wehub-resource-sync
149 lines
4.3 KiB
Python
149 lines
4.3 KiB
Python
from __future__ import annotations
|
|
|
|
from dataclasses import dataclass
|
|
|
|
|
|
@dataclass(frozen=True)
|
|
class ReplayEvent:
|
|
kind: str
|
|
payload: dict[str, object]
|
|
timestamp_ms: int
|
|
|
|
|
|
def _router_decision_payload(index: int, *, tier: str) -> dict[str, object]:
|
|
model = f"openrouter/{tier}-model-{index}"
|
|
return {
|
|
"tier": tier,
|
|
"model": model,
|
|
"baseline_model": "openrouter/frontier-baseline",
|
|
"source": "synthetic-replay",
|
|
"confidence": round(0.92 - index * 0.07, 2),
|
|
"savings_pct": 18 + index * 9,
|
|
"fallback": index == 3,
|
|
"thinking_mode": "observe" if index % 2 == 0 else "full",
|
|
"prompt_policy": "standard",
|
|
"routing_applied": index != 3,
|
|
"rollout_phase": f"phase-{index}",
|
|
}
|
|
|
|
|
|
def build_long_stream_events() -> list[ReplayEvent]:
|
|
events: list[ReplayEvent] = [
|
|
ReplayEvent(
|
|
"user_input",
|
|
{"text": "Explain the OpenSquilla routing decision stream."},
|
|
0,
|
|
),
|
|
ReplayEvent("router_decision", _router_decision_payload(0, tier="standard"), 1),
|
|
]
|
|
|
|
timestamp_ms = 2
|
|
tool_index = 0
|
|
for index in range(4_000):
|
|
section = index // 800
|
|
delta = f"stream-section-{section:02d}-delta-{index:04d}-"[:40]
|
|
events.append(ReplayEvent("text_delta", {"text": delta.ljust(40, ".")}, timestamp_ms))
|
|
timestamp_ms += 1
|
|
|
|
if (index + 1) % 800 == 0 and tool_index < 4:
|
|
tool_id = f"tool-{tool_index}"
|
|
events.append(
|
|
ReplayEvent(
|
|
"tool_start",
|
|
{
|
|
"name": "synthetic_tool",
|
|
"args": {"section": section},
|
|
"tool_use_id": tool_id,
|
|
},
|
|
timestamp_ms,
|
|
)
|
|
)
|
|
timestamp_ms += 1
|
|
events.append(
|
|
ReplayEvent(
|
|
"tool_finished",
|
|
{
|
|
"tool_use_id": tool_id,
|
|
"success": True,
|
|
"elapsed": 0.01 + tool_index * 0.001,
|
|
},
|
|
timestamp_ms,
|
|
)
|
|
)
|
|
timestamp_ms += 1
|
|
tool_index += 1
|
|
|
|
events.append(
|
|
ReplayEvent(
|
|
"done",
|
|
{
|
|
"usage": {
|
|
"input_tokens": 1_024,
|
|
"output_tokens": 42_000,
|
|
"cache_read_tokens": 512,
|
|
}
|
|
},
|
|
timestamp_ms,
|
|
)
|
|
)
|
|
return events
|
|
|
|
|
|
def build_dense_history_events() -> list[ReplayEvent]:
|
|
events: list[ReplayEvent] = []
|
|
timestamp_ms = 0
|
|
for index, tier in enumerate(("spark", "standard", "frontier", "fallback")):
|
|
events.append(
|
|
ReplayEvent(
|
|
"router_decision",
|
|
_router_decision_payload(index, tier=tier),
|
|
timestamp_ms,
|
|
)
|
|
)
|
|
timestamp_ms += 1
|
|
|
|
user_text = "user historical prompt line ".ljust(96, "u")
|
|
assistant_text = "assistant historical answer line ".ljust(160, "a")
|
|
for index in range(250):
|
|
events.append(
|
|
ReplayEvent(
|
|
"history_message",
|
|
{
|
|
"role": "user",
|
|
"content": f"{index:03d}: {user_text}",
|
|
},
|
|
timestamp_ms,
|
|
)
|
|
)
|
|
timestamp_ms += 1
|
|
events.append(
|
|
ReplayEvent(
|
|
"history_message",
|
|
{
|
|
"role": "assistant",
|
|
"content": f"{index:03d}: {assistant_text}",
|
|
},
|
|
timestamp_ms,
|
|
)
|
|
)
|
|
timestamp_ms += 1
|
|
|
|
for index in range(120):
|
|
events.append(
|
|
ReplayEvent(
|
|
"tool_card",
|
|
{
|
|
"tool_use_id": f"dense-tool-{index}",
|
|
"name": "synthetic_tool",
|
|
"summary": f"summary for dense tool card {index}",
|
|
"expanded_candidate": index < 20,
|
|
"line_count": 12 + index,
|
|
"rendered_bytes": 1_024 + index * 8,
|
|
},
|
|
timestamp_ms,
|
|
)
|
|
)
|
|
timestamp_ms += 1
|
|
|
|
return events
|