iomgaa/PolyGateway
1
0
Fork 0
Code Issues 3 Pull Requests Actions Packages Projects Releases 9 Wiki Activity

feat: add pricing table and cost calculation in telemetry emitter

Browse Source
This commit is contained in:
iomgaa
2026-07-21 00:54:10 -04:00
parent abb65c2324
commit d66299210b
4 changed files with 250 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/polygateway/client.py
+6 -1
View File
@@ -22,6 +22,7 @@ from polygateway.middleware.cache import CacheMW
from polygateway.middleware.retry import RetryMW
from polygateway.middleware.structured import StructuredMW
from polygateway.middleware.telemetry import TelemetryEmitter, TelemetryMW
from polygateway.pricing import PricingTable
from polygateway.providers import get_provider
from polygateway.sources import LeastInflightSelector, RoundRobinSelector, SourceCooldownMemo
from polygateway.transports.openai_compat import OpenAICompatTransport
@@ -68,6 +69,7 @@ class GatewayClient:
backpressure: BackpressurePolicy,
quota_full: str = "wait",
telemetry: TelemetryRecorder | None = None,
pricing: PricingTable | None = None,
cache: CacheBackend | None = None,
cache_namespace: str | None = None,
cache_ttl_s: int | None = None,
@@ -78,7 +80,7 @@ class GatewayClient:
sleep: Any = asyncio.sleep,
rng: Any = random.random,
) -> None:
emitter = TelemetryEmitter(telemetry) if telemetry is not None else None
emitter = TelemetryEmitter(telemetry, pricing=pricing) if telemetry is not None else None
terminal = RetryMW(
scope=scope,
sources=sources,
@@ -207,6 +209,9 @@ class GatewayClient:
backpressure=settings.backpressure,
quota_full=settings.quota_full,
telemetry=telemetry if telemetry is not None else _build_telemetry(settings),
pricing=PricingTable.from_file(settings.pricing_path)
if settings.pricing_path is not None
else None,
cache=cache if cache is not None else _build_cache(settings),
cache_namespace=settings.cache_namespace,
cache_ttl_s=settings.cache_ttl_s,
src/polygateway/middleware/telemetry.py
+12 -2
View File
@@ -23,14 +23,16 @@ if TYPE_CHECKING:
from collections.abc import Callable
from polygateway.ports import CallNext, TelemetryRecorder
from polygateway.pricing import PricingTable
from polygateway.types import ChatRequest, LLMResponse, SourceConfig
class TelemetryEmitter:
"""从请求与结果组装 18 字段并写入 recorder;一切写失败降级 warning。"""
def __init__(self, recorder: TelemetryRecorder) -> None:
def __init__(self, recorder: TelemetryRecorder, *, pricing: PricingTable | None = None) -> None:
self._recorder = recorder
self._pricing = pricing
async def emit_attempt(
self,
@@ -123,6 +125,14 @@ class TelemetryEmitter:
error: str | None,
) -> None:
try:
# 成本换算(M2 §6): 成功行按单价换算;缓存命中 0.0(未产生新调用);
# 失败/终态行 None;未注入价格表 = 恒 None(M1 现状)
if cache_hit:
cost: float | None = 0.0
elif error is None and model and self._pricing is not None:
cost = self._pricing.cost(model, prompt_tokens, completion_tokens)
else:
cost = None
# messages 落库前多模态摘要,与缓存 key 共用同一函数(VT R12)
messages_json = json.dumps(digest_messages(request.messages), ensure_ascii=False)
await self._recorder.record_llm_call(
@@ -143,7 +153,7 @@ class TelemetryEmitter:
max_inter_token_ms=max_inter_token_ms,
cache_hit=cache_hit,
error=error,
cost=None, # M1 无 pricing;M2 换算后填充
cost=cost,
)
except asyncio.CancelledError:
raise
src/polygateway/pricing.py
+73
View File
@@ -0,0 +1,73 @@
"""模型单价表与成本换算(M2 设计 §6;参考仓零先例,从头设计)。
**零内置单价**: 实验室走中转网关,计费非官方牌价;库内硬编码单价表
必然过时并掩盖真实成本(P5 严禁默认值掩盖错误)。价格一律由使用方
提供——JSON 文件(`PGW_PRICING_PATH`)或 dict 注入;币种由使用方全表
统一口径,库不设币种字段(18 字段冻结)。查不到的 model → cost=None
且每 model 仅首次 warning(防日志风暴),不阻塞调用。
"""
from __future__ import annotations
import json
from dataclasses import dataclass
from pathlib import Path
from typing import TYPE_CHECKING
from loguru import logger
if TYPE_CHECKING:
from collections.abc import Mapping
@dataclass(frozen=True)
class ModelPrice:
"""每百万 token 的输入/输出单价(币种由使用方口径统一)。"""
input_per_1m: float
output_per_1m: float
def __post_init__(self) -> None:
if self.input_per_1m < 0 or self.output_per_1m < 0:
raise ValueError("单价不能为负")
class PricingTable:
"""model → 单价 的只读表;cost() 是全库唯一换算点(经 TelemetryEmitter)。"""
def __init__(self, prices: Mapping[str, ModelPrice]) -> None:
self._prices = dict(prices)
self._warned: set[str] = set()
@classmethod
def from_file(cls, path: Path | str) -> PricingTable:
"""装配 fail-loud: 文件缺失/JSON 坏/条目缺字段/负单价一律 ValueError。"""
p = Path(path)
try:
data = json.loads(p.read_text(encoding="utf-8"))
except (OSError, json.JSONDecodeError) as exc:
raise ValueError(f"价格表 {p} 读取失败: {exc}") from exc
if not isinstance(data, dict):
raise ValueError(f"价格表 {p} 顶层必须是对象: model -> 单价")
prices: dict[str, ModelPrice] = {}
for model, entry in data.items():
if not isinstance(entry, dict) or not {"input_per_1m", "output_per_1m"} <= set(entry):
raise ValueError(f"价格表 {p} 条目 {model!r} 须含 input_per_1m 与 output_per_1m")
prices[model] = ModelPrice(
input_per_1m=float(entry["input_per_1m"]),
output_per_1m=float(entry["output_per_1m"]),
)
return cls(prices)
def cost(self, model: str, prompt_tokens: int, completion_tokens: int) -> float | None:
"""换算一次调用成本;未知 model 记 None 并仅首次 warning。"""
price = self._prices.get(model)
if price is None:
if model not in self._warned:
self._warned.add(model)
logger.warning("pricing 表无 model {!r} 的单价,cost 记 None", model)
return None
return (
prompt_tokens / 1_000_000 * price.input_per_1m
+ completion_tokens / 1_000_000 * price.output_per_1m
)
tests/unit/test_pricing.py
+159
View File
@@ -0,0 +1,159 @@
"""pricing.py 价格表与 TelemetryEmitter 成本换算(M2 设计 §6)。
零内置单价(实验室走中转网关,内置表必然过时掩盖真实成本,P5);
价格由使用方经 JSON 文件或 dict 注入;查不到 → cost=None + 每 model
仅首次 warning,不阻塞调用;计算点收敛在 Emitter(单一 helper 铁律)。
"""
import json
import pytest
from polygateway.middleware.telemetry import TelemetryEmitter
from polygateway.pricing import ModelPrice, PricingTable
from polygateway.types import ChatRequest, LLMResponse, SourceConfig
_TABLE = PricingTable(
{
"qwen-max": ModelPrice(input_per_1m=2.4, output_per_1m=9.6),
"free-model": ModelPrice(input_per_1m=0.0, output_per_1m=0.0),
}
)
class TestPricingTable:
def test_cost_formula_per_million(self):
# 1M prompt + 0.5M completion: 2.4 + 4.8
assert _TABLE.cost("qwen-max", 1_000_000, 500_000) == pytest.approx(7.2)
assert _TABLE.cost("qwen-max", 1000, 2000) == pytest.approx(0.0216)
assert _TABLE.cost("free-model", 999, 999) == 0.0
def test_unknown_model_returns_none(self):
assert _TABLE.cost("mystery", 100, 100) is None
def test_from_file_roundtrip(self, tmp_path):
path = tmp_path / "prices.json"
path.write_text(
json.dumps({"m1": {"input_per_1m": 1.0, "output_per_1m": 2.0}}), encoding="utf-8"
)
table = PricingTable.from_file(path)
assert table.cost("m1", 1_000_000, 1_000_000) == pytest.approx(3.0)
def test_from_file_missing_or_bad_fails_loudly(self, tmp_path):
with pytest.raises(ValueError, match="价格表"):
PricingTable.from_file(tmp_path / "nope.json")
bad = tmp_path / "bad.json"
bad.write_text("not json", encoding="utf-8")
with pytest.raises(ValueError, match="价格表"):
PricingTable.from_file(bad)
malformed = tmp_path / "malformed.json"
malformed.write_text(json.dumps({"m": {"input_per_1m": 1.0}}), encoding="utf-8")
with pytest.raises(ValueError, match="output_per_1m"):
PricingTable.from_file(malformed)
def test_negative_price_rejected(self):
with pytest.raises(ValueError):
ModelPrice(input_per_1m=-1.0, output_per_1m=0.0)
class _MemoryRecorder:
def __init__(self):
self.rows = []
async def record_llm_call(self, **fields):
self.rows.append(fields)
_REQ = ChatRequest(messages=[{"role": "user", "content": "hi"}])
def _resp(**overrides):
base = {
"content": "ok",
"thinking": "",
"model": "qwen-max",
"provider": "qwen",
"prompt_tokens": 1_000_000,
"completion_tokens": 500_000,
"latency_ms": 30,
"ttft_ms": None,
"max_inter_token_ms": None,
"cache_hit": False,
"call_id": "cid-1",
"source_name": "s1",
"usage_source": "measured",
}
base.update(overrides)
return LLMResponse(**base)
def _source(model="qwen-max"):
return SourceConfig(
name="s1",
provider="qwen",
base_url="https://gw.example/v1",
api_key="sk",
model=model,
timeout_s=10.0,
)
class TestEmitterCost:
async def test_success_row_costed(self):
rec = _MemoryRecorder()
emitter = TelemetryEmitter(rec, pricing=_TABLE)
await emitter.emit_attempt(
request=_REQ,
source=_source(),
call_id="c",
latency_ms=1,
response=_resp(),
error=None,
)
assert rec.rows[0]["cost"] == pytest.approx(7.2)
async def test_cache_hit_row_costs_zero(self):
rec = _MemoryRecorder()
emitter = TelemetryEmitter(rec, pricing=_TABLE)
await emitter.emit_cache_hit(request=_REQ, response=_resp(cache_hit=True))
assert rec.rows[0]["cost"] == 0.0
async def test_failure_row_cost_none(self):
rec = _MemoryRecorder()
emitter = TelemetryEmitter(rec, pricing=_TABLE)
await emitter.emit_attempt(
request=_REQ,
source=_source(),
call_id="c",
latency_ms=1,
response=None,
error="TransientError: boom",
)
assert rec.rows[0]["cost"] is None
async def test_unknown_model_none_without_blocking(self):
rec = _MemoryRecorder()
emitter = TelemetryEmitter(rec, pricing=_TABLE)
await emitter.emit_attempt(
request=_REQ,
source=_source(model="mystery"),
call_id="c",
latency_ms=1,
response=_resp(model="mystery"),
error=None,
)
assert rec.rows[0]["cost"] is None
async def test_no_pricing_keeps_none(self):
"""未注入价格表 = M1 现状: cost 恒 None(回归)。"""
rec = _MemoryRecorder()
emitter = TelemetryEmitter(rec)
await emitter.emit_attempt(
request=_REQ,
source=_source(),
call_id="c",
latency_ms=1,
response=_resp(),
error=None,
)
assert rec.rows[0]["cost"] is None