From 05bc14e82b9a7daa1a8a9abcffdd54a9f139e7f4 Mon Sep 17 00:00:00 2001 From: Jason Date: Tue, 9 Jun 2026 12:04:46 +0800 Subject: [PATCH] fix(usage): import billable session messages without stop_reason The local session-log scanner dropped any assistant message that lacked a stop_reason or had output_tokens==0. Claude Code Workflow / sub-agent fan-out frequently produces messages that only wrote a message_start snapshot (output=1, stop_reason=None) without a final block, yet their input + cache_read + cache_creation tokens are already billed by Anthropic (charged once the request is accepted). Dropping them under-counted usage by ~4.1% overall, 92% concentrated in workflow/subagent transcripts. Replace the stop_reason/output gate with a billable-token check (any of input/output/cache_read/cache_creation > 0). The per-message-id dedup selection is unchanged, and request_id = "session:"+msg_id PRIMARY KEY with INSERT OR IGNORE keeps each message single-inserted, so relaxing the gate cannot double-count. Add a regression test covering a stop_reason-less message with real cache cost plus an all-zero skip. This is the parser-layer half of the Workflow under-counting fixed at the collector layer in 8d332925. --- src-tauri/src/services/session_usage.rs | 67 ++++++++++++++++++++++--- 1 file changed, 59 insertions(+), 8 deletions(-) diff --git a/src-tauri/src/services/session_usage.rs b/src-tauri/src/services/session_usage.rs index 002b34d27..7d8756722 100644 --- a/src-tauri/src/services/session_usage.rs +++ b/src-tauri/src/services/session_usage.rs @@ -314,8 +314,23 @@ fn sync_single_file(db: &Database, file_path: &Path) -> Result<(u32, u32), AppEr let mut skipped: u32 = 0; for msg in messages.values() { - // 只导入有 stop_reason 的最终条目（完整的 API 调用） - if msg.stop_reason.is_none() { + // 只要产生了真实计费 token 就导入，不再强制要求 stop_reason 或 output>0。 + // + // Anthropic 在受理请求时即对 input + cache_read + cache_creation 计费 + // （这些在请求开始就确定），output 按实际生成量计。Workflow / 子 agent 的 + // 并行短命请求经常只写了 message_start 快照（output=1、stop_reason=None） + // 却没有写最终块，但其 cache/input 成本已被真实计费。旧逻辑用 stop_reason + // 非空 + output>0 双重过滤，会把这类请求整条丢弃，实测系统性低估约 4.1%， + // 且 92% 集中在 workflow/subagent。这里改为「任一计费维度 > 0 即导入」。 + // + // 去重选择逻辑（上方按 message.id 取 stop_reason 优先 / output 最大者）保持 + // 不变：它选出的代表行的 input/cache 本就准确；request_id = session:msg_id + // 主键 + INSERT OR IGNORE 保证一个 message 仍只落库一次，放宽 gate 不会双算。 + let has_billable_tokens = msg.input_tokens > 0 + || msg.output_tokens > 0 + || msg.cache_read_tokens > 0 + || msg.cache_creation_tokens > 0; + if !has_billable_tokens { continue; } @@ -325,11 +340,6 @@ fn sync_single_file(db: &Database, file_path: &Path) -> Result<(u32, u32), AppEr msg.message_id ); - // 跳过 output_tokens 为 0 的无意义条目 - if msg.output_tokens == 0 { - continue; - } - match insert_session_log_entry(db, &request_id, msg) { Ok(true) => imported += 1, Ok(false) => skipped += 1, @@ -491,7 +501,7 @@ fn insert_session_log_entry( total_cost, 0i64, // latency_ms: 会话日志无此数据 Option::::None, // first_token_ms - 200i64, // status_code: 有 stop_reason 说明请求成功 + 200i64, // status_code: 会话日志中的请求只要产生计费 token 即视为成功 Option::::None, // error_message msg.session_id, Some("session_log"), // provider_type @@ -744,4 +754,45 @@ mod tests { fs::remove_dir_all(&tmp).ok(); } + + #[test] + fn test_sync_imports_billable_message_without_stop_reason() -> Result<(), AppError> { + // 回归：stop_reason 缺失但有真实 cache/input 成本的 message（Workflow / + // 子 agent 常见的「只有 message_start 快照、没写最终块」形态）必须被计入， + // 不能因缺 stop_reason 或 output==0 而整条丢弃；全 0 token 的占位行仍应跳过。 + let db = Database::memory()?; + let tmp = std::env::temp_dir().join(format!("cc-switch-test-{}", uuid::Uuid::new_v4())); + fs::create_dir_all(&tmp).unwrap(); + let file = tmp.join("agent-wf.jsonl"); + + // 第一行：无 stop_reason、output=1，但 cache_read/cache_creation 很大 → 应导入 + // 第二行：全部 token 为 0 → 应跳过（无计费意义） + let billable = r#"{"type":"assistant","message":{"id":"msg_nostop","model":"claude-opus-4-8","usage":{"input_tokens":2,"output_tokens":1,"cache_read_input_tokens":48719,"cache_creation_input_tokens":2061}},"timestamp":"2026-06-07T13:01:23Z","sessionId":"session-wf"}"#; + let empty = r#"{"type":"assistant","message":{"id":"msg_empty","model":"claude-opus-4-8","usage":{"input_tokens":0,"output_tokens":0,"cache_read_input_tokens":0,"cache_creation_input_tokens":0}},"timestamp":"2026-06-07T13:01:24Z","sessionId":"session-wf"}"#; + fs::write(&file, format!("{billable}\n{empty}\n")).unwrap(); + + let (imported, _skipped) = sync_single_file(&db, &file)?; + assert_eq!( + imported, 1, + "有 cache 成本但无 stop_reason 的 message 必须被导入" + ); + + let conn = lock_conn!(db.conn); + let cache_read: i64 = conn.query_row( + "SELECT cache_read_tokens FROM proxy_request_logs WHERE request_id = 'session:msg_nostop'", + [], + |row| row.get(0), + )?; + assert_eq!(cache_read, 48719, "cache_read 必须被完整记录"); + let empty_exists: bool = conn.query_row( + "SELECT EXISTS(SELECT 1 FROM proxy_request_logs WHERE request_id = 'session:msg_empty')", + [], + |row| row.get(0), + )?; + assert!(!empty_exists, "全 0 token 的 message 应被跳过"); + drop(conn); + + fs::remove_dir_all(&tmp).ok(); + Ok(()) + } }