codex-app-server/cloud/cost.py

"""Cost estimation and cost recording.

Zero third-party dependencies; standard library only.

Actual token counts come from the model/provider response when
available; when they are not, token counts are estimates (see
``estimate_tokens``) and are stored separately as ``source="estimated"``
so estimates are never mixed into actual totals.
"""

import os
import sqlite3
from cloud.db import connect as _db_connect
import time
from datetime import datetime, timezone


def estimate_tokens(prompt_text, output_text) -> int:
    """Estimate total tokens as (chars of prompt + output) // 4."""
    prompt_text = prompt_text or ""
    output_text = output_text or ""
    return (len(prompt_text) + len(output_text)) // 4


def fetch_account_usage(usage_client) -> dict:
    """Fetch account usage through an injected ``usage_client`` callable.

    This is a call wrapper around the app-server account/usage/read
    concept: the real ``usage_client`` is injected by the control plane
    once the billing route is confirmed available. This function itself
    never runs a model turn and never touches the network.

    On success it returns ``{"available": True, "data": <raw data>}``.
    If ``usage_client`` raises, or returns None / an empty dict,
    it returns ``{"available": False, "data": None}`` and never raises
    to the caller. On a relay, usage being unreadable is a normal
    outcome (available=False), not a bug.
    """
    try:
        data = usage_client()
    except Exception:
        return {"available": False, "data": None}
    if data is None:
        return {"available": False, "data": None}
    if isinstance(data, dict) and not data:
        return {"available": False, "data": None}
    return {"available": True, "data": data}


class CostStore:
    """SQLite store for per-task / per-project / per-day cost records."""

    _SOURCES = ("actual", "estimated")

    def __init__(self, db_path):
        self.db_path = db_path
        if db_path != ":memory:":
            parent = os.path.dirname(os.path.abspath(db_path))
            os.makedirs(parent, exist_ok=True)
        self._conn = _db_connect(db_path)
        self._conn.execute(
            """CREATE TABLE IF NOT EXISTS cost_records (
                id INTEGER PRIMARY KEY AUTOINCREMENT,
                task_id TEXT,
                project_id TEXT,
                model TEXT,
                input_tokens INTEGER,
                output_tokens INTEGER,
                total_tokens INTEGER,
                usd_micros INTEGER,
                source TEXT,
                fallback_count INTEGER DEFAULT 0,
                day TEXT,
                created_at REAL
            )"""
        )
        self._conn.commit()

    def _new_conn(self):
        """每次写/读开一个新连接，用完即关。

        旧实现在 __init__ 里建一条连接并终身共用，单线程没问题；
        worker 并发后多个线程共用同一条 sqlite 连接会直接报
        ProgrammingError（连接是线程绑定的）。其它 store（Worker
        的 _set_status、_audit 等）本来就是每次开新连接，这里
        向它们看齐。_conn 仍保留（建表 + 旧测试直接用它读），
        但 record/_report 不再走它。
        例外：db_path 为 :memory: 时新连接是另一个空库，必须
        复用 _conn（:memory: 只用于单线程单测，并发 worker
        从不用它）。
        """
        if self.db_path == ":memory:":
            return self._conn
        conn = _db_connect(self.db_path)
        conn.execute(
            """CREATE TABLE IF NOT EXISTS cost_records (
                id INTEGER PRIMARY KEY AUTOINCREMENT,
                task_id TEXT,
                project_id TEXT,
                model TEXT,
                input_tokens INTEGER,
                output_tokens INTEGER,
                total_tokens INTEGER,
                usd_micros INTEGER,
                source TEXT,
                fallback_count INTEGER DEFAULT 0,
                day TEXT,
                created_at REAL
            )"""
        )
        conn.commit()
        return conn

    def record(self, task_id, project_id, model, input_tokens=0,
               output_tokens=0, usd_micros=None, source="actual",
               fallback_count=0):
        """Insert one cost record; returns its row id.

        ``total_tokens`` is computed as input + output, and ``day`` is
        the UTC date (YYYY-MM-DD) of ``created_at``.
        """
        if source not in self._SOURCES:
            raise ValueError(
                "source must be one of %r, got %r" % (self._SOURCES, source)
            )
        input_tokens = int(input_tokens or 0)
        output_tokens = int(output_tokens or 0)
        total_tokens = input_tokens + output_tokens
        created_at = time.time()
        day = datetime.fromtimestamp(created_at, tz=timezone.utc).strftime("%Y-%m-%d")
        conn = self._new_conn()
        try:
            cursor = conn.execute(
                """INSERT INTO cost_records
                   (task_id, project_id, model, input_tokens, output_tokens,
                    total_tokens, usd_micros, source, fallback_count, day,
                    created_at)
                   VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)""",
                (task_id, project_id, model, input_tokens, output_tokens,
                 total_tokens, usd_micros, source, int(fallback_count or 0),
                 day, created_at),
            )
            conn.commit()
            return cursor.lastrowid
        finally:
            # :memory: 时 conn 就是共享的 _conn，绝不能关掉它。
            if conn is not self._conn:
                conn.close()

    def record_estimated(self, task_id, project_id, model, prompt_text,
                         output_text):
        """Record an estimated cost row; returns its row id.

        The token split cannot be recovered from a character estimate,
        so ``input_tokens`` holds the whole estimate and
        ``output_tokens`` is 0; ``source`` is ``"estimated"`` and
        ``usd_micros`` is None.
        """
        total = estimate_tokens(prompt_text, output_text)
        return self.record(
            task_id, project_id, model,
            input_tokens=total, output_tokens=0,
            usd_micros=None, source="estimated", fallback_count=0,
        )

    def _report(self, where, params):
        conn = self._new_conn()
        try:
            row = conn.execute(
                """SELECT
                     COALESCE(SUM(total_tokens), 0),
                     COALESCE(SUM(usd_micros), 0),
                     COUNT(*),
                     COALESCE(SUM(fallback_count), 0),
                     COALESCE(SUM(CASE WHEN source = 'actual'
                                       THEN total_tokens ELSE 0 END), 0),
                     COALESCE(SUM(CASE WHEN source = 'estimated'
                                       THEN total_tokens ELSE 0 END), 0)
                   FROM cost_records WHERE %s""" % where,
                params,
            ).fetchone()
        finally:
            if conn is not self._conn:
                conn.close()
        total_tokens, usd_micros, turns, fallback_count, actual, estimated = row
        return {
            "total_tokens": int(total_tokens),
            "usd_micros": int(usd_micros),
            "turns": int(turns),
            "fallback_count": int(fallback_count),
            "actual_tokens": int(actual),
            "estimated_tokens": int(estimated),
        }

    def report_by_task(self, task_id) -> dict:
        """Aggregate cost records for one task."""
        return self._report("task_id = ?", (task_id,))

    def report_by_project(self, project_id) -> dict:
        """Aggregate cost records for one project."""
        return self._report("project_id = ?", (project_id,))

    def report_by_day(self, day) -> dict:
        """Aggregate cost records for one UTC day (YYYY-MM-DD)."""
        return self._report("day = ?", (day,))