path if isinstance(output_path, str) else None,
"request": {
"instruction": str(snapshot.get("instruction") or ""),
"style": str(snapshot.get("style") or "professional"),
@@ -250,6 +281,7 @@ def _artifact_job_response(job: dict) -> dict:
"generateImages": bool(snapshot.get("generate_images")),
"maxGeneratedImages": int(snapshot.get("max_generated_images") or 2),
"originalContent": snapshot.get("original_content") or "",
"outputPath": output_path if isinstance(output_path, str) else None,
},
}
@@ -297,11 +329,29 @@ async def _stream_durable_artifact_rewrite_job(
# Never let the browser retain a spinning timeline
# merely because the terminal job row already knows
# the failure.
result = current.get("result_snapshot") or {}
diagnostic = result.get("diagnostic") if isinstance(result, dict) else None
if not isinstance(diagnostic, dict):
diagnostic = build_document_rewrite_diagnostic(
job_id=job_id,
stage=str(current.get("phase") or "generation"),
error_code=str(current.get("error_code") or "DOCUMENT_REWRITE_FAILED"),
message=str(
current.get("error_message")
or "全文改写任务未完成,原文件未被修改。"
),
)
yield _sse(
"error",
{
"message": str(current.get("error_message") or "全文改写任务未完成,原文件未被修改。"),
"stage": str(current.get("phase") or "generation"),
"message": str(diagnostic.get("message") or "全文改写任务未完成,原文件未被修改。"),
"stage": str(diagnostic.get("stage") or current.get("phase") or "generation"),
"errorCode": str(
diagnostic.get("errorCode")
or current.get("error_code")
or "DOCUMENT_REWRITE_FAILED"
),
"diagnostic": diagnostic,
},
)
return
@@ -392,11 +442,13 @@ async def export_markdown_docx(body: MarkdownDocxExportRequest) -> Response:
persist the supplied Markdown.
"""
try:
heading_numbering_style = load_system_settings().prompt_prefix.word_heading_numbering_style
content = await asyncio.to_thread(
build_markdown_docx,
body.markdown,
body.title,
embedded_images=body.embedded_images or None,
heading_numbering_style=heading_numbering_style,
)
except WordExportFontError as exc:
raise HTTPException(status_code=503, detail=str(exc)) from exc
@@ -418,9 +470,7 @@ async def export_markdown_docx(body: MarkdownDocxExportRequest) -> Response:
)
@router.post("/document-rewrite/jobs")
@require_permission("threads", "write", owner_check=True)
async def start_durable_document_rewrite_job(
async def _start_durable_document_rewrite_job_impl(
body: DocumentRewriteRequest,
request: Request,
) -> dict:
@@ -449,12 +499,26 @@ async def start_durable_document_rewrite_job(
)
base_hash = _content_hash(original)
if body.expected_hash and body.expected_hash != base_hash:
raise HTTPException(status_code=409, detail="文件已变化,请刷新后重新发起全文改写。")
# The editor can hold a slightly stale artifact query while the
# sandbox is finishing a write. Do not reject the durable job based
# on that browser-side preflight hash: the server has just captured
# the authoritative source snapshot above. The generated result is
# written to a separate output file, so the source artifact is never
# overwritten.
logger.info(
"writing/document-rewrite client snapshot stale; using current artifact: "
"thread=%s path=%s expected_hash=%s actual_hash=%s",
body.thread_id,
requested_path,
body.expected_hash,
base_hash,
)
job_store = getattr(request.app.state, "deep_research_job_store", None)
dispatcher = getattr(request.app.state, "deep_research_dispatcher", None)
if job_store is None or dispatcher is None:
raise HTTPException(status_code=503, detail="改写任务服务暂不可用。")
_output_actual_path, output_virtual_path = _next_document_rewrite_output_path(actual_path, requested_path)
scope = _artifact_rewrite_scope(body.thread_id, requested_path)
report_outline = body.report_outline.strip() if body.report_outline else ""
instruction = _instruction_with_report_outline(body.instruction, report_outline)
@@ -464,6 +528,7 @@ async def start_durable_document_rewrite_job(
"user_id": user_id,
"thread_id": body.thread_id,
"path": requested_path,
"output_path": output_virtual_path,
"original_content": original,
"base_hash": base_hash,
"instruction": instruction,
@@ -498,7 +563,28 @@ async def start_durable_document_rewrite_job(
phase="initializing",
)
if not _artifact_job_matches(job, thread_id=body.thread_id, virtual_path=requested_path):
raise HTTPException(status_code=409, detail="该文件已有不兼容的改写任务,请先等待或停止它。")
logger.warning(
"writing/document-rewrite active-job scope collision: thread=%s path=%s "
"existing_job=%s existing_session=%s existing_kind=%s",
body.thread_id,
requested_path,
job.get("id"),
job.get("session_id"),
(job.get("input_snapshot") or {}).get("kind")
if isinstance(job.get("input_snapshot"), dict)
else None,
)
message = "该文件已有不兼容的改写任务,请先等待或停止它。"
diagnostic = build_document_rewrite_diagnostic(
job_id=str(job.get("id") or "") or None,
stage="start",
error_code="ACTIVE_JOB_SCOPE_COLLISION",
message=message,
)
raise HTTPException(
status_code=409,
detail={"message": message, "diagnostic": diagnostic},
)
if created:
dispatcher.nudge()
return {
@@ -506,9 +592,55 @@ async def start_durable_document_rewrite_job(
"status": job["status"],
"reused": not created,
"baseHash": base_hash,
"outputPath": str((job.get("input_snapshot") or {}).get("output_path") or output_virtual_path),
}
@router.post("/document-rewrite/jobs")
@require_permission("threads", "write", owner_check=True)
async def start_durable_document_rewrite_job(
body: DocumentRewriteRequest,
request: Request,
) -> dict:
"""Start a durable rewrite and return safe diagnostics for unexpected failures."""
try:
return await _start_durable_document_rewrite_job_impl(body, request)
except HTTPException as exc:
if isinstance(exc.detail, dict) or exc.status_code < 409:
raise
diagnostic = build_document_rewrite_diagnostic(
job_id=None,
stage="start",
error_code=f"HTTP_{exc.status_code}",
message=str(exc.detail or "全文改写任务启动失败。"),
)
raise HTTPException(
status_code=exc.status_code,
detail={"message": diagnostic["message"], "diagnostic": diagnostic},
headers=exc.headers,
) from exc
except Exception as exc: # noqa: BLE001
logger.exception(
"writing/document-rewrite could not start: thread=%s path=%s",
body.thread_id,
body.path,
)
diagnostic = build_document_rewrite_diagnostic(
job_id=None,
stage="start",
error_code=type(exc).__name__,
message=str(exc) or "全文改写任务启动失败。",
exc=exc,
)
raise HTTPException(
status_code=500,
detail={
"message": diagnostic["message"],
"diagnostic": diagnostic,
},
) from exc
@router.get("/document-rewrite/jobs/latest")
@require_permission("threads", "read")
async def get_latest_durable_document_rewrite_job(