Gahow Wang
825062f554
The gateway records three epoch-ms stamps per request (admitted, forwarded to backend, backend first token). They reached *-raw.jsonl but stopped there: features.csv had only total latency, and none of the figures covered time-to-first-token. Derive ttft_ms, ttft_e2e_ms and tpot_ms in one helper shared by the streaming path (preparation.py) and the in-memory path (features.py) so the two cannot drift, and add figures 14 (TTFT, backend vs end-to-end) and 15 (TPOT). The two TTFT series are plotted together because their offset is the gateway admission delay, which is then readable straight off the chart. Rows missing a backend stamp and single-token responses are recorded as 0 and filtered at plot time; left in, they would stack at the origin and pull every percentile down. Figure 15 is a single panel: TPOT is near-deterministic (p50 21.9, p90 23.0), so a p90 zoom renders a flat line and a vertical edge. Verified by recomputing ttft_ms straight from the original SLS rows for all 2190 requests of a sample: 0 mismatches. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
98 lines
2.2 KiB
Python
98 lines
2.2 KiB
Python
from dataclasses import dataclass, field
|
|
|
|
|
|
@dataclass
|
|
class MessageEvent:
|
|
role: str
|
|
content_type: str
|
|
text_len: int
|
|
has_cache_control: bool = False
|
|
item_count: int = 0
|
|
|
|
|
|
@dataclass
|
|
class ToolSpec:
|
|
name: str
|
|
tool_type: str
|
|
|
|
|
|
@dataclass
|
|
class UsageStats:
|
|
input_tokens: int = 0
|
|
output_tokens: int = 0
|
|
total_tokens: int = 0
|
|
reasoning_tokens: int = 0
|
|
cached_tokens: int = 0
|
|
|
|
|
|
@dataclass
|
|
class RequestMeta:
|
|
provider: str
|
|
line_number: int
|
|
request_id: str
|
|
session_id: str
|
|
request_model: str
|
|
time: str
|
|
status_code: str
|
|
status_name: str
|
|
request_ready_time_ms: int
|
|
request_end_time_ms: int
|
|
total_cost_time_ms: int
|
|
backend_first_request_time_ms: int = 0
|
|
backend_first_response_time_ms: int = 0
|
|
|
|
|
|
@dataclass
|
|
class TraceRecord:
|
|
meta: RequestMeta
|
|
canonical_prompt: str = ""
|
|
messages: list[MessageEvent] = field(default_factory=list)
|
|
role_sequence: list[str] = field(default_factory=list)
|
|
declared_tools: list[ToolSpec] = field(default_factory=list)
|
|
usage: UsageStats = field(default_factory=UsageStats)
|
|
raw_messages: list[dict] = field(default_factory=list)
|
|
|
|
|
|
@dataclass
|
|
class TraceFeatures:
|
|
request_id: str
|
|
session_id: str
|
|
model: str
|
|
status_code: str
|
|
time: str
|
|
message_count: int
|
|
conversation_depth: int
|
|
declared_tool_count: int
|
|
assistant_msg_count: int
|
|
tool_msg_count: int
|
|
user_msg_count: int
|
|
system_msg_count: int
|
|
assistant_to_tool_count: int
|
|
tool_to_assistant_count: int
|
|
tool_to_tool_count: int
|
|
assistant_to_user_count: int
|
|
user_to_assistant_count: int
|
|
max_consecutive_tool_msgs: int
|
|
avg_tool_burst_len: float
|
|
has_tool_loop: int
|
|
input_tokens: int
|
|
output_tokens: int
|
|
total_tokens: int
|
|
reasoning_tokens: int
|
|
cached_tokens: int
|
|
cache_hit_ratio: float
|
|
uncached_prompt_tokens: int
|
|
output_input_ratio: float
|
|
latency_ms: int
|
|
ms_per_input_token: float
|
|
ms_per_output_token: float
|
|
long_context: int
|
|
high_cache: int
|
|
tool_burst_alert: int
|
|
tool_loop_alert: int
|
|
slow_request: int = 0
|
|
ttft_ms: int = 0
|
|
ttft_e2e_ms: int = 0
|
|
tpot_ms: float = 0.0
|
|
pattern_labels: list[str] = field(default_factory=list)
|