code_pipeline/data_construction/utils/session_utils.py

from __future__ import annotations
from datetime import datetime, timedelta
import random


def format_session_as_context(sessions: list[dict], include_timestamps: bool = True) -> str:
    parts = []
    for sess in sessions:
        header = f"[Session: {sess['session_id']}"
        if include_timestamps and "timestamp" in sess:
            header += f" | {sess['timestamp']}"
        header += "]"
        parts.append(header)
        for turn in sess.get("turns", []):
            parts.append(f"  {turn['role'].capitalize()}: {turn['content']}")
        parts.append("")
    return "\n".join(parts)


def format_evidence_sessions_text(sessions: list[dict]) -> str:
    parts = []
    for i, sess in enumerate(sessions, 1):
        parts.append(f"--- Evidence Session {i} (ID: {sess['session_id']}) ---")
        for turn in sess.get("turns", []):
            parts.append(f"{turn['role'].capitalize()}: {turn['content']}")
        parts.append("")
    return "\n".join(parts)


def generate_timestamps(n: int, start_date: str = "2023/01/01", span_days: int = 180) -> list[str]:
    base = datetime.strptime(start_date, "%Y/%m/%d")
    dates = sorted([
        base + timedelta(days=random.randint(0, span_days), hours=random.randint(8, 22), minutes=random.randint(0, 59))
        for _ in range(n)
    ])
    return [d.strftime("%Y/%m/%d (%a) %H:%M") for d in dates]


def build_memory_bank(sessions: list[dict]) -> list[dict]:
    return [
        {
            "session_id": sess["session_id"],
            "timestamp": sess.get("timestamp", ""),
            "turns": sess["turns"],
        }
        for sess in sessions
    ]