loreal-utilisation-dept/backend/app/services/ai_context.py

"""Builds the dashboard-context system prompt the AI chat reads from.

Mirrors `src/lib/aiContextBuilder.ts` in the original SPA — kept small
and human-readable so the LLM can cite specific numbers back to the
user.
"""

from __future__ import annotations

from datetime import date, timedelta
from typing import Any, Iterable

from app.services.project_types import build_project_types


def _is_complete(s: str | None) -> bool:
    if not s:
        return False
    n = s.upper()
    return n in {"APPROVED", "COMPLETE", "CANCELLED", "REJECTED", "DECLINED"}


def build_dashboard_context(
    *,
    logs: list[dict[str, Any]],
    project_summary: list[dict[str, Any]] | None = None,
) -> str:
    today = date.today().isoformat()
    lines: list[str] = [f"Today: {today}", ""]

    # Time-log summary.
    people: set[str] = set()
    projects: set[str] = set()
    total_hours = 0.0
    dates: list[str] = []
    for r in logs:
        ident = r.get("submitterEmail") or r.get("submitter") or ""
        if ident:
            people.add(str(ident))
        title = r.get("projectTitle") or r.get("projectNumber") or ""
        if title:
            projects.add(str(title))
        total_hours += float(r.get("hoursLogged") or r.get("hours") or 0)
        d = r.get("date")
        if d:
            dates.append(str(d))
    dates.sort()
    lines.append("## TIME LOG")
    lines.append(f"{len(logs):,} entries | {len(people)} people | {len(projects)} projects | {total_hours:.0f}h total")
    if dates:
        lines.append(f"Date range: {dates[0]} to {dates[-1]}")
    lines.append("")

    # Benchmark by project type (top types).
    types = build_project_types(logs=logs, project_summary=project_summary or None)
    stats = types.get("stats", []) or []
    if stats:
        lines.append(f"## BENCHMARK ({sum(s['projectCount'] for s in stats)} projects across {len(stats)} types)")
        for s in stats[:12]:
            lines.append(
                f"  {s['projectType']}: "
                f"{s['avgHoursPerAsset']:.1f}h/asset | "
                f"{s['projectCount']} projects | "
                f"{s['totalAssets']:.0f} assets | "
                f"avg {s['avgDurationDays']:.0f}d"
            )
        lines.append("")

    # Active pipeline from Project Summary.
    if project_summary:
        active = [ps for ps in project_summary
                  if not _is_complete(ps.get("projectStatus"))
                  and ps.get("projectStartDate")
                  and ps["projectStartDate"] <= today
                  and (not ps.get("projectEndDate") or ps["projectEndDate"] >= today)]
        upcoming = [ps for ps in project_summary
                    if not _is_complete(ps.get("projectStatus"))
                    and (ps.get("projectStartDate") or "") > today]
        # Exiting this week (Mon-Fri).
        d = date.today()
        mon = d - timedelta(days=d.weekday())
        fri = mon + timedelta(days=4)
        exiting = [ps for ps in project_summary
                   if not _is_complete(ps.get("projectStatus"))
                   and ps.get("projectEndDate")
                   and mon.isoformat() <= ps["projectEndDate"] <= fri.isoformat()]
        total_active_assets = sum(float(ps.get("assetCount") or 0) for ps in active)
        lines.append("## ACTIVE PIPELINE")
        lines.append(
            f"Active: {len(active)} projects ({total_active_assets:,.0f} assets) | "
            f"Upcoming: {len(upcoming)} | Exiting this week: {len(exiting)}"
        )
        if exiting:
            lines.append("Exiting this week:")
            for ps in exiting[:20]:
                lines.append(
                    f"  {ps.get('projectNumber','')} | {ps.get('projectTitle','')} | "
                    f"{(ps.get('assetCount') or 0):.0f} assets | ends {ps.get('projectEndDate')}"
                )
        lines.append("")

    return "\n".join(lines)