309a051038
数据库与模型：新增多版迁移（章节证据快照、对话血缘、记忆事实/时间线 lineage 等），把「成稿 ↔ 对话/记忆」的溯源信息落到表结构里。 业务链路：会话与 WS、回忆录/故事流水线、记忆写入与 enrichment 等跟着接上线索与快照；新增章节证据快照与评测侧 EvalTraceService 等模块，方便组评审用的证据包。 内部评测：自动化 run 与手工 memoir 评审共用可追溯证据；rubric/ judge 相关脚本与文档有配套调整。 app-eval-web：Memoir/实验详情里能展开看证据摘要与 evidence_trace（含对话轮次 id）；Vite 代理与 development.sh 注入的 API 端口与当前默认内部评测端口一致，避免改端口后页面连错服务。 工程杂项：GitHub Actions / 仓库说明有更新；各适配器与支付/配额/plan 等多处为小改动或跟随主改动的收尾；新增/扩充了?
287 lines
10 KiB
Python
287 lines
10 KiB
Python
"""
|
||
Liblib 图生 SDK 封装，位于 adapters 层；实现细节不暴露给 feature。
|
||
Feature 通过 port ImageGenerator 使用，本模块仅被 app.adapters.image_gen.liblib 使用。
|
||
"""
|
||
|
||
import base64
|
||
import hmac
|
||
import logging
|
||
import re
|
||
import time
|
||
import uuid
|
||
from hashlib import sha1
|
||
from urllib.parse import urlparse
|
||
|
||
import httpx
|
||
|
||
from app.core.config import settings
|
||
from app.core.logging import get_logger
|
||
|
||
logger = get_logger(__name__)
|
||
|
||
DEFAULT_LIBLIB_TEMPLATE_UUID = "5d7e67009b344550bc1aa6ccbfa1d7f4"
|
||
|
||
_SENSITIVE_QUERY_PARAMS = ("AccessKey", "Signature", "Timestamp", "SignatureNonce")
|
||
_SENSITIVE_QUERY_RE = re.compile(
|
||
r"([?&])(" + "|".join(_SENSITIVE_QUERY_PARAMS) + r")=([^&\s]+)"
|
||
)
|
||
|
||
_SIZE_TO_ASPECT_RATIO = {
|
||
"1024x1024": "square",
|
||
"768x1024": "portrait",
|
||
"1280x720": "landscape",
|
||
}
|
||
_DEFAULT_WIDTH, _DEFAULT_HEIGHT = 1024, 1024
|
||
|
||
|
||
def _parse_size(size: str) -> tuple[int, int]:
|
||
try:
|
||
w_str, h_str = size.lower().split("x", 1)
|
||
w = max(512, min(2048, int(w_str)))
|
||
h = max(512, min(2048, int(h_str)))
|
||
return w, h
|
||
except (ValueError, AttributeError):
|
||
return _DEFAULT_WIDTH, _DEFAULT_HEIGHT
|
||
|
||
|
||
class LiblibImageProvider:
|
||
"""Liblib (https://openapi.liblibai.cloud) image generation — adapter 层实现。"""
|
||
|
||
def __init__(
|
||
self,
|
||
http_client: httpx.Client | None = None,
|
||
access_key: str | None = None,
|
||
secret_key: str | None = None,
|
||
base_url: str | None = None,
|
||
template_uuid: str | None = None,
|
||
allowed_download_hosts: tuple[str, ...] | None = None,
|
||
):
|
||
_install_http_log_redaction()
|
||
self._owns_http_client = http_client is None
|
||
self.http_client = http_client or httpx.Client(timeout=120)
|
||
self.access_key = access_key or (settings.liblib_access_key or "")
|
||
self.secret_key = secret_key or (settings.liblib_secret_key or "")
|
||
self.base_url = (
|
||
base_url or settings.liblib_base_url or "https://openapi.liblibai.cloud"
|
||
).rstrip("/")
|
||
self.template_uuid = template_uuid or (
|
||
settings.liblib_template_uuid or DEFAULT_LIBLIB_TEMPLATE_UUID
|
||
)
|
||
self.allowed_download_hosts = _build_allowed_download_hosts(
|
||
self.base_url,
|
||
allowed_download_hosts=allowed_download_hosts,
|
||
)
|
||
|
||
def _build_url(self, uri: str) -> str:
|
||
return f"{self.base_url}{uri}"
|
||
|
||
def _sign(self, uri: str) -> dict[str, str]:
|
||
timestamp = str(int(time.time() * 1000))
|
||
nonce = str(uuid.uuid4())
|
||
content = "&".join((uri, timestamp, nonce))
|
||
digest = hmac.new(self.secret_key.encode(), content.encode(), sha1).digest()
|
||
signature = base64.urlsafe_b64encode(digest).rstrip(b"=").decode()
|
||
return {
|
||
"AccessKey": self.access_key,
|
||
"Signature": signature,
|
||
"Timestamp": timestamp,
|
||
"SignatureNonce": nonce,
|
||
}
|
||
|
||
def submit_generation(self, prompt: str, size: str, style: str) -> dict:
|
||
uri = "/api/generate/webui/text2img/ultra"
|
||
url = self._build_url(uri)
|
||
params = self._sign(uri)
|
||
styled_prompt = _apply_style_to_prompt(prompt, style)
|
||
aspect_ratio = _SIZE_TO_ASPECT_RATIO.get(size)
|
||
generate_params: dict = {
|
||
"prompt": styled_prompt,
|
||
"imgCount": 1,
|
||
"steps": 30,
|
||
}
|
||
if aspect_ratio:
|
||
generate_params["aspectRatio"] = aspect_ratio
|
||
else:
|
||
w, h = _parse_size(size)
|
||
generate_params["imageSize"] = {"width": w, "height": h}
|
||
body = {
|
||
"templateUuid": self.template_uuid,
|
||
"generateParams": generate_params,
|
||
}
|
||
response = self.http_client.post(
|
||
url,
|
||
params=params,
|
||
headers={"Content-Type": "application/json"},
|
||
json=body,
|
||
)
|
||
response.raise_for_status()
|
||
data = response.json()
|
||
if data.get("code") != 0:
|
||
raise RuntimeError(f"Liblib submit failed: {data.get('msg', data)}")
|
||
generate_uuid = data["data"]["generateUuid"]
|
||
return {"status": "processing", "job_id": generate_uuid, "image_url": None}
|
||
|
||
def poll_until_complete(
|
||
self, job: dict, poll_interval_seconds: int, max_attempts: int
|
||
) -> dict:
|
||
uri = "/api/generate/webui/status"
|
||
for attempt in range(max_attempts):
|
||
url = self._build_url(uri)
|
||
params = self._sign(uri)
|
||
response = self.http_client.post(
|
||
url,
|
||
params=params,
|
||
headers={"Content-Type": "application/json"},
|
||
json={"generateUuid": job["job_id"]},
|
||
)
|
||
response.raise_for_status()
|
||
data = response.json()
|
||
if data.get("code") != 0:
|
||
raise RuntimeError(
|
||
f"Liblib status query failed: {data.get('msg', data)}"
|
||
)
|
||
result = data.get("data", {})
|
||
status = result.get("generateStatus")
|
||
if status == 5:
|
||
images = result.get("images") or []
|
||
if images:
|
||
return {
|
||
"status": "completed",
|
||
"image_url": images[0]["imageUrl"],
|
||
"job_id": job["job_id"],
|
||
}
|
||
raise RuntimeError(
|
||
f"Liblib returned success but no images for {job['job_id']}"
|
||
)
|
||
if status == 6:
|
||
raise RuntimeError(
|
||
f"Liblib generation failed: {result.get('generateMsg', 'unknown')}"
|
||
)
|
||
if status == 7:
|
||
raise TimeoutError(
|
||
f"Liblib returned undocumented status 7 for {job['job_id']}"
|
||
)
|
||
logger.debug(
|
||
"Liblib poll attempt {}/{}, status={}, job={}",
|
||
attempt + 1,
|
||
max_attempts,
|
||
status,
|
||
job["job_id"],
|
||
)
|
||
time.sleep(poll_interval_seconds)
|
||
raise TimeoutError(
|
||
f"Liblib image generation timed out after {max_attempts} attempts for {job['job_id']}"
|
||
)
|
||
|
||
def download_image(self, job: dict) -> bytes:
|
||
image_url = job["image_url"]
|
||
_validate_download_url(image_url, self.allowed_download_hosts)
|
||
response = self.http_client.get(image_url)
|
||
response.raise_for_status()
|
||
return response.content
|
||
|
||
def download_image_from_url(self, image_url: str) -> bytes:
|
||
"""按 URL 下载图片（用于 port 的 download_image）。"""
|
||
return self.download_image({"image_url": image_url})
|
||
|
||
def close(self) -> None:
|
||
if self._owns_http_client:
|
||
self.http_client.close()
|
||
|
||
def __enter__(self) -> "LiblibImageProvider":
|
||
return self
|
||
|
||
def __exit__(self, exc_type, exc, tb) -> None:
|
||
self.close()
|
||
|
||
|
||
class _LiblibAuthRedactionFilter(logging.Filter):
|
||
def filter(self, record: logging.LogRecord) -> bool:
|
||
record.msg = _redact_sensitive_query_values(record.msg)
|
||
if record.args:
|
||
if isinstance(record.args, dict):
|
||
record.args = {
|
||
k: _redact_sensitive_query_values(v) for k, v in record.args.items()
|
||
}
|
||
else:
|
||
record.args = tuple(
|
||
_redact_sensitive_query_values(v) for v in record.args
|
||
)
|
||
return True
|
||
|
||
|
||
def _redact_sensitive_query_values(value):
|
||
if isinstance(value, str):
|
||
return _SENSITIVE_QUERY_RE.sub(r"\1\2=[REDACTED]", value)
|
||
return value
|
||
|
||
|
||
def _install_http_log_redaction() -> None:
|
||
"""对 httpx/httpcore 的标准库 Logger 挂 Filter（非 loguru get_logger）。"""
|
||
for logger_name in (
|
||
"httpx",
|
||
"httpcore",
|
||
"httpcore.connection",
|
||
"httpcore.http11",
|
||
"httpcore.proxy",
|
||
):
|
||
target_logger = logging.getLogger(logger_name)
|
||
if getattr(target_logger, "_liblib_auth_redaction_installed", False):
|
||
continue
|
||
target_logger.addFilter(_LiblibAuthRedactionFilter())
|
||
target_logger._liblib_auth_redaction_installed = True
|
||
|
||
|
||
def _build_allowed_download_hosts(
|
||
base_url: str,
|
||
allowed_download_hosts: tuple[str, ...] | None = None,
|
||
) -> tuple[str, ...]:
|
||
configured_hosts = allowed_download_hosts
|
||
if configured_hosts is None:
|
||
configured_hosts = tuple(
|
||
host.strip().lower()
|
||
for host in (settings.memoir_image_download_hosts or "").split(",")
|
||
if host.strip()
|
||
)
|
||
base_hostname = (urlparse(base_url).hostname or "").lower()
|
||
default_hosts: set[str] = set()
|
||
if base_hostname:
|
||
default_hosts.add(base_hostname)
|
||
if (
|
||
base_hostname.endswith(".liblibai.cloud")
|
||
or base_hostname == "liblibai.cloud"
|
||
):
|
||
default_hosts.add("liblibai.cloud")
|
||
default_hosts.add("liblib.cloud")
|
||
return tuple(sorted(default_hosts.union(configured_hosts)))
|
||
|
||
|
||
def _validate_download_url(image_url: str, allowed_hosts: tuple[str, ...]) -> None:
|
||
parsed = urlparse(image_url)
|
||
hostname = (parsed.hostname or "").lower()
|
||
if parsed.scheme != "https" or not hostname:
|
||
raise ValueError(f"Unsupported image download URL: {image_url}")
|
||
if not any(
|
||
_hostname_matches(hostname, allowed_host) for allowed_host in allowed_hosts
|
||
):
|
||
raise ValueError(f"Image download host is not allowed: {hostname}")
|
||
|
||
|
||
def _hostname_matches(hostname: str, allowed_host: str) -> bool:
|
||
normalized_allowed = allowed_host.strip().lower()
|
||
if not normalized_allowed:
|
||
return False
|
||
return hostname == normalized_allowed or hostname.endswith(f".{normalized_allowed}")
|
||
|
||
|
||
def _apply_style_to_prompt(prompt: str, style: str) -> str:
|
||
cleaned_prompt = (prompt or "").strip()
|
||
cleaned_style = (style or "").strip()
|
||
if not cleaned_style:
|
||
return cleaned_prompt
|
||
if cleaned_style.lower() in cleaned_prompt.lower():
|
||
return cleaned_prompt
|
||
if not cleaned_prompt:
|
||
return cleaned_style
|
||
return f"{cleaned_style}, {cleaned_prompt}"
|