mirror of
https://github.com/farion1231/cc-switch.git
synced 2026-06-03 21:32:45 +08:00
2ee7cb4101
Proxy writes and session-log sync wrote to proxy_request_logs with
mismatched request_ids: only Claude on a native Anthropic backend used the
shared `session:{message_id}` key. Codex/Gemini and Claude-through-OpenAI
providers always produced distinct ids, so primary-key dedup never fired
and every real request was recorded twice.
Adds a 7-dim fingerprint dedup (app_type, 4 token counts, 2xx status,
model with case-insensitive match, ±10min window) wired into three layers:
- Write path: should_skip_session_insert() blocks duplicate session rows
before INSERT, unifying the previously-divergent Claude/Codex/Gemini
paths through a single DedupKey-based helper.
- Read path: effective_usage_log_filter() excludes already-covered session
rows from every aggregation query.
- Rollup path: same filter applied so usage_daily_rollups never absorbs
duplicates.
Also adds a covering index (idx_request_logs_dedup_lookup) so the EXISTS
subquery stays index-only, and a transform.rs regression test that pins
openai_to_anthropic id preservation - the missing piece that lets
Claude+OpenAI-compatible providers reuse the session: id scheme.
868 lines
28 KiB
Rust
868 lines
28 KiB
Rust
//! Codex 会话日志使用追踪
|
||
//!
|
||
//! 从 ~/.codex/sessions/ 下的 JSONL 会话文件中提取精确 token 使用数据，
|
||
//! 替代原有的 state_5.sqlite 估算方案。
|
||
//!
|
||
//! ## 数据流
|
||
//! ```text
|
||
//! ~/.codex/sessions/YYYY/MM/DD/*.jsonl → 增量解析 → delta 计算 → 费用计算 → proxy_request_logs 表
|
||
//! ```
|
||
//!
|
||
//! ## 解析的事件类型
|
||
//! - `session_meta` → 提取 session_id
|
||
//! - `turn_context` → 提取当前 model
|
||
//! - `event_msg` (type=token_count) → 提取累计 token 用量，计算 delta
|
||
|
||
use crate::codex_config::get_codex_config_dir;
|
||
use crate::database::{lock_conn, Database};
|
||
use crate::error::AppError;
|
||
use crate::proxy::usage::calculator::{CostCalculator, ModelPricing};
|
||
use crate::proxy::usage::parser::TokenUsage;
|
||
use crate::services::session_usage::SessionSyncResult;
|
||
use crate::services::usage_stats::{should_skip_session_insert, DedupKey};
|
||
use rust_decimal::Decimal;
|
||
use std::fs;
|
||
use std::io::{BufRead, BufReader};
|
||
use std::path::{Path, PathBuf};
|
||
use std::time::SystemTime;
|
||
|
||
/// 累计 token 用量（跟踪 total_token_usage 字段）
|
||
#[derive(Debug, Clone, Default)]
|
||
struct CumulativeTokens {
|
||
input: u64,
|
||
cached_input: u64,
|
||
output: u64,
|
||
}
|
||
|
||
/// 单次 API 调用的 token 增量
|
||
#[derive(Debug)]
|
||
struct DeltaTokens {
|
||
input: u32,
|
||
cached_input: u32,
|
||
output: u32,
|
||
}
|
||
|
||
impl DeltaTokens {
|
||
fn is_zero(&self) -> bool {
|
||
self.input == 0 && self.cached_input == 0 && self.output == 0
|
||
}
|
||
}
|
||
|
||
/// 单文件解析时的运行状态
|
||
struct FileParseState {
|
||
session_id: Option<String>,
|
||
current_model: String,
|
||
prev_total: Option<CumulativeTokens>,
|
||
event_index: u32,
|
||
}
|
||
|
||
/// 归一化 Codex 模型名
|
||
///
|
||
/// 处理规则（按顺序）：
|
||
/// 1. 转小写：`GLM-4.6` → `glm-4.6`
|
||
/// 2. 剥离 provider 前缀：`openai/gpt-5.4` → `gpt-5.4`
|
||
/// 3. 剥离 ISO 日期后缀：`gpt-5.4-2026-03-05` → `gpt-5.4`
|
||
/// 4. 剥离紧凑日期后缀：`gpt-5.4-20260305` → `gpt-5.4`
|
||
fn normalize_codex_model(raw: &str) -> String {
|
||
// Step 1: 小写
|
||
let mut name = raw.to_lowercase();
|
||
|
||
// Step 2: 剥离 "provider/" 前缀（如 openai/, azure/）
|
||
if let Some(pos) = name.rfind('/') {
|
||
name = name[pos + 1..].to_string();
|
||
}
|
||
|
||
// Step 3: 剥离 ISO 日期后缀 -YYYY-MM-DD（正好 11 字符）
|
||
if name.len() > 11 {
|
||
let suffix = &name[name.len() - 11..];
|
||
if suffix.as_bytes()[0] == b'-'
|
||
&& suffix[1..5].chars().all(|c| c.is_ascii_digit())
|
||
&& suffix.as_bytes()[5] == b'-'
|
||
&& suffix[6..8].chars().all(|c| c.is_ascii_digit())
|
||
&& suffix.as_bytes()[8] == b'-'
|
||
&& suffix[9..11].chars().all(|c| c.is_ascii_digit())
|
||
{
|
||
name.truncate(name.len() - 11);
|
||
}
|
||
}
|
||
|
||
// Step 4: 剥离紧凑日期后缀 -YYYYMMDD（正好 9 字符）
|
||
if name.len() > 9 {
|
||
let parts: Vec<&str> = name.rsplitn(2, '-').collect();
|
||
if parts.len() == 2 {
|
||
if let Some(suffix) = parts.first() {
|
||
if suffix.len() == 8 && suffix.chars().all(|c| c.is_ascii_digit()) {
|
||
name = parts[1].to_string();
|
||
}
|
||
}
|
||
}
|
||
}
|
||
|
||
name
|
||
}
|
||
|
||
/// 计算两次累计值之间的 delta
|
||
fn compute_delta(prev: &Option<CumulativeTokens>, current: &CumulativeTokens) -> DeltaTokens {
|
||
match prev {
|
||
None => DeltaTokens {
|
||
input: current.input as u32,
|
||
cached_input: current.cached_input as u32,
|
||
output: current.output as u32,
|
||
},
|
||
Some(p) => DeltaTokens {
|
||
input: current.input.saturating_sub(p.input) as u32,
|
||
cached_input: current.cached_input.saturating_sub(p.cached_input) as u32,
|
||
output: current.output.saturating_sub(p.output) as u32,
|
||
},
|
||
}
|
||
}
|
||
|
||
/// 从 JSON Value 中提取累计 token 用量
|
||
fn parse_cumulative_tokens(total_usage: &serde_json::Value) -> Option<CumulativeTokens> {
|
||
if total_usage.is_null() || !total_usage.is_object() {
|
||
return None;
|
||
}
|
||
Some(CumulativeTokens {
|
||
input: total_usage
|
||
.get("input_tokens")
|
||
.and_then(|v| v.as_u64())
|
||
.unwrap_or(0),
|
||
cached_input: total_usage
|
||
.get("cached_input_tokens")
|
||
.or_else(|| total_usage.get("cache_read_input_tokens"))
|
||
.and_then(|v| v.as_u64())
|
||
.unwrap_or(0),
|
||
output: total_usage
|
||
.get("output_tokens")
|
||
.and_then(|v| v.as_u64())
|
||
.unwrap_or(0),
|
||
})
|
||
}
|
||
|
||
/// 同步 Codex 使用数据（从 JSONL 会话日志）
|
||
pub fn sync_codex_usage(db: &Database) -> Result<SessionSyncResult, AppError> {
|
||
let codex_dir = get_codex_config_dir();
|
||
|
||
let files = collect_codex_session_files(&codex_dir);
|
||
|
||
let mut result = SessionSyncResult {
|
||
imported: 0,
|
||
skipped: 0,
|
||
files_scanned: files.len() as u32,
|
||
errors: vec![],
|
||
};
|
||
|
||
if files.is_empty() {
|
||
return Ok(result);
|
||
}
|
||
|
||
for file_path in &files {
|
||
match sync_single_codex_file(db, file_path) {
|
||
Ok((imported, skipped)) => {
|
||
result.imported += imported;
|
||
result.skipped += skipped;
|
||
}
|
||
Err(e) => {
|
||
let msg = format!("Codex 会话文件解析失败 {}: {e}", file_path.display());
|
||
log::warn!("[CODEX-SYNC] {msg}");
|
||
result.errors.push(msg);
|
||
}
|
||
}
|
||
}
|
||
|
||
if result.imported > 0 {
|
||
log::info!(
|
||
"[CODEX-SYNC] 同步完成: 导入 {} 条, 跳过 {} 条, <20><>描 {} 个文件",
|
||
result.imported,
|
||
result.skipped,
|
||
result.files_scanned
|
||
);
|
||
}
|
||
|
||
Ok(result)
|
||
}
|
||
|
||
/// 收集所有 Codex 会话 JSONL 文件
|
||
fn collect_codex_session_files(codex_dir: &Path) -> Vec<PathBuf> {
|
||
let mut files = Vec::new();
|
||
|
||
// 1. 扫描 sessions/YYYY/MM/DD/*.jsonl（日期分区目录<E79BAE><E5BD95>
|
||
let sessions_dir = codex_dir.join("sessions");
|
||
if sessions_dir.is_dir() {
|
||
collect_jsonl_recursive(&sessions_dir, &mut files, 0, 3);
|
||
}
|
||
|
||
// 2. 扫描 archived_sessions/*.jsonl（扁平归档目录）
|
||
let archived_dir = codex_dir.join("archived_sessions");
|
||
if archived_dir.is_dir() {
|
||
if let Ok(entries) = fs::read_dir(&archived_dir) {
|
||
for entry in entries.flatten() {
|
||
let path = entry.path();
|
||
if path.extension().and_then(|e| e.to_str()) == Some("jsonl") {
|
||
files.push(path);
|
||
}
|
||
}
|
||
}
|
||
}
|
||
|
||
files
|
||
}
|
||
|
||
/// 递归扫描目录下的 .jsonl 文件（限制最大深度）
|
||
fn collect_jsonl_recursive(dir: &Path, files: &mut Vec<PathBuf>, depth: u32, max_depth: u32) {
|
||
let entries = match fs::read_dir(dir) {
|
||
Ok(e) => e,
|
||
Err(_) => return,
|
||
};
|
||
|
||
for entry in entries.flatten() {
|
||
let path = entry.path();
|
||
if path.is_dir() && depth < max_depth {
|
||
collect_jsonl_recursive(&path, files, depth + 1, max_depth);
|
||
} else if path.extension().and_then(|e| e.to_str()) == Some("jsonl") {
|
||
files.push(path);
|
||
}
|
||
}
|
||
}
|
||
|
||
/// 同步单<E6ADA5><E58D95> Codex JSONL 文件，返回 (imported, skipped)
|
||
fn sync_single_codex_file(db: &Database, file_path: &Path) -> Result<(u32, u32), AppError> {
|
||
let file_path_str = file_path.to_string_lossy().to_string();
|
||
|
||
// 获取文件元数据
|
||
let metadata = fs::metadata(file_path)
|
||
.map_err(|e| AppError::Config(format!("无法读取文<EFBFBD><EFBFBD><EFBFBD>元数据: {e}")))?;
|
||
let file_modified = metadata
|
||
.modified()
|
||
.ok()
|
||
.and_then(|t| t.duration_since(SystemTime::UNIX_EPOCH).ok())
|
||
.map(|d| d.as_secs() as i64)
|
||
.unwrap_or(0);
|
||
|
||
// 检查同步状态
|
||
let (last_modified, last_offset) = get_sync_state(db, &file_path_str)?;
|
||
|
||
// 文件未变化则跳过
|
||
if file_modified <= last_modified {
|
||
return Ok((0, 0));
|
||
}
|
||
|
||
// 打开文件逐行解析
|
||
let file =
|
||
fs::File::open(file_path).map_err(|e| AppError::Config(format!("无法打开文件: {e}")))?;
|
||
let reader = BufReader::new(file);
|
||
|
||
let mut state = FileParseState {
|
||
session_id: None,
|
||
current_model: "unknown".to_string(),
|
||
prev_total: None,
|
||
event_index: 0,
|
||
};
|
||
|
||
let mut line_offset: i64 = 0;
|
||
let mut imported: u32 = 0;
|
||
let mut skipped: u32 = 0;
|
||
|
||
for line_result in reader.lines() {
|
||
line_offset += 1;
|
||
|
||
let line = match line_result {
|
||
Ok(l) => l,
|
||
Err(_) => continue, // 容忍不完整的最后一行
|
||
};
|
||
|
||
if line.trim().is_empty() {
|
||
continue;
|
||
}
|
||
|
||
// 快速过滤：在 JSON 反序列化前跳过无关行
|
||
let is_event_msg = line.contains("\"event_msg\"");
|
||
let is_turn_context = line.contains("\"turn_context\"");
|
||
let is_session_meta = line.contains("\"session_meta\"");
|
||
|
||
if !is_event_msg && !is_turn_context && !is_session_meta {
|
||
continue;
|
||
}
|
||
if is_event_msg && !line.contains("\"token_count\"") {
|
||
continue;
|
||
}
|
||
|
||
let value: serde_json::Value = match serde_json::from_str(&line) {
|
||
Ok(v) => v,
|
||
Err(_) => continue,
|
||
};
|
||
|
||
let event_type = match value.get("type").and_then(|t| t.as_str()) {
|
||
Some(t) => t,
|
||
None => continue,
|
||
};
|
||
|
||
match event_type {
|
||
"session_meta" if state.session_id.is_none() => {
|
||
let payload = value.get("payload");
|
||
state.session_id = payload
|
||
.and_then(|p| {
|
||
p.get("session_id")
|
||
.or_else(|| p.get("sessionId"))
|
||
.or_else(|| p.get("id"))
|
||
})
|
||
.and_then(|v| v.as_str())
|
||
.map(|s| s.to_string());
|
||
}
|
||
"turn_context" => {
|
||
if let Some(payload) = value.get("payload") {
|
||
// model 可能在 payload.model 或 payload.info.model
|
||
if let Some(model) = payload
|
||
.get("model")
|
||
.or_else(|| payload.get("info").and_then(|info| info.get("model")))
|
||
.and_then(|v| v.as_str())
|
||
{
|
||
state.current_model = normalize_codex_model(model);
|
||
}
|
||
}
|
||
}
|
||
"event_msg" => {
|
||
let payload = match value.get("payload") {
|
||
Some(p) => p,
|
||
None => continue,
|
||
};
|
||
|
||
// 只处理 token_count 类型
|
||
if payload.get("type").and_then(|t| t.as_str()) != Some("token_count") {
|
||
continue;
|
||
}
|
||
|
||
let info = match payload.get("info") {
|
||
Some(i) if !i.is_null() => i,
|
||
_ => continue, // info 为 null 的首个事件跳<E4BBB6><E8B7B3>
|
||
};
|
||
|
||
// 提取模型（token_count 事件也可能携带 model）
|
||
if let Some(model) = info
|
||
.get("model")
|
||
.or_else(|| info.get("model_name"))
|
||
.or_else(|| payload.get("model"))
|
||
.and_then(|v| v.as_str())
|
||
{
|
||
state.current_model = normalize_codex_model(model);
|
||
}
|
||
|
||
// 优先用 total_token_usage（累计值），fallback 到 last_token_usage（增量值）
|
||
let (cumulative, is_total) = if let Some(total) = info.get("total_token_usage") {
|
||
(parse_cumulative_tokens(total), true)
|
||
} else if let Some(last) = info.get("last_token_usage") {
|
||
(parse_cumulative_tokens(last), false)
|
||
} else {
|
||
continue;
|
||
};
|
||
|
||
let cumulative = match cumulative {
|
||
Some(c) => c,
|
||
None => continue,
|
||
};
|
||
|
||
let delta = if is_total {
|
||
// 累计值模式：计算与上次的 delta
|
||
let d = compute_delta(&state.prev_total, &cumulative);
|
||
state.prev_total = Some(cumulative);
|
||
d
|
||
} else {
|
||
// 增量值模式：直接使用 last_token_usage 的值
|
||
DeltaTokens {
|
||
input: cumulative.input as u32,
|
||
cached_input: cumulative.cached_input as u32,
|
||
output: cumulative.output as u32,
|
||
}
|
||
};
|
||
|
||
// 钳制：cached 不应超过 input（防护异常数据）
|
||
let delta = DeltaTokens {
|
||
cached_input: delta.cached_input.min(delta.input),
|
||
..delta
|
||
};
|
||
|
||
if delta.is_zero() {
|
||
continue; // 跳过 task 边界的零 delta 事件
|
||
}
|
||
|
||
state.event_index += 1;
|
||
|
||
// 跳过已处理的行（但仍需解析以恢复状态）
|
||
if line_offset <= last_offset {
|
||
continue;
|
||
}
|
||
|
||
// 生成唯一 request_id
|
||
let session_id_str = state.session_id.as_deref().unwrap_or("unknown");
|
||
let request_id = format!("codex_session:{}:{}", session_id_str, state.event_index);
|
||
|
||
// 提取时间戳
|
||
let timestamp = value
|
||
.get("timestamp")
|
||
.and_then(|v| v.as_str())
|
||
.map(|s| s.to_string());
|
||
|
||
match insert_codex_session_entry(
|
||
db,
|
||
&request_id,
|
||
&delta,
|
||
&state.current_model,
|
||
state.session_id.as_deref(),
|
||
timestamp.as_deref(),
|
||
) {
|
||
Ok(true) => imported += 1,
|
||
Ok(false) => skipped += 1,
|
||
Err(e) => {
|
||
log::warn!("[CODEX-SYNC] 插入失败 ({}): {e}", request_id);
|
||
skipped += 1;
|
||
}
|
||
}
|
||
}
|
||
_ => {}
|
||
}
|
||
}
|
||
|
||
// 更新同步状态
|
||
update_sync_state(db, &file_path_str, file_modified, line_offset)?;
|
||
|
||
Ok((imported, skipped))
|
||
}
|
||
|
||
/// 插入单条 Codex 会话记录到 proxy_request_logs
|
||
fn insert_codex_session_entry(
|
||
db: &Database,
|
||
request_id: &str,
|
||
delta: &DeltaTokens,
|
||
model: &str,
|
||
session_id: Option<&str>,
|
||
timestamp: Option<&str>,
|
||
) -> Result<bool, AppError> {
|
||
let conn = lock_conn!(db.conn);
|
||
|
||
let created_at = timestamp
|
||
.and_then(|ts| {
|
||
chrono::DateTime::parse_from_rfc3339(ts)
|
||
.ok()
|
||
.map(|dt| dt.timestamp())
|
||
})
|
||
.unwrap_or_else(|| {
|
||
SystemTime::now()
|
||
.duration_since(SystemTime::UNIX_EPOCH)
|
||
.map(|d| d.as_secs() as i64)
|
||
.unwrap_or(0)
|
||
});
|
||
|
||
let dedup_key = DedupKey {
|
||
app_type: "codex",
|
||
model,
|
||
input_tokens: delta.input,
|
||
output_tokens: delta.output,
|
||
cache_read_tokens: delta.cached_input,
|
||
cache_creation_tokens: 0,
|
||
created_at,
|
||
};
|
||
if should_skip_session_insert(&conn, request_id, &dedup_key)? {
|
||
return Ok(false);
|
||
}
|
||
|
||
// 计算费用
|
||
let usage = TokenUsage {
|
||
input_tokens: delta.input,
|
||
output_tokens: delta.output,
|
||
cache_read_tokens: delta.cached_input,
|
||
cache_creation_tokens: 0,
|
||
model: Some(model.to_string()),
|
||
message_id: None,
|
||
};
|
||
|
||
let pricing = find_codex_pricing(&conn, model);
|
||
let multiplier = Decimal::from(1);
|
||
let (input_cost, output_cost, cache_read_cost, cache_creation_cost, total_cost) = match pricing
|
||
{
|
||
Some(p) => {
|
||
let cost = CostCalculator::calculate(&usage, &p, multiplier);
|
||
(
|
||
cost.input_cost.to_string(),
|
||
cost.output_cost.to_string(),
|
||
cost.cache_read_cost.to_string(),
|
||
cost.cache_creation_cost.to_string(),
|
||
cost.total_cost.to_string(),
|
||
)
|
||
}
|
||
None => (
|
||
"0".to_string(),
|
||
"0".to_string(),
|
||
"0".to_string(),
|
||
"0".to_string(),
|
||
"0".to_string(),
|
||
),
|
||
};
|
||
|
||
conn.execute(
|
||
"INSERT OR IGNORE INTO proxy_request_logs (
|
||
request_id, provider_id, app_type, model, request_model,
|
||
input_tokens, output_tokens, cache_read_tokens, cache_creation_tokens,
|
||
input_cost_usd, output_cost_usd, cache_read_cost_usd, cache_creation_cost_usd, total_cost_usd,
|
||
latency_ms, first_token_ms, status_code, error_message, session_id,
|
||
provider_type, is_streaming, cost_multiplier, created_at, data_source
|
||
) VALUES (?1, ?2, ?3, ?4, ?5, ?6, ?7, ?8, ?9, ?10, ?11, ?12, ?13, ?14, ?15, ?16, ?17, ?18, ?19, ?20, ?21, ?22, ?23, ?24)",
|
||
rusqlite::params![
|
||
request_id,
|
||
"_codex_session", // provider_id
|
||
"codex", // app_type
|
||
model,
|
||
model, // request_model = model
|
||
delta.input,
|
||
delta.output,
|
||
delta.cached_input,
|
||
0i64, // cache_creation_tokens: Codex 日志无此数据
|
||
input_cost,
|
||
output_cost,
|
||
cache_read_cost,
|
||
cache_creation_cost,
|
||
total_cost,
|
||
0i64, // latency_ms
|
||
Option::<i64>::None, // first_token_ms
|
||
200i64, // status_code
|
||
Option::<String>::None, // error_message
|
||
session_id.map(|s| s.to_string()),
|
||
Some("codex_session"), // provider_type
|
||
1i64, // is_streaming
|
||
"1.0", // cost_multiplier
|
||
created_at,
|
||
"codex_session", // data_source
|
||
],
|
||
)
|
||
.map_err(|e| AppError::Database(format!("插入 Codex 会话日志失败: {e}")))?;
|
||
|
||
Ok(true)
|
||
}
|
||
|
||
/// 获取文件的同步状态
|
||
fn get_sync_state(db: &Database, file_path: &str) -> Result<(i64, i64), AppError> {
|
||
let conn = lock_conn!(db.conn);
|
||
let result = conn.query_row(
|
||
"SELECT last_modified, last_line_offset FROM session_log_sync WHERE file_path = ?1",
|
||
rusqlite::params![file_path],
|
||
|row| Ok((row.get::<_, i64>(0)?, row.get::<_, i64>(1)?)),
|
||
);
|
||
Ok(result.unwrap_or((0, 0)))
|
||
}
|
||
|
||
/// 更新文件的同步状态
|
||
fn update_sync_state(
|
||
db: &Database,
|
||
file_path: &str,
|
||
last_modified: i64,
|
||
last_offset: i64,
|
||
) -> Result<(), AppError> {
|
||
let now = SystemTime::now()
|
||
.duration_since(SystemTime::UNIX_EPOCH)
|
||
.map(|d| d.as_secs() as i64)
|
||
.unwrap_or(0);
|
||
|
||
let conn = lock_conn!(db.conn);
|
||
conn.execute(
|
||
"INSERT OR REPLACE INTO session_log_sync (file_path, last_modified, last_line_offset, last_synced_at)
|
||
VALUES (?1, ?2, ?3, ?4)",
|
||
rusqlite::params![file_path, last_modified, last_offset, now],
|
||
)
|
||
.map_err(|e| AppError::Database(format!("更新同步状态失败: {e}")))?;
|
||
Ok(())
|
||
}
|
||
|
||
/// <20><>找 Codex 模型定价（带归一化）
|
||
fn find_codex_pricing(conn: &rusqlite::Connection, model_id: &str) -> Option<ModelPricing> {
|
||
let normalized = normalize_codex_model(model_id);
|
||
|
||
// 1. 精确匹配（归一化后的名称）
|
||
if let Some(pricing) = try_find_pricing(conn, &normalized) {
|
||
return Some(pricing);
|
||
}
|
||
|
||
// 2. LIKE 模糊匹配（兜底）
|
||
let pattern = format!("{normalized}%");
|
||
conn.query_row(
|
||
"SELECT input_cost_per_million, output_cost_per_million,
|
||
cache_read_cost_per_million, cache_creation_cost_per_million
|
||
FROM model_pricing WHERE model_id LIKE ?1 LIMIT 1",
|
||
rusqlite::params![pattern],
|
||
|row| {
|
||
Ok((
|
||
row.get::<_, String>(0)?,
|
||
row.get::<_, String>(1)?,
|
||
row.get::<_, String>(2)?,
|
||
row.get::<_, String>(3)?,
|
||
))
|
||
},
|
||
)
|
||
.ok()
|
||
.and_then(|(i, o, cr, cc)| ModelPricing::from_strings(&i, &o, &cr, &cc).ok())
|
||
}
|
||
|
||
/// 精确匹配定价查询
|
||
fn try_find_pricing(conn: &rusqlite::Connection, model_id: &str) -> Option<ModelPricing> {
|
||
conn.query_row(
|
||
"SELECT input_cost_per_million, output_cost_per_million,
|
||
cache_read_cost_per_million, cache_creation_cost_per_million
|
||
FROM model_pricing WHERE model_id = ?1",
|
||
rusqlite::params![model_id],
|
||
|row| {
|
||
Ok((
|
||
row.get::<_, String>(0)?,
|
||
row.get::<_, String>(1)?,
|
||
row.get::<_, String>(2)?,
|
||
row.get::<_, String>(3)?,
|
||
))
|
||
},
|
||
)
|
||
.ok()
|
||
.and_then(|(i, o, cr, cc)| ModelPricing::from_strings(&i, &o, &cr, &cc).ok())
|
||
}
|
||
|
||
#[cfg(test)]
|
||
mod tests {
|
||
use super::*;
|
||
|
||
#[test]
|
||
fn test_delta_first_event() {
|
||
let prev = None;
|
||
let current = CumulativeTokens {
|
||
input: 17934,
|
||
cached_input: 9600,
|
||
output: 454,
|
||
};
|
||
let delta = compute_delta(&prev, ¤t);
|
||
assert_eq!(delta.input, 17934);
|
||
assert_eq!(delta.cached_input, 9600);
|
||
assert_eq!(delta.output, 454);
|
||
assert!(!delta.is_zero());
|
||
}
|
||
|
||
#[test]
|
||
fn test_delta_subsequent_event() {
|
||
let prev = Some(CumulativeTokens {
|
||
input: 17934,
|
||
cached_input: 9600,
|
||
output: 454,
|
||
});
|
||
let current = CumulativeTokens {
|
||
input: 36722,
|
||
cached_input: 27904,
|
||
output: 804,
|
||
};
|
||
let delta = compute_delta(&prev, ¤t);
|
||
assert_eq!(delta.input, 36722 - 17934);
|
||
assert_eq!(delta.cached_input, 27904 - 9600);
|
||
assert_eq!(delta.output, 804 - 454);
|
||
}
|
||
|
||
#[test]
|
||
fn test_delta_zero_at_task_boundary() {
|
||
let prev = Some(CumulativeTokens {
|
||
input: 58346,
|
||
cached_input: 46976,
|
||
output: 1045,
|
||
});
|
||
// task 边界：相同的累计值
|
||
let current = CumulativeTokens {
|
||
input: 58346,
|
||
cached_input: 46976,
|
||
output: 1045,
|
||
};
|
||
let delta = compute_delta(&prev, ¤t);
|
||
assert!(delta.is_zero());
|
||
}
|
||
|
||
#[test]
|
||
fn test_delta_saturating_sub() {
|
||
// 异常情况：当前值小于前值（不应发生，但需防护）
|
||
let prev = Some(CumulativeTokens {
|
||
input: 100,
|
||
cached_input: 50,
|
||
output: 30,
|
||
});
|
||
let current = CumulativeTokens {
|
||
input: 80,
|
||
cached_input: 40,
|
||
output: 20,
|
||
};
|
||
let delta = compute_delta(&prev, ¤t);
|
||
assert_eq!(delta.input, 0);
|
||
assert_eq!(delta.cached_input, 0);
|
||
assert_eq!(delta.output, 0);
|
||
assert!(delta.is_zero());
|
||
}
|
||
|
||
#[test]
|
||
fn test_parse_cumulative_tokens_valid() {
|
||
let json: serde_json::Value = serde_json::json!({
|
||
"input_tokens": 17934,
|
||
"cached_input_tokens": 9600,
|
||
"output_tokens": 454,
|
||
"reasoning_output_tokens": 233,
|
||
"total_tokens": 18388
|
||
});
|
||
let tokens = parse_cumulative_tokens(&json).unwrap();
|
||
assert_eq!(tokens.input, 17934);
|
||
assert_eq!(tokens.cached_input, 9600);
|
||
assert_eq!(tokens.output, 454);
|
||
}
|
||
|
||
#[test]
|
||
fn test_parse_cumulative_tokens_null() {
|
||
let json = serde_json::Value::Null;
|
||
assert!(parse_cumulative_tokens(&json).is_none());
|
||
}
|
||
|
||
#[test]
|
||
fn test_parse_cumulative_tokens_alt_field_names() {
|
||
// 某些版本可能使用 cache_read_input_tokens 而非 cached_input_tokens
|
||
let json: serde_json::Value = serde_json::json!({
|
||
"input_tokens": 1000,
|
||
"cache_read_input_tokens": 500,
|
||
"output_tokens": 200
|
||
});
|
||
let tokens = parse_cumulative_tokens(&json).unwrap();
|
||
assert_eq!(tokens.cached_input, 500);
|
||
}
|
||
|
||
#[test]
|
||
fn test_collect_codex_session_files_nonexistent() {
|
||
let files = collect_codex_session_files(Path::new("/nonexistent/path"));
|
||
assert!(files.is_empty());
|
||
}
|
||
|
||
#[test]
|
||
fn test_insert_codex_session_skips_matching_proxy_log() -> Result<(), AppError> {
|
||
let db = Database::memory()?;
|
||
{
|
||
let conn = lock_conn!(db.conn);
|
||
conn.execute(
|
||
"INSERT INTO proxy_request_logs (
|
||
request_id, provider_id, app_type, model, request_model,
|
||
input_tokens, output_tokens, cache_read_tokens, cache_creation_tokens,
|
||
total_cost_usd, latency_ms, status_code, created_at, data_source
|
||
) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)",
|
||
rusqlite::params![
|
||
"codex-proxy",
|
||
"openai",
|
||
"codex",
|
||
"gpt-5.4",
|
||
"gpt-5.4",
|
||
10,
|
||
2,
|
||
1,
|
||
7,
|
||
"0.01",
|
||
100,
|
||
200,
|
||
1000,
|
||
"proxy"
|
||
],
|
||
)?;
|
||
}
|
||
|
||
let delta = DeltaTokens {
|
||
input: 10,
|
||
cached_input: 1,
|
||
output: 2,
|
||
};
|
||
let inserted = insert_codex_session_entry(
|
||
&db,
|
||
"codex-session-dup",
|
||
&delta,
|
||
"gpt-5.4",
|
||
Some("session-1"),
|
||
Some("1970-01-01T00:16:45Z"),
|
||
)?;
|
||
assert!(!inserted);
|
||
|
||
let conn = lock_conn!(db.conn);
|
||
let count: i64 = conn.query_row("SELECT COUNT(*) FROM proxy_request_logs", [], |row| {
|
||
row.get(0)
|
||
})?;
|
||
assert_eq!(count, 1);
|
||
|
||
Ok(())
|
||
}
|
||
|
||
// ── 模型名归一化测试 ──
|
||
|
||
#[test]
|
||
fn test_normalize_codex_model_lowercase() {
|
||
assert_eq!(normalize_codex_model("GLM-4.6"), "glm-4.6");
|
||
assert_eq!(normalize_codex_model("DeepSeek-Chat"), "deepseek-chat");
|
||
assert_eq!(normalize_codex_model("GPT-5.4"), "gpt-5.4");
|
||
}
|
||
|
||
#[test]
|
||
fn test_normalize_codex_model_strip_prefix() {
|
||
assert_eq!(normalize_codex_model("openai/gpt-5.4"), "gpt-5.4");
|
||
assert_eq!(
|
||
normalize_codex_model("azure/gpt-5.2-codex"),
|
||
"gpt-5.2-codex"
|
||
);
|
||
assert_eq!(normalize_codex_model("OPENAI/GPT-5.4"), "gpt-5.4");
|
||
}
|
||
|
||
#[test]
|
||
fn test_normalize_codex_model_strip_iso_date() {
|
||
assert_eq!(normalize_codex_model("gpt-5.4-2026-03-05"), "gpt-5.4");
|
||
assert_eq!(
|
||
normalize_codex_model("gpt-5.4-pro-2026-03-05"),
|
||
"gpt-5.4-pro"
|
||
);
|
||
}
|
||
|
||
#[test]
|
||
fn test_normalize_codex_model_strip_compact_date() {
|
||
assert_eq!(normalize_codex_model("gpt-5.4-20260305"), "gpt-5.4");
|
||
assert_eq!(
|
||
normalize_codex_model("claude-opus-4-6-20260206"),
|
||
"claude-opus-4-6"
|
||
);
|
||
}
|
||
|
||
#[test]
|
||
fn test_normalize_codex_model_no_change() {
|
||
assert_eq!(normalize_codex_model("gpt-5.4"), "gpt-5.4");
|
||
assert_eq!(normalize_codex_model("gpt-5.2-codex"), "gpt-5.2-codex");
|
||
assert_eq!(normalize_codex_model("o3"), "o3");
|
||
assert_eq!(normalize_codex_model("deepseek-chat"), "deepseek-chat");
|
||
}
|
||
|
||
#[test]
|
||
fn test_normalize_codex_model_combined() {
|
||
// prefix + uppercase + ISO date
|
||
assert_eq!(
|
||
normalize_codex_model("openai/GPT-5.4-2026-03-05"),
|
||
"gpt-5.4"
|
||
);
|
||
// prefix + compact date
|
||
assert_eq!(normalize_codex_model("openai/gpt-5.4-20260305"), "gpt-5.4");
|
||
}
|
||
|
||
#[test]
|
||
fn test_cached_clamped_to_input() {
|
||
// cached > input 的异常场景应被 min() 钳制
|
||
let prev = Some(CumulativeTokens {
|
||
input: 100,
|
||
cached_input: 0,
|
||
output: 50,
|
||
});
|
||
let current = CumulativeTokens {
|
||
input: 110, // delta = 10
|
||
cached_input: 80, // delta = 80（异常：大于 input delta）
|
||
output: 60,
|
||
};
|
||
let delta = compute_delta(&prev, ¤t);
|
||
// 钳制前：cached_input = 80, input = 10
|
||
assert_eq!(delta.cached_input, 80);
|
||
assert_eq!(delta.input, 10);
|
||
// 实际钳制在调用侧：delta.cached_input.min(delta.input)
|
||
let clamped = delta.cached_input.min(delta.input);
|
||
assert_eq!(clamped, 10);
|
||
}
|
||
}
|