iomgaa
a1c4273a8b
resolve_thinking now takes an Effort instead of a tri-state bool, and the four gates become five. The new one sits ahead of the generic tier check on purpose: asking for `none` on GLM-5.3 used to fall through to "none is not supported, pick low/high/max", which loses both the fact that the model cannot stop reasoning and the one tier the caller could switch to right now. Without that alternative, downstream goes looking for extra_body — which is how issue #20 happened in the first place. The return type is a ThinkingResolution rather than the payload alone. Under fallback="nearest" the tier that goes out is not the tier that was asked for, and telemetry has to record the one that ran, or task 10 files a call under a tier it never used. Ties in that mapping go to the weaker side: a silent medium -> max is a multiple of the bill, and the library does not raise a caller's price on its own. Two readings the design left implicit, both settled the way its own compatibility promise requires: - `auto` is exempt from the tier list. It means "on, no tier named", which in the body is the absence of the effort key, not a value of it. Checking it against the list would break every existing source that sets ENABLE_THINKING=true against deepseek-v4 or glm-5.3. - `none` is never a mapping target. Turning "think less" into "do not think" reverses the decision instead of cheapening it; a switch-only model maps to `auto` and a model that only has `none` still errors. Both call sites convert enable_thinking in place for now; task 5 folds that into effective_effort along with the source- and call-level tiers.
632 lines
27 KiB
Python
632 lines
27 KiB
Python
"""OpenAI 兼容 transport(D2 默认): 手写 httpx + SSE 解析 + 错误翻译。
|
|
|
|
SSE 纯函数移植 VT `adapters/llm.py:51-124`;错误翻译移植 CHS
|
|
`app/providers/invokers.py:127-227`。职责只到"一次原始调用"——重试/限流/
|
|
缓存归中间件。看门狗活性口径: content 与 reasoning_content 增量都作为流
|
|
元素产出,思考流天然刷新计时(CHS 迁移约束 R1)。
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import json
|
|
import re
|
|
import time
|
|
from typing import TYPE_CHECKING, Any
|
|
|
|
import httpx
|
|
from loguru import logger
|
|
|
|
from polygateway.errors import (
|
|
PolyGatewayError,
|
|
RequestRejectedError,
|
|
ResultInvalidError,
|
|
SourceDeadError,
|
|
TransientError,
|
|
)
|
|
from polygateway.providers import ProviderProfile, get_provider
|
|
from polygateway.streaming import StreamLivenessTimeout, stream_with_liveness_timeouts
|
|
from polygateway.thinking import (
|
|
ThinkingCapability,
|
|
ThinkingUnsupportedError,
|
|
get_capability,
|
|
observe_thinking,
|
|
reconcile_thinking,
|
|
resolve_thinking,
|
|
)
|
|
from polygateway.transports._http_errors import compose_message, summarize_body
|
|
from polygateway.types import (
|
|
Effort,
|
|
EmbeddingTransportResult,
|
|
SourceConfig,
|
|
TransportResult,
|
|
)
|
|
|
|
if TYPE_CHECKING:
|
|
from collections.abc import AsyncIterator, Callable, Mapping
|
|
|
|
_THINK_PATTERN = re.compile(r"<think>(.*?)</think>", re.DOTALL)
|
|
|
|
# —— SSE 纯函数(VT llm.py 同款)——
|
|
|
|
|
|
def _sse_data_payload(raw: str) -> str | None:
|
|
"""提取 SSE data 行载荷;ping(: 开头)/空行/非 data 行返回 None 跳过。"""
|
|
line = raw.strip()
|
|
if not line or line.startswith(":") or not line.startswith("data:"):
|
|
return None
|
|
return line[len("data:") :].strip()
|
|
|
|
|
|
def _sse_delta(chunk: dict[str, Any], usage_sink: dict[str, Any]) -> tuple[bool, str] | None:
|
|
"""从 chunk 提取增量: (True, content) 或 (False, reasoning);usage 帧旁路进 sink。"""
|
|
if chunk.get("usage"):
|
|
usage_sink["usage"] = chunk["usage"]
|
|
if "model" not in usage_sink:
|
|
# 首个**有效**值即固定: 末帧的异常值不得覆盖它;但首帧报空串也不能锁死
|
|
# sink——否则后续真实版本会丢(issue #3)
|
|
reported = _coerce_model_reported(chunk.get("model"))
|
|
if reported is not None:
|
|
usage_sink["model"] = reported
|
|
choices = chunk.get("choices") or []
|
|
if not choices:
|
|
return None
|
|
delta = choices[0].get("delta") or {}
|
|
content = delta.get("content")
|
|
if content:
|
|
return (True, content)
|
|
reasoning = delta.get("reasoning_content")
|
|
if reasoning:
|
|
return (False, reasoning)
|
|
return None
|
|
|
|
|
|
async def _iter_sse_deltas(
|
|
lines: AsyncIterator[str], usage_sink: dict[str, Any]
|
|
) -> AsyncIterator[tuple[bool, str]]:
|
|
"""逐行解析 SSE 流;[DONE] 置 sink["done"];畸形 JSON 帧 → 瞬时错误(可重试)。"""
|
|
async for raw in lines:
|
|
data = _sse_data_payload(raw)
|
|
if data is None:
|
|
continue
|
|
if data == "[DONE]":
|
|
usage_sink["done"] = True
|
|
return
|
|
try:
|
|
chunk = json.loads(data)
|
|
except json.JSONDecodeError as exc:
|
|
raise TransientError(
|
|
f"SSE 帧畸形(malformed_json): {data[:80]!r}", operation="chat"
|
|
) from exc
|
|
delta = _sse_delta(chunk, usage_sink)
|
|
if delta is not None:
|
|
yield delta
|
|
|
|
|
|
# —— 错误翻译(CHS invokers.py 同款)——
|
|
|
|
|
|
def _parse_retry_after(raw: str | None) -> float | None:
|
|
"""解析 Retry-After 头;仅支持秒数形态,HTTP-date 返回 None(CHS 同款)。"""
|
|
if raw is None:
|
|
return None
|
|
try:
|
|
seconds = float(raw.strip())
|
|
except ValueError:
|
|
return None
|
|
return seconds if seconds > 0 else None
|
|
|
|
|
|
def _translate_429(
|
|
source: SourceConfig, body_text: str, headers: Mapping[str, str], ctx: dict[str, Any]
|
|
) -> Exception:
|
|
"""429 细分。`body_text` 必须是**未截断的原文**——`ctx["body_text"]` 是摘要,
|
|
头尾保留会破坏 JSON 结构,拿它解析会让超长 body 的配额耗尽退化成普通限速
|
|
(该源不再 force_open),把一个诊断改进变成治理 bug(issue #10 实现红线)。
|
|
"""
|
|
try:
|
|
err_type = json.loads(body_text).get("error", {}).get("type", "")
|
|
except (json.JSONDecodeError, AttributeError):
|
|
err_type = ""
|
|
summary = ctx["body_text"]
|
|
if err_type == "insufficient_quota":
|
|
return SourceDeadError(
|
|
compose_message(f"{source.name} 配额耗尽(insufficient_quota)", summary), **ctx
|
|
)
|
|
return TransientError(
|
|
compose_message(f"{source.name} 限速: 429", summary),
|
|
retry_after_s=_parse_retry_after(headers.get("retry-after")),
|
|
**ctx,
|
|
)
|
|
|
|
|
|
def _classify(status: int) -> tuple[type[PolyGatewayError], str]:
|
|
"""状态码 → (错误类, message 标签);映射与 ARCH §6.2 逐条相同,本次零变更。"""
|
|
if status in (401, 403):
|
|
return SourceDeadError, "凭据失效/欠费"
|
|
if status == 400:
|
|
return RequestRejectedError, "请求被拒"
|
|
if status >= 500:
|
|
return TransientError, "瞬时错误"
|
|
return RequestRejectedError, "客户端错误"
|
|
|
|
|
|
def _status_to_error(
|
|
source: SourceConfig, status: int, body_text: str, headers: Mapping[str, str]
|
|
) -> Exception:
|
|
"""非 2xx → 领域错误,**全部分支**携带响应体摘要(issue #10)。
|
|
|
|
摘要只算一次,message 与 `body_text` 共用同一份串: 两份不同长度会让"遥测里
|
|
看到的"与"下游 catch 到的"对不上,排查时反而多一层困惑。
|
|
"""
|
|
summary = summarize_body(body_text)
|
|
ctx: dict[str, Any] = {
|
|
"source_name": source.name,
|
|
"status_code": status,
|
|
"operation": "chat",
|
|
"body_text": summary,
|
|
}
|
|
if status == 429:
|
|
return _translate_429(source, body_text, headers, ctx)
|
|
cls, label = _classify(status)
|
|
return cls(compose_message(f"{source.name} {label}: {status}", summary), **ctx)
|
|
|
|
|
|
def _strip_think(content: str) -> tuple[str, str]:
|
|
"""剥离 <think> 标签(qwen 系),返回 (正文, 思考流)。VT llm.py:147-164 同款。"""
|
|
match = _THINK_PATTERN.search(content)
|
|
if match is None:
|
|
return content, ""
|
|
return _THINK_PATTERN.sub("", content).strip(), match.group(1).strip()
|
|
|
|
|
|
def _resolve_usage(usage: dict[str, Any]) -> tuple[int, int, str]:
|
|
"""usage 帧读取;缺失/非法记 0/0 并标 unavailable(est_tokens 解耦设计 §3.2 #3)。
|
|
|
|
不再拿 `est_tokens` 兜底: 它按 CHS 定义是"最坏情形上界",拿上界当实测值
|
|
只会系统性高估账单;宁可把用量记成显式的"不可得"(cost 随之为 NULL),
|
|
让缺口可被统计,也不编一个看似有效的数字。用量口径自此不依赖源配置,
|
|
故不再收 `SourceConfig`。
|
|
"""
|
|
prompt, completion = usage.get("prompt_tokens"), usage.get("completion_tokens")
|
|
if isinstance(prompt, int) and isinstance(completion, int) and prompt + completion > 0:
|
|
return prompt, completion, "measured"
|
|
return 0, 0, "unavailable"
|
|
|
|
|
|
def _coerce_cached_tokens(usage: Any) -> int | None:
|
|
"""取 usage.prompt_tokens_details.cached_tokens(issue #3);形态异常一律 None。
|
|
|
|
`0` 与 `None` 必须可区分: 前者是"该源上报了一次真实零命中",后者是"该源
|
|
不报这个数",下游对两者的处置不同(后者不可做缓存成本校正)。故只把
|
|
**负数与非整数**归 None,`0` 如实保留。`bool` 显式排除——isinstance(True, int)
|
|
在 Python 里为真,放行会把 `True` 记成 1 个命中 token。
|
|
"""
|
|
if not isinstance(usage, dict):
|
|
return None
|
|
details = usage.get("prompt_tokens_details")
|
|
if not isinstance(details, dict):
|
|
return None
|
|
cached = details.get("cached_tokens")
|
|
if isinstance(cached, bool) or not isinstance(cached, int) or cached < 0:
|
|
return None
|
|
return cached
|
|
|
|
|
|
def _coerce_reasoning_tokens(usage: Any) -> int | None:
|
|
"""取 usage.completion_tokens_details.reasoning_tokens(issue #6);形态异常一律 None。
|
|
|
|
与 `_coerce_cached_tokens` 逐条同构(两者是 OpenAI 兼容 usage 里对称的一对):
|
|
`0` 如实保留、负数与非整数归 None、`bool` 显式排除。差别只在语义——本字段
|
|
的 None 是"**本次调用**未上报"而非"该源不上报": 中转在上游不返回 usage 时
|
|
会本地补算并整体替换 usage 对象,把 details 一并吃掉(findings §4c)。
|
|
"""
|
|
if not isinstance(usage, dict):
|
|
return None
|
|
details = usage.get("completion_tokens_details")
|
|
if not isinstance(details, dict):
|
|
return None
|
|
reasoning = details.get("reasoning_tokens")
|
|
if isinstance(reasoning, bool) or not isinstance(reasoning, int) or reasoning < 0:
|
|
return None
|
|
return reasoning
|
|
|
|
|
|
def _coerce_model_reported(value: Any) -> str | None:
|
|
"""取响应体的 model 字段(issue #3);非 str 或空白串一律 None,收口时去空白。
|
|
|
|
去空白不是洁癖: 下游拿这个串做实验快照的 key,`" m "` 与 `"m"` 会造成假分叉。
|
|
"""
|
|
if not isinstance(value, str) or not value.strip():
|
|
return None
|
|
return value.strip()
|
|
|
|
|
|
def _resolve_stream_usage(sink: dict[str, Any], salvaged: bool) -> tuple[int, int, str]:
|
|
"""流式用量口径: 打捞路径把 measured 降级为 estimated,unavailable 原样保留。
|
|
|
|
前置条件不可省(解耦设计 §3.2 #4): usage 帧本就缺失时 `0/0` 会被洗成
|
|
`estimated`,进而按 token 换算出一个假的 `0.0` 成本。
|
|
"""
|
|
prompt, completion, usage_source = _resolve_usage(sink.get("usage") or {})
|
|
if salvaged and usage_source == "measured":
|
|
# 收到 usage 帧但流被截断: 数字真实、可信度降级(M1 设计 §6)
|
|
usage_source = "estimated"
|
|
return prompt, completion, usage_source
|
|
|
|
|
|
def _extract_vectors(
|
|
data: dict[str, Any], source: SourceConfig, expected_count: int, ctx: dict[str, Any]
|
|
) -> list[list[float]]:
|
|
"""按 data[].index 重排提取向量并校验条数/维度一致性。"""
|
|
try:
|
|
vectors = [
|
|
[float(x) for x in item["embedding"]]
|
|
for item in sorted(data["data"], key=lambda it: int(it["index"]))
|
|
]
|
|
except (KeyError, TypeError, ValueError) as exc:
|
|
raise ResultInvalidError(f"{source.name} embedding 响应形态异常: {exc}", **ctx) from exc
|
|
if len(vectors) != expected_count:
|
|
raise ResultInvalidError(
|
|
f"{source.name} 返回 {len(vectors)} 条向量,与输入 {expected_count} 条不符", **ctx
|
|
)
|
|
if len({len(v) for v in vectors}) != 1 or not vectors[0]:
|
|
raise ResultInvalidError(
|
|
f"{source.name} 向量维度异常: {sorted({len(v) for v in vectors})}", **ctx
|
|
)
|
|
return vectors
|
|
|
|
|
|
def _resolve_embedding_usage(data: dict[str, Any]) -> tuple[int, str]:
|
|
"""usage 读取;缺失/非法记 0 并标 unavailable(与 chat 同口径,设计 §3.2 #3)。"""
|
|
prompt = (data.get("usage") or {}).get("prompt_tokens")
|
|
if isinstance(prompt, int) and prompt > 0:
|
|
return prompt, "measured"
|
|
return 0, "unavailable"
|
|
|
|
|
|
def _parse_embedding_payload(
|
|
resp: httpx.Response, source: SourceConfig, expected_count: int
|
|
) -> EmbeddingTransportResult:
|
|
"""解析 /embeddings 响应;一切形态异常归 ResultInvalidError(坏结果≠坏服务)。"""
|
|
ctx: dict[str, Any] = {"source_name": source.name, "operation": "embedding"}
|
|
try:
|
|
data = resp.json()
|
|
except json.JSONDecodeError as exc:
|
|
raise ResultInvalidError(f"{source.name} embedding 响应非 JSON: {exc}", **ctx) from exc
|
|
vectors = _extract_vectors(data, source, expected_count, ctx)
|
|
prompt_tokens, usage_source = _resolve_embedding_usage(data)
|
|
return EmbeddingTransportResult(
|
|
vectors=vectors,
|
|
dim=len(vectors[0]),
|
|
prompt_tokens=prompt_tokens,
|
|
usage_source=usage_source,
|
|
raw={"id": data.get("id")},
|
|
)
|
|
|
|
|
|
def _default_client_factory(source: SourceConfig) -> httpx.AsyncClient:
|
|
return httpx.AsyncClient(
|
|
headers={"Authorization": f"Bearer {source.api_key}"},
|
|
timeout=httpx.Timeout(source.timeout_s),
|
|
trust_env=source.trust_env,
|
|
)
|
|
|
|
|
|
class OpenAICompatTransport:
|
|
"""默认 transport: 每源一个预配 httpx client,懒创建,aclose 统一释放。"""
|
|
|
|
def __init__(
|
|
self,
|
|
*,
|
|
registry: Mapping[str, ProviderProfile] | None = None,
|
|
capabilities: Mapping[str, ThinkingCapability] | None = None,
|
|
client_factory: Callable[[SourceConfig], httpx.AsyncClient] | None = None,
|
|
) -> None:
|
|
self._registry = registry
|
|
self._capabilities = capabilities
|
|
# 未登记模型只喊一次: 装配期已喊过,逐次调用再喊是日志洪水。
|
|
# 实例级而非模块级 —— 模块级可变状态违反纯 asyncio 中立铁律
|
|
self._warned_models: set[str] = set()
|
|
# 对账告警独立节流,**不复用** `_warned_models`: 两者语义不同(那个 set 记
|
|
# 的是"未登记能力已告警过",这个记的是"某源某方向的矛盾已告警过"),共用
|
|
# 一个容器会让两种告警的生命周期纠缠在一起——将来任一侧想加清空/过期策略,
|
|
# 都会连带改掉另一侧的行为。(键空间恰好不相交,故当下**不会**互相压制;
|
|
# 分开维护的理由是语义,不是碰撞)
|
|
self._warned_mismatches: set[tuple[str, str, bool | None]] = set()
|
|
self._client_factory = client_factory or _default_client_factory
|
|
self._clients: dict[str, httpx.AsyncClient] = {}
|
|
|
|
def _client_for(self, source: SourceConfig) -> httpx.AsyncClient:
|
|
client = self._clients.get(source.name)
|
|
if client is None:
|
|
client = self._client_factory(source)
|
|
self._clients[source.name] = client
|
|
return client
|
|
|
|
def _build_payload(
|
|
self,
|
|
*,
|
|
messages: list[dict[str, Any]],
|
|
source: SourceConfig,
|
|
profile: ProviderProfile,
|
|
stream: bool,
|
|
overlay: dict[str, Any],
|
|
) -> dict[str, Any]:
|
|
payload: dict[str, Any] = {"model": source.model, "messages": messages, "stream": stream}
|
|
if stream:
|
|
payload["stream_options"] = {"include_usage": True} # 强制 usage 帧(三项目同款)
|
|
# 形态(provider 级)与能力(model 级)在此相遇;不可满足时 ValueError,
|
|
# 由 complete() 翻译为四分类之一(issue #5)
|
|
capability = get_capability(source.model, table=self._capabilities)
|
|
first_time = source.model not in self._warned_models
|
|
self._warned_models.add(source.model)
|
|
# `enable_thinking` 的档位语法糖(True → auto,False → none,None 不表态);
|
|
# 就地转换是过渡形态,T5 起由 thinking.effective_effort() 统一收口并接上
|
|
# 源级/请求级档位(设计 §4.2)
|
|
enabled = source.enable_thinking
|
|
effort = None if enabled is None else (Effort.AUTO if enabled else Effort.NONE)
|
|
payload.update(
|
|
resolve_thinking(
|
|
profile,
|
|
capability,
|
|
effort,
|
|
model=source.model,
|
|
warn_unregistered=first_time,
|
|
).payload
|
|
)
|
|
# 顺序即优先级(issue #4 设计决策 A): 配置级 extra_body 在前,调用级
|
|
# overlay(含结构化注入)在后覆盖之。两行不可调换
|
|
payload.update(source.extra_body)
|
|
payload.update(overlay)
|
|
return payload
|
|
|
|
async def complete(
|
|
self,
|
|
*,
|
|
messages: list[dict[str, Any]],
|
|
source: SourceConfig,
|
|
stream: bool,
|
|
overlay: dict[str, Any],
|
|
call_id: str,
|
|
) -> TransportResult:
|
|
"""一次原始调用;HTTP/线路/流式异常按 ARCH §6.2 翻译为领域错误。"""
|
|
profile = get_provider(source.provider, registry=self._registry)
|
|
try:
|
|
payload = self._build_payload(
|
|
messages=messages, source=source, profile=profile, stream=stream, overlay=overlay
|
|
)
|
|
except ThinkingUnsupportedError as exc:
|
|
# 推理开关不可满足是**请求本身**的问题: 换源重试都救不了它。只捕这个
|
|
# 专用类型而非宽 catch ValueError —— 后者会把序列化等无关错误误贴标签
|
|
raise RequestRejectedError(
|
|
f"{source.name} 推理开关无法满足: {exc}",
|
|
source_name=source.name,
|
|
operation="chat",
|
|
) from exc
|
|
url = source.base_url.rstrip("/") + "/chat/completions"
|
|
client = self._client_for(source)
|
|
ctx: dict[str, Any] = {"source_name": source.name, "operation": "chat"}
|
|
try:
|
|
if stream:
|
|
result = await self._complete_stream(client, url, payload, source, profile)
|
|
else:
|
|
result = await self._complete_once(client, url, payload, source, profile)
|
|
except StreamLivenessTimeout as exc:
|
|
raise TransientError(f"{source.name} 流活性超时({exc.kind})", **ctx) from exc
|
|
except httpx.TimeoutException as exc:
|
|
raise TransientError(f"{source.name} 超时: {exc}", **ctx) from exc
|
|
except httpx.TransportError as exc:
|
|
# VT 宽集: 覆盖断连/协议错误/读写失败(设计 §9 行 8)
|
|
raise TransientError(f"{source.name} 网络错误: {exc}", **ctx) from exc
|
|
# 此处是唯一同时握有请求方向与响应结果的地方,对账只能落在这里
|
|
self._warn_on_thinking_mismatch(source, result)
|
|
return result
|
|
|
|
def _warn_on_thinking_mismatch(self, source: SourceConfig, result: TransportResult) -> None:
|
|
"""声明与观测矛盾即 warning;按 (source, model, direction) 节流,同组合只喊一次。
|
|
|
|
三段缺一不可。**方向**: 同一模型的开、关两档是两个独立的矛盾。**源名**:
|
|
多源多账号是本库的核心场景,同一 model 跨 N 个源是常态,而每个源背后是
|
|
独立的账号/网关,一个源的行为不代表另一个——漏掉源名,5 个源里第一个出
|
|
问题的喊完一次,其余四个永久静音。逐次调用刷屏会把告警变成噪声,噪声等于
|
|
没有告警。
|
|
|
|
**先判键再对账**: `reconcile_thinking` 会拼含完整 `evidence` 的长字符串,
|
|
而非流式档每次调用都命中这一分支,节流后再拼是纯粹的热路径浪费。
|
|
"""
|
|
key = (source.name, source.model, source.enable_thinking)
|
|
if key in self._warned_mismatches:
|
|
return
|
|
message = reconcile_thinking(
|
|
enable_thinking=source.enable_thinking,
|
|
observation=result.thinking_observation,
|
|
capability=get_capability(source.model, table=self._capabilities),
|
|
model=source.model,
|
|
)
|
|
if message is None:
|
|
return
|
|
self._warned_mismatches.add(key)
|
|
# 源名拼在调用点而不是加进 `reconcile_thinking` 的签名: 那是纯判定函数,
|
|
# 输入只该含判定依据(声明/观测/能力/模型),源名是**定位信息**,进不了判据。
|
|
# 单参数传入 loguru: 文案里带 `thinking:{type:disabled}` 这类字面花括号
|
|
# (能力表 evidence),将来有人给这行加个格式化参数就会炸在成功调用的返回
|
|
# 路径上(与 telemetry/sqlite.py 的缺列告警同一先例)
|
|
logger.warning("源 {} —— {}", source.name, message)
|
|
|
|
async def embed(
|
|
self, *, texts: list[str], source: SourceConfig, call_id: str
|
|
) -> EmbeddingTransportResult:
|
|
"""一次原始 embedding 调用(M2 §7): POST /embeddings,错误翻译同 chat。
|
|
|
|
响应按 data[].index 重排保序(GovDoc embedding.py:149 / VT :164 同款);
|
|
空 data/长度不符/维度不一致 → ResultInvalidError(坏结果不熔断)。
|
|
"""
|
|
if not texts:
|
|
raise ValueError("texts 不能为空(空输入由 EmbeddingClient 短路)")
|
|
url = source.base_url.rstrip("/") + "/embeddings"
|
|
client = self._client_for(source)
|
|
ctx: dict[str, Any] = {"source_name": source.name, "operation": "embedding"}
|
|
try:
|
|
resp = await client.post(url, json={"model": source.model, "input": texts})
|
|
except httpx.TimeoutException as exc:
|
|
raise TransientError(f"{source.name} 超时: {exc}", **ctx) from exc
|
|
except httpx.TransportError as exc:
|
|
raise TransientError(f"{source.name} 网络错误: {exc}", **ctx) from exc
|
|
if resp.status_code != 200:
|
|
raise _status_to_error(source, resp.status_code, resp.text, resp.headers)
|
|
return _parse_embedding_payload(resp, source, len(texts))
|
|
|
|
async def _complete_stream(
|
|
self,
|
|
client: httpx.AsyncClient,
|
|
url: str,
|
|
payload: dict[str, Any],
|
|
source: SourceConfig,
|
|
profile: ProviderProfile,
|
|
) -> TransportResult:
|
|
started = time.monotonic()
|
|
async with client.stream("POST", url, json=payload) as resp:
|
|
if resp.status_code != 200:
|
|
body = (await resp.aread()).decode("utf-8", errors="replace")
|
|
raise _status_to_error(source, resp.status_code, body, resp.headers)
|
|
sink: dict[str, Any] = {}
|
|
guarded = stream_with_liveness_timeouts(
|
|
_iter_sse_deltas(resp.aiter_lines(), sink),
|
|
ttft_s=source.ttft_timeout_s or source.timeout_s,
|
|
inter_token_s=source.inter_token_timeout_s or source.timeout_s,
|
|
total_s=source.timeout_s,
|
|
)
|
|
content_parts: list[str] = []
|
|
thinking_parts: list[str] = []
|
|
ttft_ms: float | None = None
|
|
last = started
|
|
max_gap = 0.0
|
|
async for is_content, text in guarded:
|
|
now = time.monotonic()
|
|
if ttft_ms is None:
|
|
ttft_ms = (now - started) * 1000
|
|
else:
|
|
max_gap = max(max_gap, (now - last) * 1000)
|
|
last = now
|
|
(content_parts if is_content else thinking_parts).append(text)
|
|
salvaged = self._check_done(sink, content_parts, thinking_parts, source)
|
|
content, thinking = self._finalize_text(content_parts, thinking_parts, profile)
|
|
self._reject_empty_completion(content, source)
|
|
prompt, completion, usage_source = _resolve_stream_usage(sink, salvaged)
|
|
reasoning_tokens = _coerce_reasoning_tokens(sink.get("usage"))
|
|
return TransportResult(
|
|
content=content,
|
|
thinking=thinking,
|
|
prompt_tokens=prompt,
|
|
completion_tokens=completion,
|
|
usage_source=usage_source,
|
|
ttft_ms=ttft_ms,
|
|
max_inter_token_ms=(max_gap if ttft_ms is not None else None),
|
|
raw={"usage": sink.get("usage")},
|
|
cached_prompt_tokens=_coerce_cached_tokens(sink.get("usage")),
|
|
model_reported=_coerce_model_reported(sink.get("model")),
|
|
reasoning_tokens=reasoning_tokens,
|
|
# 两条组装路径必须同口径裁定: 只在一条路径上给结论,下游就得靠
|
|
# "这次是不是流式"去猜可观测性,那正是 issue #16/#17 的根因形态
|
|
thinking_observation=observe_thinking(
|
|
thinking=thinking, reasoning_tokens=reasoning_tokens
|
|
),
|
|
)
|
|
|
|
def _check_done(
|
|
self,
|
|
sink: dict[str, Any],
|
|
content_parts: list[str],
|
|
thinking_parts: list[str],
|
|
source: SourceConfig,
|
|
) -> bool:
|
|
"""缺 [DONE] 语义(设计 §6): 零内容恒 retry;有内容按 missing_done 策略。"""
|
|
if sink.get("done"):
|
|
return False
|
|
ctx: dict[str, Any] = {"source_name": source.name, "operation": "chat"}
|
|
if not content_parts and not thinking_parts:
|
|
raise TransientError(f"{source.name} SSE early_eof: 零内容断流", **ctx)
|
|
if source.missing_done == "retry":
|
|
raise TransientError(f"{source.name} SSE missing_done: 截断且无 [DONE]", **ctx)
|
|
return True
|
|
|
|
def _reject_empty_completion(self, content: str, source: SourceConfig) -> None:
|
|
"""空补全 → 瞬时错误(2026-07-20 人类裁决,M1 验证发现)。
|
|
|
|
服务 200 且流程完整([DONE]/usage 正常)但 content 为空——MiniMax 等
|
|
网关的间歇异常形态。视为服务抖动: 退避重试/换源,**绝不缓存空响应**;
|
|
承 CHS "VLM 零 content"归瞬时的先例(invokers.py:309)。
|
|
"""
|
|
if not content.strip():
|
|
raise TransientError(
|
|
f"{source.name} 空补全(empty_completion): 流程完整但零内容",
|
|
source_name=source.name,
|
|
operation="chat",
|
|
)
|
|
|
|
def _finalize_text(
|
|
self, content_parts: list[str], thinking_parts: list[str], profile: ProviderProfile
|
|
) -> tuple[str, str]:
|
|
content = "".join(content_parts)
|
|
thinking = "".join(thinking_parts)
|
|
if profile.strip_think_tags and "<think>" in content:
|
|
content, tag_thinking = _strip_think(content)
|
|
if tag_thinking:
|
|
thinking = tag_thinking
|
|
return content, thinking
|
|
|
|
async def _complete_once(
|
|
self,
|
|
client: httpx.AsyncClient,
|
|
url: str,
|
|
payload: dict[str, Any],
|
|
source: SourceConfig,
|
|
profile: ProviderProfile,
|
|
) -> TransportResult:
|
|
"""非流式快路径(三项目均无,库新增): 单 JSON 响应,仅 total 超时。"""
|
|
resp = await client.post(url, json=payload)
|
|
if resp.status_code != 200:
|
|
raise _status_to_error(source, resp.status_code, resp.text, resp.headers)
|
|
try:
|
|
body = resp.json()
|
|
except json.JSONDecodeError as exc:
|
|
raise TransientError(
|
|
f"{source.name} 非流式响应非法 JSON", source_name=source.name, operation="chat"
|
|
) from exc
|
|
choices = body.get("choices") or []
|
|
if not choices:
|
|
raise TransientError(
|
|
f"{source.name} 响应缺 choices", source_name=source.name, operation="chat"
|
|
)
|
|
message = choices[0].get("message") or {}
|
|
content, thinking = self._finalize_text(
|
|
[message.get("content") or ""], [message.get("reasoning_content") or ""], profile
|
|
)
|
|
self._reject_empty_completion(content, source)
|
|
prompt, completion, usage_source = _resolve_usage(body.get("usage") or {})
|
|
reasoning_tokens = _coerce_reasoning_tokens(body.get("usage"))
|
|
return TransportResult(
|
|
content=content,
|
|
thinking=thinking,
|
|
prompt_tokens=prompt,
|
|
completion_tokens=completion,
|
|
usage_source=usage_source,
|
|
ttft_ms=None,
|
|
max_inter_token_ms=None,
|
|
raw={"usage": body.get("usage")},
|
|
cached_prompt_tokens=_coerce_cached_tokens(body.get("usage")),
|
|
model_reported=_coerce_model_reported(body.get("model")),
|
|
reasoning_tokens=reasoning_tokens,
|
|
# 本路径的裁定多半落 UNKNOWN(M3 实测: 推理已计费却正文与 details 双
|
|
# 缺)。如实标记"观测不到",好过让下游误读成"没推理"
|
|
thinking_observation=observe_thinking(
|
|
thinking=thinking, reasoning_tokens=reasoning_tokens
|
|
),
|
|
)
|
|
|
|
async def aclose(self) -> None:
|
|
"""幂等关闭全部源 client。"""
|
|
clients, self._clients = self._clients, {}
|
|
for client in clients.values():
|
|
await client.aclose()
|