08dc1f5c11
Bring the language-agnostic optimization engine from codeflash-next-gen. Includes protocols, strategy, config, models, AI client, telemetry, ranking, verification, and diff modules. TUI artifacts stripped.
34 lines
1.2 KiB
Python
34 lines
1.2 KiB
Python
from __future__ import annotations
|
|
|
|
from typing import TYPE_CHECKING
|
|
|
|
if TYPE_CHECKING:
|
|
from codeflash_core.models import BenchmarkResults, ScoredCandidate
|
|
|
|
|
|
def compute_speedup(baseline: BenchmarkResults, candidate: BenchmarkResults) -> float:
|
|
"""Compute speedup as percentage improvement: (baseline - candidate) / candidate.
|
|
|
|
Matches original codeflash performance_gain formula.
|
|
Returns 0.0 if candidate time is zero (no improvement measurable).
|
|
A positive value means the candidate is faster (e.g. 1.0 = 100% faster).
|
|
"""
|
|
if candidate.total_time <= 0:
|
|
return 0.0
|
|
return (baseline.total_time - candidate.total_time) / candidate.total_time
|
|
|
|
|
|
def score_candidate(speedup: float) -> float:
|
|
"""Score a candidate based on its speedup.
|
|
|
|
Currently score == speedup. This is the extension point for adding
|
|
more signals (code complexity, diff size, etc.) in the future.
|
|
"""
|
|
return speedup
|
|
|
|
|
|
def select_best(candidates: list[ScoredCandidate]) -> ScoredCandidate | None:
|
|
"""Select the best candidate by score. Returns None if list is empty."""
|
|
if not candidates:
|
|
return None
|
|
return max(candidates, key=lambda c: c.score)
|