mirror of
https://github.com/fawney19/Aether.git
synced 2026-09-02 17:30:23 +08:00
6ec8df97e8
- StreamContext.append_text 增加 16KB 上限，超出后仅计数不存储， 避免长流式响应导致内存持续增长；token 估算改用 collected_text_length - 降低 DB 连接池上限 (30->15) 和 HTTP 连接池上限 (200->100) - tiktoken 编码器缓存从 32 缩减到 4（实际编码种类只有几种） - dev.sh 添加开发环境低配连接池默认值，uvicorn 热重载仅监视 src 目录
132 lines
4.2 KiB
Python
132 lines
4.2 KiB
Python
from src.api.handlers.base import stream_context
|
||
from src.api.handlers.base.stream_context import StreamContext
|
||
|
||
|
||
def test_collected_text_append_and_property() -> None:
|
||
ctx = StreamContext(model="test-model", api_format="openai:chat")
|
||
assert ctx.collected_text == ""
|
||
assert ctx.collected_text_length == 0
|
||
|
||
ctx.append_text("hello")
|
||
ctx.append_text(" ")
|
||
ctx.append_text("world")
|
||
assert ctx.collected_text == "hello world"
|
||
assert ctx.collected_text_length == len("hello world")
|
||
|
||
|
||
def test_collected_text_is_capped_but_total_length_is_preserved() -> None:
|
||
ctx = StreamContext(model="test-model", api_format="openai:chat")
|
||
cap = stream_context._MAX_COLLECTED_TEXT_CHARS
|
||
|
||
ctx.append_text("a" * (cap - 4))
|
||
ctx.append_text("b" * 10)
|
||
|
||
assert len(ctx.collected_text) == cap
|
||
assert ctx.collected_text == ("a" * (cap - 4)) + ("b" * 4)
|
||
assert ctx.collected_text_length == cap + 6
|
||
|
||
|
||
def test_reset_for_retry_clears_state() -> None:
|
||
ctx = StreamContext(model="test-model", api_format="openai:chat")
|
||
ctx.append_text("x")
|
||
ctx.update_usage(input_tokens=10, output_tokens=5)
|
||
ctx.parsed_chunks.append({"type": "chunk"})
|
||
ctx.chunk_count = 3
|
||
ctx.data_count = 2
|
||
ctx.has_completion = True
|
||
ctx.status_code = 418
|
||
ctx.error_message = "boom"
|
||
|
||
ctx.reset_for_retry()
|
||
|
||
assert ctx.collected_text == ""
|
||
assert ctx.input_tokens == 0
|
||
assert ctx.output_tokens == 0
|
||
assert ctx.parsed_chunks == []
|
||
assert ctx.chunk_count == 0
|
||
assert ctx.data_count == 0
|
||
assert ctx.has_completion is False
|
||
assert ctx.status_code == 200
|
||
assert ctx.error_message is None
|
||
|
||
|
||
def test_record_first_byte_time(monkeypatch) -> None:
|
||
"""测试记录首字时间"""
|
||
ctx = StreamContext(model="claude-3", api_format="claude_messages")
|
||
start_time = 100.0
|
||
monkeypatch.setattr(stream_context.time, "time", lambda: 100.0123) # 12.3ms
|
||
|
||
# 记录首字时间
|
||
ctx.record_first_byte_time(start_time)
|
||
|
||
# 验证首字时间已记录
|
||
assert ctx.first_byte_time_ms == 12
|
||
|
||
|
||
def test_record_first_byte_time_idempotent(monkeypatch) -> None:
|
||
"""测试首字时间只记录一次"""
|
||
ctx = StreamContext(model="claude-3", api_format="claude_messages")
|
||
start_time = 100.0
|
||
|
||
# 第一次记录
|
||
monkeypatch.setattr(stream_context.time, "time", lambda: 100.010)
|
||
ctx.record_first_byte_time(start_time)
|
||
first_value = ctx.first_byte_time_ms
|
||
|
||
# 第二次记录（应该被忽略）
|
||
monkeypatch.setattr(stream_context.time, "time", lambda: 100.020)
|
||
ctx.record_first_byte_time(start_time)
|
||
second_value = ctx.first_byte_time_ms
|
||
|
||
# 验证值没有改变
|
||
assert first_value == second_value
|
||
|
||
|
||
def test_reset_for_retry_clears_first_byte_time(monkeypatch) -> None:
|
||
"""测试重试时清除首字时间"""
|
||
ctx = StreamContext(model="claude-3", api_format="claude_messages")
|
||
start_time = 100.0
|
||
|
||
# 记录首字时间
|
||
monkeypatch.setattr(stream_context.time, "time", lambda: 100.010)
|
||
ctx.record_first_byte_time(start_time)
|
||
assert ctx.first_byte_time_ms is not None
|
||
|
||
# 重置
|
||
ctx.reset_for_retry()
|
||
|
||
# 验证首字时间已清除
|
||
assert ctx.first_byte_time_ms is None
|
||
|
||
|
||
def test_get_log_summary_with_first_byte_time() -> None:
|
||
"""测试日志摘要包含首字时间"""
|
||
ctx = StreamContext(model="claude-3", api_format="claude_messages")
|
||
ctx.provider_name = "anthropic"
|
||
ctx.input_tokens = 100
|
||
ctx.output_tokens = 50
|
||
ctx.first_byte_time_ms = 123
|
||
|
||
summary = ctx.get_log_summary("request-id-123", 456)
|
||
|
||
# 验证包含首字时间和总时间（大写格式）
|
||
assert "TTFB: 123ms" in summary
|
||
assert "Total: 456ms" in summary
|
||
assert "in:100 out:50" in summary
|
||
|
||
|
||
def test_get_log_summary_without_first_byte_time() -> None:
|
||
"""测试日志摘要在没有首字时间时的格式"""
|
||
ctx = StreamContext(model="claude-3", api_format="claude_messages")
|
||
ctx.provider_name = "anthropic"
|
||
ctx.input_tokens = 100
|
||
ctx.output_tokens = 50
|
||
# first_byte_time_ms 保持为 None
|
||
|
||
summary = ctx.get_log_summary("request-id-123", 456)
|
||
|
||
# 验证不包含首字时间标记,但有总时间（使用大写 TTFB 和 Total）
|
||
assert "TTFB:" not in summary
|
||
assert "Total: 456ms" in summary
|
||
assert "in:100 out:50" in summary
|