mirror of
https://github.com/farion1231/cc-switch.git
synced 2026-07-27 08:14:33 +08:00
Jason
63aa310576
Copilot routes through OpenAI-compatible endpoints that reject Anthropic's thinking and redacted_thinking blocks. Previously the request would fail upstream, burning one premium interaction, and only then trigger thinking_rectifier to retry. This adds a proactive strip_thinking_blocks pass in the Copilot optimization pipeline (step 3.5, after tool_result merging). Signature fields and top-level thinking are left alone — those are the reactive rectifier's job on the error path. Also fixes a default-value inconsistency where CopilotOptimizerConfig's Default impl used "gpt-4o-mini" while the serde default function returned "gpt-5-mini" (aligned to gpt-5-mini, matching the reference implementation). Aligned with yuegongzi/copilot-api's /v1/messages handler behavior.
1544 lines
58 KiB
Rust
1544 lines
58 KiB
Rust
//! Copilot 请求优化器
|
||
//!
|
||
//! 解决 GitHub Copilot 代理消耗量异常问题（Issue #1813）。
|
||
//!
|
||
//! Copilot 使用 `x-initiator` 请求头区分「用户发起」和「agent 续写」：
|
||
//! - `user`：计为一次 premium interaction（扣额度）
|
||
//! - `agent`：视为上一次交互的延续（不额外扣费）
|
||
//!
|
||
//! 参考实现: https://github.com/caozhiyuan/copilot-api
|
||
|
||
use std::collections::HashSet;
|
||
|
||
use serde_json::Value;
|
||
use sha2::{Digest, Sha256};
|
||
use uuid::Uuid;
|
||
|
||
/// 请求分类结果
|
||
#[derive(Debug, Clone)]
|
||
pub struct CopilotClassification {
|
||
/// "user" 或 "agent" — 映射到 x-initiator 请求头
|
||
pub initiator: &'static str,
|
||
/// 是否为 warmup/探针请求（可降级到小模型）
|
||
pub is_warmup: bool,
|
||
/// 是否为上下文压缩请求
|
||
pub is_compact: bool,
|
||
/// 是否为 Claude Code 子代理请求（Agent tool 生成的 subagent）
|
||
/// 子代理请求应设置 x-interaction-type=conversation-subagent，不计 premium interaction
|
||
pub is_subagent: bool,
|
||
}
|
||
|
||
/// 分类 Anthropic 格式的请求体，决定 Copilot 请求头。
|
||
///
|
||
/// 分类算法（只检查最后一条消息，与参考实现 caozhiyuan/copilot-api 对齐）：
|
||
/// 1. 无消息 → "user"（安全默认，首次请求）
|
||
/// 2. 最后消息 role=user：
|
||
/// - content 中存在非 tool_result 类型 block → "user"
|
||
/// - content 全部是 tool_result → "agent"
|
||
/// - 匹配 compact 模式 → "agent"
|
||
/// 3. 最后消息 role 非 user → "user"（安全默认）
|
||
///
|
||
/// Warmup 检测（与参考实现对齐）：
|
||
/// - 请求头中有 `anthropic-beta` + 无 tools + 非 compact → warmup
|
||
///
|
||
/// `compact_detection`：是否启用 compact 检测。为 false 时跳过，
|
||
/// 确保 `CopilotOptimizerConfig.compact_detection` 开关真正生效。
|
||
///
|
||
/// `subagent_detection`：是否启用子代理检测。为 true 时，会扫描首条用户消息
|
||
/// 中的 `__SUBAGENT_MARKER__` 标记，将子代理请求标记为不计费。
|
||
pub fn classify_request(
|
||
body: &Value,
|
||
has_anthropic_beta: bool,
|
||
compact_detection: bool,
|
||
subagent_detection: bool,
|
||
) -> CopilotClassification {
|
||
let is_compact = compact_detection && is_compact_request(body);
|
||
let is_subagent = subagent_detection && detect_subagent(body);
|
||
|
||
let messages = match body.get("messages").and_then(|m| m.as_array()) {
|
||
Some(msgs) if !msgs.is_empty() => msgs,
|
||
_ => {
|
||
return CopilotClassification {
|
||
initiator: "user",
|
||
is_warmup: is_warmup_request(body, has_anthropic_beta, false),
|
||
is_compact: false,
|
||
is_subagent,
|
||
}
|
||
}
|
||
};
|
||
|
||
let last_msg = &messages[messages.len() - 1];
|
||
let role = last_msg.get("role").and_then(|r| r.as_str()).unwrap_or("");
|
||
|
||
// 只有 role=user 的消息需要细分
|
||
if role != "user" {
|
||
return CopilotClassification {
|
||
initiator: if is_subagent { "agent" } else { "user" },
|
||
is_warmup: false,
|
||
is_compact,
|
||
is_subagent,
|
||
};
|
||
}
|
||
|
||
// 判定逻辑（与 copilot-api 的 merge-then-classify 效果对齐）：
|
||
// 只要 content 数组中包含 tool_result → 视为工具续写 → agent
|
||
// 这覆盖了 skill/edit hook/plan follow-up 等常见场景，
|
||
// 它们的 content 通常是 [tool_result, text] 混合形态。
|
||
// copilot-api 通过先 merge（text 吸收进 tool_result）再 classify 实现同等效果；
|
||
// 直接在分类层处理更稳健，不依赖 merge 启用状态和执行顺序。
|
||
let is_user_initiated = match last_msg.get("content") {
|
||
Some(content) if content.is_array() => {
|
||
let blocks = content.as_array().unwrap();
|
||
// 含有 tool_result → 工具续写（agent），否则 → 用户发起（user）
|
||
!blocks
|
||
.iter()
|
||
.any(|block| block.get("type").and_then(|t| t.as_str()) == Some("tool_result"))
|
||
}
|
||
Some(content) if content.is_string() => true,
|
||
_ => false,
|
||
};
|
||
|
||
// 子代理请求始终标记为 agent（即使首条消息包含用户文本）
|
||
let initiator = if is_subagent || !is_user_initiated || is_compact {
|
||
"agent"
|
||
} else {
|
||
"user"
|
||
};
|
||
|
||
CopilotClassification {
|
||
initiator,
|
||
is_warmup: initiator == "user" && is_warmup_request(body, has_anthropic_beta, is_compact),
|
||
is_compact,
|
||
is_subagent,
|
||
}
|
||
}
|
||
|
||
/// 检测是否为 warmup/探针请求（适合降级到小模型）。
|
||
///
|
||
/// 与参考实现对齐，三个条件同时满足：
|
||
/// 1. 请求头有 `anthropic-beta`（Claude Code warmup 探针的标志）
|
||
/// 2. 无 tools 定义
|
||
/// 3. 非 compact 请求
|
||
fn is_warmup_request(body: &Value, has_anthropic_beta: bool, is_compact: bool) -> bool {
|
||
if !has_anthropic_beta || is_compact {
|
||
return false;
|
||
}
|
||
// 无工具定义
|
||
!matches!(body.get("tools"), Some(tools) if tools.is_array() && !tools.as_array().unwrap().is_empty())
|
||
}
|
||
|
||
/// 检测是否为 Claude Code 上下文压缩/compact 请求。
|
||
///
|
||
/// 只匹配 Claude Code **内部生成**的机器特征，不匹配用户可能手动输入的通用短语，
|
||
/// 避免将真实用户请求误标为 agent。
|
||
///
|
||
/// 强特征来源：
|
||
/// 1. system prompt — Claude Code compact 模式会设置专用 system prompt，用户无法手动设置
|
||
/// 2. "CRITICAL: Respond with TEXT ONLY. Do NOT call any tools." — 机器指令
|
||
/// 3. 同时包含 "Pending Tasks:" 和 "Current Work:" — Claude Code compact 的结构标记
|
||
fn is_compact_request(body: &Value) -> bool {
|
||
// 信号 1: system prompt 以 Claude Code compact 专用前缀开头
|
||
// 用户在 Claude Code 中无法直接控制 system prompt，这是最可靠的信号
|
||
let system_text = extract_system_text(body);
|
||
if system_text
|
||
.starts_with("You are a helpful AI assistant tasked with summarizing conversations")
|
||
{
|
||
return true;
|
||
}
|
||
|
||
// 信号 2 & 3: 检查最后一条用户消息中的机器生成特征
|
||
let messages = match body.get("messages").and_then(|m| m.as_array()) {
|
||
Some(msgs) => msgs,
|
||
None => return false,
|
||
};
|
||
|
||
if let Some(last_msg) = messages.last() {
|
||
if last_msg.get("role").and_then(|r| r.as_str()) != Some("user") {
|
||
return false;
|
||
}
|
||
|
||
let text = extract_text_from_message(last_msg);
|
||
|
||
// 信号 2: Claude Code compact 的机器指令（大小写敏感，精确匹配）
|
||
if text.contains("CRITICAL: Respond with TEXT ONLY. Do NOT call any tools.") {
|
||
return true;
|
||
}
|
||
|
||
// 信号 3: Claude Code compact 的结构标记（两个同时出现才算）
|
||
if text.contains("Pending Tasks:") && text.contains("Current Work:") {
|
||
return true;
|
||
}
|
||
}
|
||
|
||
false
|
||
}
|
||
|
||
/// 合并用户消息中的 tool_result 和 text block。
|
||
///
|
||
/// 与参考实现 `mergeToolResultForClaude` 对齐：
|
||
///
|
||
/// **消息内部合并**（核心）：在单条 user 消息内，将 text block 吸收进 tool_result block，
|
||
/// 使整条消息只剩 tool_result 类型 block。这样 Copilot 不会将其视为用户发起的交互。
|
||
///
|
||
/// 场景：Claude Code 在 skill 调用、edit hook、plan 提醒等场景下，会发送混合了
|
||
/// tool_result + text 的用户消息。text block 的存在让 Copilot 将其计为 premium request。
|
||
///
|
||
/// **跨消息合并**（补充）：连续的 tool_result-only 用户消息合并为一条。
|
||
pub fn merge_tool_results(mut body: Value) -> Value {
|
||
let messages = match body.get_mut("messages").and_then(|m| m.as_array_mut()) {
|
||
Some(msgs) if !msgs.is_empty() => msgs,
|
||
_ => return body,
|
||
};
|
||
|
||
// Phase 1: 消息内部合并 — 将 text block 吸收进 tool_result block
|
||
for msg in messages.iter_mut() {
|
||
if msg.get("role").and_then(|r| r.as_str()) != Some("user") {
|
||
continue;
|
||
}
|
||
let content = match msg.get("content").and_then(|c| c.as_array()) {
|
||
Some(blocks) => blocks,
|
||
None => continue,
|
||
};
|
||
|
||
// 分离 tool_result 和 text block
|
||
let mut tool_results: Vec<Value> = Vec::new();
|
||
let mut text_blocks: Vec<Value> = Vec::new();
|
||
let mut valid = true;
|
||
|
||
for block in content {
|
||
match block.get("type").and_then(|t| t.as_str()) {
|
||
Some("tool_result") => tool_results.push(block.clone()),
|
||
Some("text") => text_blocks.push(block.clone()),
|
||
_ => {
|
||
// 存在其他类型 block → 跳过此消息
|
||
valid = false;
|
||
break;
|
||
}
|
||
}
|
||
}
|
||
|
||
// 必须同时有 tool_result 和 text 才需要合并
|
||
if !valid || tool_results.is_empty() || text_blocks.is_empty() {
|
||
continue;
|
||
}
|
||
|
||
// 合并策略（与参考实现对齐）
|
||
let merged = merge_blocks_into_tool_results(tool_results, text_blocks);
|
||
msg["content"] = Value::Array(merged);
|
||
}
|
||
|
||
// Phase 2: 跨消息合并 — 连续的 tool_result-only 用户消息合并
|
||
let messages = body["messages"].as_array().unwrap().clone();
|
||
if messages.len() <= 1 {
|
||
return body;
|
||
}
|
||
|
||
let mut merged_msgs: Vec<Value> = Vec::with_capacity(messages.len());
|
||
let mut i = 0;
|
||
|
||
while i < messages.len() {
|
||
if is_tool_result_only_message(&messages[i]) {
|
||
let mut combined_content: Vec<Value> = Vec::new();
|
||
while i < messages.len() && is_tool_result_only_message(&messages[i]) {
|
||
if let Some(content) = messages[i].get("content").and_then(|c| c.as_array()) {
|
||
combined_content.extend(content.iter().cloned());
|
||
}
|
||
i += 1;
|
||
}
|
||
if !combined_content.is_empty() {
|
||
merged_msgs.push(serde_json::json!({
|
||
"role": "user",
|
||
"content": combined_content
|
||
}));
|
||
}
|
||
} else {
|
||
merged_msgs.push(messages[i].clone());
|
||
i += 1;
|
||
}
|
||
}
|
||
|
||
body["messages"] = Value::Array(merged_msgs);
|
||
body
|
||
}
|
||
|
||
/// 基于最后一条用户消息内容生成确定性 Request ID。
|
||
///
|
||
/// CC Switch 额外策略（参考项目 copilot-api 使用随机 UUID）：
|
||
/// - 哈希输入: sessionId + lastUserContent（排除 tool_result 和 cache_control）
|
||
/// - 相同内容产生相同 ID，可能帮助 Copilot 去重
|
||
/// - 找不到用户内容时退化为随机 UUID
|
||
/// - 使用 UUID v4 格式
|
||
pub fn deterministic_request_id(body: &Value, session_id: &str) -> String {
|
||
let last_user_content = find_last_user_content(body);
|
||
|
||
match last_user_content {
|
||
Some(content) => {
|
||
let mut hasher = Sha256::new();
|
||
hasher.update(session_id.as_bytes());
|
||
hasher.update(content.as_bytes());
|
||
let result = hasher.finalize();
|
||
|
||
let mut bytes = [0u8; 16];
|
||
bytes.copy_from_slice(&result[..16]);
|
||
// UUID v4 版本位和变体位（与参考实现一致）
|
||
bytes[6] = (bytes[6] & 0x0f) | 0x40; // version 4
|
||
bytes[8] = (bytes[8] & 0x3f) | 0x80; // variant 1
|
||
|
||
Uuid::from_bytes(bytes).to_string()
|
||
}
|
||
None => Uuid::new_v4().to_string(),
|
||
}
|
||
}
|
||
|
||
/// 基于 session ID 生成稳定的 Interaction ID。
|
||
///
|
||
/// 与参考实现（copilot-api session.ts）对齐：
|
||
/// - 同一主对话的所有请求共享同一个 interaction ID
|
||
/// - 哈希输入: 仅 session ID（不包含消息内容，与 request ID 不同）
|
||
/// - Copilot 用此 ID 将请求聚合为同一个 "interaction"，影响 premium 计费归属
|
||
/// - 空 session ID 时返回 None（不应注入随机值，避免 interaction 碎片化）
|
||
pub fn deterministic_interaction_id(session_id: &str) -> Option<String> {
|
||
if session_id.is_empty() {
|
||
return None;
|
||
}
|
||
|
||
let mut hasher = Sha256::new();
|
||
hasher.update(b"interaction:");
|
||
hasher.update(session_id.as_bytes());
|
||
let result = hasher.finalize();
|
||
|
||
let mut bytes = [0u8; 16];
|
||
bytes.copy_from_slice(&result[..16]);
|
||
bytes[6] = (bytes[6] & 0x0f) | 0x40; // version 4
|
||
bytes[8] = (bytes[8] & 0x3f) | 0x80; // variant 1
|
||
|
||
Some(Uuid::from_bytes(bytes).to_string())
|
||
}
|
||
|
||
/// 检测请求是否来自 Claude Code 子代理（Agent tool 生成的 subagent）。
|
||
///
|
||
/// Claude Code 的 Agent tool 会在子代理首条用户消息的 `<system-reminder>` 标签中
|
||
/// 注入 `__SUBAGENT_MARKER__` JSON 标记，格式如：
|
||
/// ```json
|
||
/// {"__SUBAGENT_MARKER__": {"session_id": "...", "agent_id": "...", "agent_type": "..."}}
|
||
/// ```
|
||
///
|
||
/// 扫描策略（与 copilot-api 的 subagent-marker.ts 对齐）：
|
||
/// 1. 遍历所有 user 消息（不仅是第一条，因为 context 压缩可能重排消息）
|
||
/// 2. 在消息文本中查找 `__SUBAGENT_MARKER__` 关键字
|
||
/// 3. 找到即判定为子代理请求
|
||
fn detect_subagent(body: &Value) -> bool {
|
||
// 信号 1: 显式 __SUBAGENT_MARKER__（Claude Code 2.x+ 自动注入）
|
||
if extract_system_text(body).contains("__SUBAGENT_MARKER__") {
|
||
return true;
|
||
}
|
||
|
||
if let Some(messages) = body.get("messages").and_then(|m| m.as_array()) {
|
||
for msg in messages {
|
||
if msg.get("role").and_then(|r| r.as_str()) != Some("user") {
|
||
continue;
|
||
}
|
||
let text = extract_text_from_message(msg);
|
||
if text.contains("__SUBAGENT_MARKER__") {
|
||
return true;
|
||
}
|
||
}
|
||
}
|
||
|
||
// 信号 2（fallback）: metadata.user_id 包含子代理标识
|
||
// Claude Code 的 Agent tool 会将 subagent session 标记为
|
||
// "parentSessionId_agent_agentId" 格式，检测 "_agent_" 后缀
|
||
if let Some(user_id) = body.pointer("/metadata/user_id").and_then(|v| v.as_str()) {
|
||
// "_agent_" 是 Claude Code Agent tool 的内部标记
|
||
if user_id.contains("_agent_") {
|
||
return true;
|
||
}
|
||
}
|
||
|
||
// 信号 3（fallback）: system prompt 包含 Claude Code 子代理的典型框架文本
|
||
// Agent tool 生成的子代理会在 system prompt 中包含由 Agent tool 注入的任务描述，
|
||
// 但主对话的 system prompt 由 Claude Code CLI 直接生成，两者格式不同
|
||
// 这个信号不够可靠（用户 prompt 也可能包含这些词），因此只作为辅助判据
|
||
// 暂不启用，预留接口
|
||
|
||
false
|
||
}
|
||
|
||
/// 清理孤立的 tool_result — 没有对应 tool_use 的 tool_result 转为 text block。
|
||
///
|
||
/// 场景：上下文压缩、消息截断等可能导致 assistant 消息中的 tool_use 被删除，
|
||
/// 但后续 user 消息中的 tool_result 仍在。上游 API 可能因不匹配而报错/重试。
|
||
///
|
||
/// 与 copilot-api 的 `sanitizeOrphanToolResults` 对齐。
|
||
pub fn sanitize_orphan_tool_results(mut body: Value) -> Value {
|
||
let messages = match body.get_mut("messages").and_then(|m| m.as_array_mut()) {
|
||
Some(msgs) if msgs.len() >= 2 => msgs,
|
||
_ => return body,
|
||
};
|
||
|
||
// Anthropic 协议要求 tool_result 紧跟其对应 tool_use 所在的 assistant turn。
|
||
// 只检查 messages[i-1]（紧邻上一条 assistant）来判定是否 orphan，
|
||
// 与参考实现 sanitizeOrphanToolResults 对齐。
|
||
for i in 1..messages.len() {
|
||
if messages[i].get("role").and_then(|r| r.as_str()) != Some("user") {
|
||
continue;
|
||
}
|
||
|
||
// 收集紧邻上一条 assistant 的 tool_use id
|
||
let prev_tool_use_ids: HashSet<String> =
|
||
if messages[i - 1].get("role").and_then(|r| r.as_str()) == Some("assistant") {
|
||
messages[i - 1]
|
||
.get("content")
|
||
.and_then(|c| c.as_array())
|
||
.map(|blocks| {
|
||
blocks
|
||
.iter()
|
||
.filter(|b| b.get("type").and_then(|t| t.as_str()) == Some("tool_use"))
|
||
.filter_map(|b| b.get("id").and_then(|i| i.as_str()).map(String::from))
|
||
.collect()
|
||
})
|
||
.unwrap_or_default()
|
||
} else {
|
||
// 上一条不是 assistant → 这条 user 中的所有 tool_result 都是 orphan
|
||
HashSet::new()
|
||
};
|
||
|
||
let content = match messages[i]
|
||
.get_mut("content")
|
||
.and_then(|c| c.as_array_mut())
|
||
{
|
||
Some(blocks) => blocks,
|
||
None => continue,
|
||
};
|
||
|
||
for block in content.iter_mut() {
|
||
if block.get("type").and_then(|t| t.as_str()) != Some("tool_result") {
|
||
continue;
|
||
}
|
||
let tool_use_id = block
|
||
.get("tool_use_id")
|
||
.and_then(|id| id.as_str())
|
||
.unwrap_or("");
|
||
// 空 tool_use_id 或不在紧邻 assistant 的 tool_use 中 → orphan
|
||
if tool_use_id.is_empty() || !prev_tool_use_ids.contains(tool_use_id) {
|
||
let content_text = match block.get("content") {
|
||
Some(c) if c.is_string() => c.as_str().unwrap_or("").to_string(),
|
||
Some(c) if c.is_array() => c
|
||
.as_array()
|
||
.unwrap()
|
||
.iter()
|
||
.filter_map(|b| b.get("text").and_then(|t| t.as_str()))
|
||
.collect::<Vec<_>>()
|
||
.join("\n"),
|
||
_ => String::new(),
|
||
};
|
||
*block = serde_json::json!({
|
||
"type": "text",
|
||
"text": format!("[Tool result for {}]: {}", tool_use_id, content_text)
|
||
});
|
||
}
|
||
}
|
||
}
|
||
|
||
body
|
||
}
|
||
|
||
/// 请求前主动剥离所有 assistant 消息里的 thinking / redacted_thinking block
|
||
///
|
||
/// Copilot 的三条目标端点（`/chat/completions`、`/v1/responses`、`/v1/chat/completions`）
|
||
/// 均为 OpenAI 兼容格式，不识别 Anthropic 的 thinking block。若原样转发，上游会
|
||
/// 拒绝并返回 invalid_request_error —— 届时 `thinking_rectifier` 才做反应式清理并
|
||
/// 重试。那次已经失败的请求依旧消耗一次 premium quota，所以此处提前剥离。
|
||
///
|
||
/// 与 `thinking_rectifier::rectify_anthropic_request` 的区别：
|
||
/// - 本函数只剥 thinking / redacted_thinking 两类 block，不触碰 signature，也不
|
||
/// 移除顶层 thinking 字段——那些是错误路径上的激进整流，常规路径不需要。
|
||
/// - 保持与 `merge_tool_results` / `sanitize_orphan_tool_results` 一致的"消费 body、
|
||
/// 返回新 body"签名，便于接入 forwarder 管道。
|
||
pub fn strip_thinking_blocks(mut body: Value) -> Value {
|
||
let Some(messages) = body.get_mut("messages").and_then(|m| m.as_array_mut()) else {
|
||
return body;
|
||
};
|
||
|
||
for msg in messages.iter_mut() {
|
||
if msg.get("role").and_then(|r| r.as_str()) != Some("assistant") {
|
||
continue;
|
||
}
|
||
let Some(content) = msg.get_mut("content").and_then(|c| c.as_array_mut()) else {
|
||
continue;
|
||
};
|
||
content.retain(|block| {
|
||
!matches!(
|
||
block.get("type").and_then(|t| t.as_str()),
|
||
Some("thinking") | Some("redacted_thinking")
|
||
)
|
||
});
|
||
}
|
||
|
||
body
|
||
}
|
||
|
||
// ─── 内部辅助 ─────────────────────────────────
|
||
|
||
/// 从请求体的 `system` 字段提取文本（处理 string/array 两种格式）。
|
||
fn extract_system_text(body: &Value) -> String {
|
||
match body.get("system") {
|
||
Some(s) if s.is_string() => s.as_str().unwrap_or("").to_string(),
|
||
Some(arr) if arr.is_array() => arr
|
||
.as_array()
|
||
.unwrap()
|
||
.iter()
|
||
.filter_map(|b| b.get("text").and_then(|t| t.as_str()))
|
||
.collect::<Vec<_>>()
|
||
.join(" "),
|
||
_ => String::new(),
|
||
}
|
||
}
|
||
|
||
/// 查找最后一条 user 消息的非 tool_result 文本内容。
|
||
///
|
||
/// 与参考实现的 `findLastUserContent` 对齐：
|
||
/// - 从后往前遍历消息
|
||
/// - 排除 tool_result block
|
||
/// - 排除 cache_control 字段
|
||
fn find_last_user_content(body: &Value) -> Option<String> {
|
||
let messages = body.get("messages").and_then(|m| m.as_array())?;
|
||
|
||
for msg in messages.iter().rev() {
|
||
if msg.get("role").and_then(|r| r.as_str()) != Some("user") {
|
||
continue;
|
||
}
|
||
let content = msg.get("content")?;
|
||
|
||
if let Some(s) = content.as_str() {
|
||
return Some(s.to_string());
|
||
}
|
||
|
||
if let Some(blocks) = content.as_array() {
|
||
// 过滤 tool_result，保留其他 block（去掉 cache_control）
|
||
let filtered: Vec<Value> = blocks
|
||
.iter()
|
||
.filter(|b| b.get("type").and_then(|t| t.as_str()) != Some("tool_result"))
|
||
.map(|b| {
|
||
let mut b = b.clone();
|
||
if let Some(obj) = b.as_object_mut() {
|
||
obj.remove("cache_control");
|
||
}
|
||
b
|
||
})
|
||
.collect();
|
||
|
||
if !filtered.is_empty() {
|
||
return Some(serde_json::to_string(&filtered).unwrap_or_default());
|
||
}
|
||
}
|
||
}
|
||
|
||
None
|
||
}
|
||
|
||
/// 将 text block 合并进 tool_result block。
|
||
///
|
||
/// 两种合并策略（与参考实现对齐）：
|
||
/// - 数量相等：一一对应，text 追加到对应 tool_result 的 content 中
|
||
/// - 数量不等：所有 text 追加到最后一个 tool_result 的 content 中
|
||
fn merge_blocks_into_tool_results(
|
||
mut tool_results: Vec<Value>,
|
||
text_blocks: Vec<Value>,
|
||
) -> Vec<Value> {
|
||
if tool_results.len() == text_blocks.len() {
|
||
// 一一对应合并
|
||
for (tr, tb) in tool_results.iter_mut().zip(text_blocks.iter()) {
|
||
append_text_to_tool_result(tr, tb);
|
||
}
|
||
} else {
|
||
// 所有 text 追加到最后一个 tool_result
|
||
if let Some(last_tr) = tool_results.last_mut() {
|
||
for tb in &text_blocks {
|
||
append_text_to_tool_result(last_tr, tb);
|
||
}
|
||
}
|
||
}
|
||
tool_results
|
||
}
|
||
|
||
/// 将 text block 的内容追加到 tool_result 的 content 中
|
||
fn append_text_to_tool_result(tool_result: &mut Value, text_block: &Value) {
|
||
let text = text_block
|
||
.get("text")
|
||
.and_then(|t| t.as_str())
|
||
.unwrap_or("");
|
||
if text.trim().is_empty() {
|
||
return;
|
||
}
|
||
|
||
// tool_result 的 content 可以是字符串或数组
|
||
match tool_result.get("content") {
|
||
Some(c) if c.is_string() => {
|
||
let existing = c.as_str().unwrap_or("");
|
||
tool_result["content"] = Value::String(format!("{existing}\n{text}"));
|
||
}
|
||
Some(c) if c.is_array() => {
|
||
let arr = tool_result["content"].as_array_mut().unwrap();
|
||
arr.push(serde_json::json!({"type": "text", "text": text}));
|
||
}
|
||
_ => {
|
||
// content 缺失或 null — 直接设置
|
||
tool_result["content"] = Value::String(text.to_string());
|
||
}
|
||
}
|
||
}
|
||
|
||
/// 从消息中提取文本内容
|
||
fn extract_text_from_message(msg: &Value) -> String {
|
||
match msg.get("content") {
|
||
Some(content) if content.is_string() => content.as_str().unwrap_or("").to_string(),
|
||
Some(content) if content.is_array() => {
|
||
let blocks = content.as_array().unwrap();
|
||
blocks
|
||
.iter()
|
||
.filter_map(|block| {
|
||
if block.get("type").and_then(|t| t.as_str()) == Some("text") {
|
||
block.get("text").and_then(|t| t.as_str())
|
||
} else {
|
||
None
|
||
}
|
||
})
|
||
.collect::<Vec<_>>()
|
||
.join(" ")
|
||
}
|
||
_ => String::new(),
|
||
}
|
||
}
|
||
|
||
/// 判断消息是否为 tool_result-only 的用户消息
|
||
fn is_tool_result_only_message(msg: &Value) -> bool {
|
||
if msg.get("role").and_then(|r| r.as_str()) != Some("user") {
|
||
return false;
|
||
}
|
||
match msg.get("content").and_then(|c| c.as_array()) {
|
||
Some(blocks) if !blocks.is_empty() => blocks
|
||
.iter()
|
||
.all(|block| block.get("type").and_then(|t| t.as_str()) == Some("tool_result")),
|
||
_ => false,
|
||
}
|
||
}
|
||
|
||
// ─── 测试 ─────────────────────────────────────
|
||
|
||
#[cfg(test)]
|
||
mod tests {
|
||
use super::*;
|
||
use serde_json::json;
|
||
|
||
// === classify_request 测试 ===
|
||
|
||
#[test]
|
||
fn test_classify_user_text_message() {
|
||
let body = json!({
|
||
"model": "claude-sonnet-4-20250514",
|
||
"messages": [
|
||
{"role": "user", "content": "Hello, please help me write some code"}
|
||
]
|
||
});
|
||
let result = classify_request(&body, false, true, false);
|
||
assert_eq!(result.initiator, "user");
|
||
assert!(!result.is_compact);
|
||
}
|
||
|
||
#[test]
|
||
fn test_classify_user_text_array_message() {
|
||
let body = json!({
|
||
"model": "claude-sonnet-4-20250514",
|
||
"messages": [
|
||
{"role": "user", "content": [
|
||
{"type": "text", "text": "Please explain this code"}
|
||
]}
|
||
]
|
||
});
|
||
let result = classify_request(&body, false, true, false);
|
||
assert_eq!(result.initiator, "user");
|
||
}
|
||
|
||
#[test]
|
||
fn test_classify_tool_result_only() {
|
||
let body = json!({
|
||
"model": "claude-sonnet-4-20250514",
|
||
"tools": [{"name": "Read", "description": "Read a file", "input_schema": {}}],
|
||
"messages": [
|
||
{"role": "user", "content": "Read the file"},
|
||
{"role": "assistant", "content": [
|
||
{"type": "text", "text": "I'll read that file."},
|
||
{"type": "tool_use", "id": "toolu_123", "name": "Read", "input": {"path": "/tmp/test.rs"}}
|
||
]},
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "toolu_123", "content": "file contents here"}
|
||
]}
|
||
]
|
||
});
|
||
let result = classify_request(&body, true, true, false);
|
||
assert_eq!(result.initiator, "agent");
|
||
assert!(!result.is_warmup);
|
||
}
|
||
|
||
#[test]
|
||
fn test_classify_tool_result_with_text_block() {
|
||
// tool_result + text block（skill/edit hook/plan follow-up 的常见形态）
|
||
// 含有 tool_result → 视为工具续写 → agent
|
||
// 与 copilot-api 的 merge-then-classify 效果对齐
|
||
let body = json!({
|
||
"model": "claude-sonnet-4-20250514",
|
||
"messages": [
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "toolu_123", "content": "file contents"},
|
||
{"type": "text", "text": "Now please refactor this code"}
|
||
]}
|
||
]
|
||
});
|
||
let result = classify_request(&body, false, true, false);
|
||
assert_eq!(result.initiator, "agent");
|
||
}
|
||
|
||
#[test]
|
||
fn test_classify_empty_messages() {
|
||
let body = json!({
|
||
"model": "claude-sonnet-4-20250514",
|
||
"messages": []
|
||
});
|
||
let result = classify_request(&body, false, true, false);
|
||
assert_eq!(result.initiator, "user");
|
||
}
|
||
|
||
#[test]
|
||
fn test_classify_no_messages() {
|
||
let body = json!({"model": "claude-sonnet-4-20250514"});
|
||
let result = classify_request(&body, false, true, false);
|
||
assert_eq!(result.initiator, "user");
|
||
}
|
||
|
||
#[test]
|
||
fn test_classify_compact_request_system_prompt() {
|
||
// compact 通过 system prompt 强特征检测
|
||
let body = json!({
|
||
"model": "claude-sonnet-4-20250514",
|
||
"system": "You are a helpful AI assistant tasked with summarizing conversations. Please create a summary.",
|
||
"messages": [
|
||
{"role": "user", "content": "Here is the conversation history to summarize..."}
|
||
]
|
||
});
|
||
let result = classify_request(&body, false, true, false);
|
||
assert_eq!(result.initiator, "agent");
|
||
assert!(result.is_compact);
|
||
}
|
||
|
||
#[test]
|
||
fn test_classify_compact_request_critical_marker() {
|
||
// compact 通过 CRITICAL 机器指令检测
|
||
let body = json!({
|
||
"model": "claude-sonnet-4-20250514",
|
||
"messages": [
|
||
{"role": "user", "content": [
|
||
{"type": "text", "text": "CRITICAL: Respond with TEXT ONLY. Do NOT call any tools. Summarize the conversation."}
|
||
]}
|
||
]
|
||
});
|
||
let result = classify_request(&body, false, true, false);
|
||
assert_eq!(result.initiator, "agent");
|
||
assert!(result.is_compact);
|
||
}
|
||
|
||
#[test]
|
||
fn test_classify_compact_disabled_by_config() {
|
||
// compact_detection=false 时，即使内容匹配也不标记为 compact
|
||
let body = json!({
|
||
"model": "claude-sonnet-4-20250514",
|
||
"system": "You are a helpful AI assistant tasked with summarizing conversations.",
|
||
"messages": [
|
||
{"role": "user", "content": "Summarize"}
|
||
]
|
||
});
|
||
let result = classify_request(&body, false, false, false); // compact_detection=false
|
||
assert_eq!(result.initiator, "user"); // 不被标记为 agent
|
||
assert!(!result.is_compact);
|
||
}
|
||
|
||
#[test]
|
||
fn test_no_false_positive_on_user_summarize_request() {
|
||
// P1 修复验证：用户手动输入 "summarize the conversation" 不应被误判为 compact
|
||
let body = json!({
|
||
"model": "claude-sonnet-4-20250514",
|
||
"messages": [
|
||
{"role": "user", "content": "Please summarize the conversation so far into a concise summary."}
|
||
]
|
||
});
|
||
let result = classify_request(&body, false, true, false);
|
||
// 没有 system prompt 强特征，也没有 CRITICAL 指令 → 不是 compact → user
|
||
assert_eq!(result.initiator, "user");
|
||
assert!(!result.is_compact);
|
||
}
|
||
|
||
// === warmup 测试（与参考实现对齐） ===
|
||
|
||
#[test]
|
||
fn test_warmup_with_anthropic_beta_no_tools() {
|
||
let body = json!({
|
||
"model": "claude-sonnet-4-20250514",
|
||
"messages": [
|
||
{"role": "user", "content": "Hello"}
|
||
]
|
||
});
|
||
// has_anthropic_beta=true, 无 tools → warmup
|
||
let result = classify_request(&body, true, true, false);
|
||
assert!(result.is_warmup);
|
||
}
|
||
|
||
#[test]
|
||
fn test_not_warmup_without_anthropic_beta() {
|
||
let body = json!({
|
||
"model": "claude-sonnet-4-20250514",
|
||
"messages": [
|
||
{"role": "user", "content": "Hello"}
|
||
]
|
||
});
|
||
// has_anthropic_beta=false → 不是 warmup
|
||
let result = classify_request(&body, false, true, false);
|
||
assert!(!result.is_warmup);
|
||
}
|
||
|
||
#[test]
|
||
fn test_not_warmup_with_tools() {
|
||
let body = json!({
|
||
"model": "claude-sonnet-4-20250514",
|
||
"tools": [{"name": "Read", "description": "Read a file", "input_schema": {}}],
|
||
"messages": [
|
||
{"role": "user", "content": "Hello"}
|
||
]
|
||
});
|
||
// 有 tools → 不是 warmup（即使有 anthropic-beta）
|
||
let result = classify_request(&body, true, true, false);
|
||
assert!(!result.is_warmup);
|
||
}
|
||
|
||
#[test]
|
||
fn test_not_warmup_when_agent() {
|
||
// tool_result → agent → 不判定 warmup
|
||
let body = json!({
|
||
"model": "claude-sonnet-4-20250514",
|
||
"messages": [
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "toolu_123", "content": "ok"}
|
||
]}
|
||
]
|
||
});
|
||
let result = classify_request(&body, true, true, false);
|
||
assert_eq!(result.initiator, "agent");
|
||
assert!(!result.is_warmup);
|
||
}
|
||
|
||
// === merge_tool_results 测试 ===
|
||
|
||
#[test]
|
||
fn test_merge_intra_message_tool_result_text() {
|
||
// 核心场景：消息内部 tool_result + text → text 被吸收进 tool_result
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "t1", "content": "file contents"},
|
||
{"type": "text", "text": "skill output here"}
|
||
]}
|
||
]
|
||
});
|
||
let result = merge_tool_results(body);
|
||
let content = result["messages"][0]["content"].as_array().unwrap();
|
||
// 应只剩 1 个 tool_result block（text 被吸收）
|
||
assert_eq!(content.len(), 1);
|
||
assert_eq!(content[0]["type"], "tool_result");
|
||
// tool_result 的 content 应包含原始内容 + 吸收的 text
|
||
let tr_content = content[0]["content"].as_str().unwrap();
|
||
assert!(tr_content.contains("file contents"));
|
||
assert!(tr_content.contains("skill output here"));
|
||
}
|
||
|
||
#[test]
|
||
fn test_merge_intra_message_equal_count() {
|
||
// 数量相等：一一对应合并
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "t1", "content": "result1"},
|
||
{"type": "text", "text": "text1"},
|
||
{"type": "tool_result", "tool_use_id": "t2", "content": "result2"},
|
||
{"type": "text", "text": "text2"}
|
||
]}
|
||
]
|
||
});
|
||
let result = merge_tool_results(body);
|
||
let content = result["messages"][0]["content"].as_array().unwrap();
|
||
assert_eq!(content.len(), 2);
|
||
assert!(content[0]["content"].as_str().unwrap().contains("text1"));
|
||
assert!(content[1]["content"].as_str().unwrap().contains("text2"));
|
||
}
|
||
|
||
#[test]
|
||
fn test_merge_intra_message_empty_text_ignored() {
|
||
// 空 text block 不追加内容
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "t1", "content": "result"},
|
||
{"type": "text", "text": ""}
|
||
]}
|
||
]
|
||
});
|
||
let result = merge_tool_results(body);
|
||
let content = result["messages"][0]["content"].as_array().unwrap();
|
||
assert_eq!(content.len(), 1);
|
||
// 空 text 不改变原始 content
|
||
assert_eq!(content[0]["content"], "result");
|
||
}
|
||
|
||
#[test]
|
||
fn test_merge_intra_skips_other_block_types() {
|
||
// 有非 tool_result/text 的 block → 跳过整条消息
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "t1", "content": "result"},
|
||
{"type": "image", "source": {"data": "..."}},
|
||
{"type": "text", "text": "caption"}
|
||
]}
|
||
]
|
||
});
|
||
let result = merge_tool_results(body);
|
||
let content = result["messages"][0]["content"].as_array().unwrap();
|
||
// 未合并，保持原样 3 个 block
|
||
assert_eq!(content.len(), 3);
|
||
}
|
||
|
||
#[test]
|
||
fn test_merge_cross_message_consecutive() {
|
||
// 跨消息合并：连续 tool_result-only 用户消息
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "user", "content": "Read files"},
|
||
{"role": "assistant", "content": [
|
||
{"type": "tool_use", "id": "t1", "name": "Read", "input": {}},
|
||
{"type": "tool_use", "id": "t2", "name": "Read", "input": {}}
|
||
]},
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "t1", "content": "file1"}
|
||
]},
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "t2", "content": "file2"}
|
||
]}
|
||
]
|
||
});
|
||
let result = merge_tool_results(body);
|
||
let messages = result["messages"].as_array().unwrap();
|
||
assert_eq!(messages.len(), 3);
|
||
let merged_content = messages[2]["content"].as_array().unwrap();
|
||
assert_eq!(merged_content.len(), 2);
|
||
}
|
||
|
||
#[test]
|
||
fn test_merge_does_not_affect_normal_messages() {
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "user", "content": "Hello"},
|
||
{"role": "assistant", "content": "Hi!"},
|
||
{"role": "user", "content": "How are you?"}
|
||
]
|
||
});
|
||
let result = merge_tool_results(body.clone());
|
||
assert_eq!(result["messages"], body["messages"]);
|
||
}
|
||
|
||
// === deterministic_request_id 测试 ===
|
||
|
||
#[test]
|
||
fn test_deterministic_id_stable() {
|
||
let body = json!({
|
||
"model": "claude-sonnet-4-20250514",
|
||
"messages": [{"role": "user", "content": "Hello"}]
|
||
});
|
||
let id1 = deterministic_request_id(&body, "session1");
|
||
let id2 = deterministic_request_id(&body, "session1");
|
||
assert_eq!(id1, id2);
|
||
}
|
||
|
||
#[test]
|
||
fn test_deterministic_id_varies_by_content() {
|
||
let body1 = json!({
|
||
"messages": [{"role": "user", "content": "Hello"}]
|
||
});
|
||
let body2 = json!({
|
||
"messages": [{"role": "user", "content": "Goodbye"}]
|
||
});
|
||
let id1 = deterministic_request_id(&body1, "session1");
|
||
let id2 = deterministic_request_id(&body2, "session1");
|
||
assert_ne!(id1, id2);
|
||
}
|
||
|
||
#[test]
|
||
fn test_deterministic_id_varies_by_session() {
|
||
let body = json!({
|
||
"messages": [{"role": "user", "content": "Hello"}]
|
||
});
|
||
let id1 = deterministic_request_id(&body, "session1");
|
||
let id2 = deterministic_request_id(&body, "session2");
|
||
assert_ne!(id1, id2);
|
||
}
|
||
|
||
#[test]
|
||
fn test_deterministic_id_ignores_tool_result() {
|
||
// tool_result 内容不同，但 user text 相同 → 相同 ID
|
||
let body1 = json!({
|
||
"messages": [
|
||
{"role": "user", "content": "Hello"},
|
||
{"role": "assistant", "content": "Hi"},
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "t1", "content": "version_A"}
|
||
]},
|
||
{"role": "user", "content": "do something"}
|
||
]
|
||
});
|
||
let body2 = json!({
|
||
"messages": [
|
||
{"role": "user", "content": "Hello"},
|
||
{"role": "assistant", "content": "Hi"},
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "t1", "content": "version_B"}
|
||
]},
|
||
{"role": "user", "content": "do something"}
|
||
]
|
||
});
|
||
let id1 = deterministic_request_id(&body1, "s");
|
||
let id2 = deterministic_request_id(&body2, "s");
|
||
assert_eq!(id1, id2);
|
||
}
|
||
|
||
#[test]
|
||
fn test_deterministic_id_fallback_when_no_user_content() {
|
||
// 无用户消息 → 退化为随机 UUID（每次不同）
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "assistant", "content": "Hi"}
|
||
]
|
||
});
|
||
let id1 = deterministic_request_id(&body, "s");
|
||
let id2 = deterministic_request_id(&body, "s");
|
||
// 随机 UUID，每次应不同
|
||
assert_ne!(id1, id2);
|
||
}
|
||
|
||
#[test]
|
||
fn test_deterministic_id_is_valid_uuid() {
|
||
let body = json!({
|
||
"messages": [{"role": "user", "content": "test"}]
|
||
});
|
||
let id = deterministic_request_id(&body, "session");
|
||
assert!(Uuid::parse_str(&id).is_ok());
|
||
}
|
||
|
||
// === deterministic_interaction_id 测试 ===
|
||
|
||
#[test]
|
||
fn test_interaction_id_stable_for_same_session() {
|
||
let id1 = deterministic_interaction_id("session_abc");
|
||
let id2 = deterministic_interaction_id("session_abc");
|
||
assert_eq!(id1, id2);
|
||
}
|
||
|
||
#[test]
|
||
fn test_interaction_id_differs_across_sessions() {
|
||
let id1 = deterministic_interaction_id("session_abc");
|
||
let id2 = deterministic_interaction_id("session_def");
|
||
assert_ne!(id1, id2);
|
||
}
|
||
|
||
#[test]
|
||
fn test_interaction_id_differs_from_request_id() {
|
||
let body = json!({
|
||
"messages": [{"role": "user", "content": "Hello"}]
|
||
});
|
||
let interaction = deterministic_interaction_id("session_abc").unwrap();
|
||
let request = deterministic_request_id(&body, "session_abc");
|
||
assert_ne!(interaction, request);
|
||
}
|
||
|
||
#[test]
|
||
fn test_interaction_id_empty_session_is_none() {
|
||
// 无 session 时不应生成 interaction ID（避免碎片化）
|
||
assert!(deterministic_interaction_id("").is_none());
|
||
}
|
||
|
||
#[test]
|
||
fn test_interaction_id_is_valid_uuid() {
|
||
let id = deterministic_interaction_id("test_session").unwrap();
|
||
assert!(Uuid::parse_str(&id).is_ok());
|
||
}
|
||
|
||
// === compact 检测增强测试 ===
|
||
|
||
#[test]
|
||
fn test_compact_detection_system_prompt() {
|
||
let body = json!({
|
||
"system": "You are a helpful AI assistant tasked with summarizing conversations. Please provide a concise summary.",
|
||
"messages": [
|
||
{"role": "user", "content": "Here is the conversation to summarize..."}
|
||
]
|
||
});
|
||
assert!(is_compact_request(&body));
|
||
}
|
||
|
||
#[test]
|
||
fn test_compact_detection_critical_keyword() {
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "user", "content": "CRITICAL: Respond with TEXT ONLY. Do NOT call any tools. Summarize this conversation."}
|
||
]
|
||
});
|
||
assert!(is_compact_request(&body));
|
||
}
|
||
|
||
#[test]
|
||
fn test_compact_detection_structural_markers() {
|
||
// Claude Code compact 特有的结构标记
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "user", "content": "Summary of conversation:\n\nPending Tasks:\n- Fix bug\n\nCurrent Work:\n- Implementing feature"}
|
||
]
|
||
});
|
||
assert!(is_compact_request(&body));
|
||
}
|
||
|
||
#[test]
|
||
fn test_compact_no_false_positive_on_generic_summary() {
|
||
// 通用短语不应触发 compact 检测
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "user", "content": "Your task is to create a detailed summary of the conversation so far."}
|
||
]
|
||
});
|
||
assert!(!is_compact_request(&body));
|
||
}
|
||
|
||
#[test]
|
||
fn test_compact_detection_negative() {
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "user", "content": "What is the weather today?"}
|
||
]
|
||
});
|
||
assert!(!is_compact_request(&body));
|
||
}
|
||
|
||
#[test]
|
||
fn test_compact_detection_system_array() {
|
||
let body = json!({
|
||
"system": [
|
||
{"type": "text", "text": "You are a helpful AI assistant tasked with summarizing conversations."}
|
||
],
|
||
"messages": [
|
||
{"role": "user", "content": "Summarize"}
|
||
]
|
||
});
|
||
assert!(is_compact_request(&body));
|
||
}
|
||
|
||
// === detect_subagent 测试 ===
|
||
|
||
#[test]
|
||
fn test_detect_subagent_with_marker_in_user_message() {
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "user", "content": [
|
||
{"type": "text", "text": "<system-reminder>\n{\"__SUBAGENT_MARKER__\":{\"session_id\":\"abc123\",\"agent_id\":\"explore-1\",\"agent_type\":\"Explore\"}}\n</system-reminder>\nPlease search the codebase for auth handlers"}
|
||
]}
|
||
]
|
||
});
|
||
assert!(detect_subagent(&body));
|
||
}
|
||
|
||
#[test]
|
||
fn test_detect_subagent_with_marker_in_system() {
|
||
let body = json!({
|
||
"system": "You are an agent. {\"__SUBAGENT_MARKER__\":{\"session_id\":\"abc\",\"agent_id\":\"plan-1\",\"agent_type\":\"Plan\"}}",
|
||
"messages": [
|
||
{"role": "user", "content": "Design the implementation plan"}
|
||
]
|
||
});
|
||
assert!(detect_subagent(&body));
|
||
}
|
||
|
||
#[test]
|
||
fn test_detect_subagent_no_marker() {
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "user", "content": "Hello, please help me write code"}
|
||
]
|
||
});
|
||
assert!(!detect_subagent(&body));
|
||
}
|
||
|
||
#[test]
|
||
fn test_detect_subagent_via_metadata_user_id() {
|
||
// fallback 信号: metadata.user_id 包含 "_agent_" 标记
|
||
let body = json!({
|
||
"metadata": {
|
||
"user_id": "session_abc123_agent_explore-1"
|
||
},
|
||
"messages": [
|
||
{"role": "user", "content": "Search for files"}
|
||
]
|
||
});
|
||
assert!(detect_subagent(&body));
|
||
}
|
||
|
||
#[test]
|
||
fn test_detect_subagent_normal_user_id_not_matched() {
|
||
// 普通 session ID 不应被误判
|
||
let body = json!({
|
||
"metadata": {
|
||
"user_id": "session_abc123"
|
||
},
|
||
"messages": [
|
||
{"role": "user", "content": "Hello"}
|
||
]
|
||
});
|
||
assert!(!detect_subagent(&body));
|
||
}
|
||
|
||
#[test]
|
||
fn test_classify_subagent_sets_agent_initiator() {
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "user", "content": [
|
||
{"type": "text", "text": "<system-reminder>\n{\"__SUBAGENT_MARKER__\":{\"session_id\":\"abc\",\"agent_id\":\"explore-1\",\"agent_type\":\"Explore\"}}\n</system-reminder>\nSearch for files"}
|
||
]}
|
||
]
|
||
});
|
||
let result = classify_request(&body, false, true, true);
|
||
assert_eq!(result.initiator, "agent");
|
||
assert!(result.is_subagent);
|
||
}
|
||
|
||
#[test]
|
||
fn test_classify_subagent_disabled_flag() {
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "user", "content": [
|
||
{"type": "text", "text": "<system-reminder>\n{\"__SUBAGENT_MARKER__\":{\"session_id\":\"abc\",\"agent_id\":\"explore-1\",\"agent_type\":\"Explore\"}}\n</system-reminder>\nSearch for files"}
|
||
]}
|
||
]
|
||
});
|
||
// subagent_detection=false → 不检测子代理
|
||
let result = classify_request(&body, false, true, false);
|
||
assert_eq!(result.initiator, "user");
|
||
assert!(!result.is_subagent);
|
||
}
|
||
|
||
// === sanitize_orphan_tool_results 测试 ===
|
||
|
||
#[test]
|
||
fn test_sanitize_orphan_tool_results_converts_orphans() {
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "user", "content": "Help me"},
|
||
{"role": "assistant", "content": [
|
||
{"type": "tool_use", "id": "tool_1", "name": "read_file", "input": {}}
|
||
]},
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "tool_1", "content": "file contents"},
|
||
{"type": "tool_result", "tool_use_id": "tool_orphan", "content": "orphan data"}
|
||
]}
|
||
]
|
||
});
|
||
let result = sanitize_orphan_tool_results(body);
|
||
let msgs = result["messages"].as_array().unwrap();
|
||
let last_content = msgs[2]["content"].as_array().unwrap();
|
||
// tool_1 保留为 tool_result
|
||
assert_eq!(last_content[0]["type"], "tool_result");
|
||
// tool_orphan 转为 text
|
||
assert_eq!(last_content[1]["type"], "text");
|
||
assert!(last_content[1]["text"]
|
||
.as_str()
|
||
.unwrap()
|
||
.contains("tool_orphan"));
|
||
}
|
||
|
||
#[test]
|
||
fn test_sanitize_orphan_tool_results_no_orphans() {
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "assistant", "content": [
|
||
{"type": "tool_use", "id": "tool_1", "name": "read_file", "input": {}}
|
||
]},
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "tool_1", "content": "ok"}
|
||
]}
|
||
]
|
||
});
|
||
let result = sanitize_orphan_tool_results(body.clone());
|
||
// 无孤立 tool_result，不应有变化
|
||
assert_eq!(result["messages"][1]["content"][0]["type"], "tool_result");
|
||
}
|
||
|
||
#[test]
|
||
fn test_sanitize_orphan_non_adjacent_assistant_tool_use_is_orphan() {
|
||
// tool_use 在更早的 assistant 中，但 tool_result 的紧邻上一条是另一个 assistant
|
||
// → 对 Anthropic 协议来说这个 tool_result 是 orphan
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "user", "content": "step 1"},
|
||
{"role": "assistant", "content": [
|
||
{"type": "tool_use", "id": "old_tool", "name": "search", "input": {}}
|
||
]},
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "old_tool", "content": "found it"}
|
||
]},
|
||
{"role": "assistant", "content": [
|
||
{"type": "text", "text": "OK, now let me think..."}
|
||
]},
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "old_tool", "content": "stale ref"}
|
||
]}
|
||
]
|
||
});
|
||
let result = sanitize_orphan_tool_results(body);
|
||
let msgs = result["messages"].as_array().unwrap();
|
||
// messages[2]: 紧邻 assistant 有 old_tool → 保留
|
||
assert_eq!(msgs[2]["content"][0]["type"], "tool_result");
|
||
// messages[4]: 紧邻 assistant 无 tool_use → orphan → text
|
||
assert_eq!(msgs[4]["content"][0]["type"], "text");
|
||
}
|
||
|
||
#[test]
|
||
fn test_sanitize_orphan_prev_not_assistant() {
|
||
// tool_result 紧邻上一条是 user（非 assistant）→ 全部 orphan
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "user", "content": "first"},
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "t1", "content": "data"}
|
||
]}
|
||
]
|
||
});
|
||
let result = sanitize_orphan_tool_results(body);
|
||
assert_eq!(result["messages"][1]["content"][0]["type"], "text");
|
||
}
|
||
|
||
/// 关键场景：orphan tool_result（上下文压缩丢失了紧邻 tool_use）
|
||
/// 在分类时仍应被视为 agent continuation，不能因为后续的 sanitize
|
||
/// 将其转为 text 而变成 user 请求。
|
||
///
|
||
/// 这个测试验证 classify_request 在原始（未 sanitize）的 body 上
|
||
/// 正确识别 orphan tool_result 为 agent。
|
||
#[test]
|
||
fn test_orphan_tool_result_classified_as_agent_before_sanitize() {
|
||
// 场景：最后一条 user 消息全是 tool_result，但紧邻的 assistant
|
||
// 消息里没有对应的 tool_use（因上下文压缩丢失了）
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "assistant", "content": "I'll help you with that."},
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "orphan_tool_1", "content": "file contents here"},
|
||
{"type": "tool_result", "tool_use_id": "orphan_tool_2", "content": "another result"}
|
||
]}
|
||
]
|
||
});
|
||
// 在原始 body 上分类 → 全是 tool_result → agent
|
||
let classification = classify_request(&body, false, false, false);
|
||
assert_eq!(classification.initiator, "agent");
|
||
|
||
// sanitize 后 → tool_result 变为 text → 如果再分类就会变成 user
|
||
let sanitized = sanitize_orphan_tool_results(body);
|
||
let classification_after = classify_request(&sanitized, false, false, false);
|
||
assert_eq!(
|
||
classification_after.initiator, "user",
|
||
"sanitize 后 orphan tool_result 变为 text，分类变成 user — \
|
||
这就是为什么分类必须在 sanitize 之前执行"
|
||
);
|
||
}
|
||
|
||
/// orphan tool_result + text 混合场景：
|
||
/// 分类器直接识别含 tool_result 的消息为 agent（无论是否有 text block），
|
||
/// 不依赖 merge 的执行顺序。即使 orphan tool_result 后续被 sanitize 转为 text，
|
||
/// 分类结果在此之前已经确定为 agent。
|
||
#[test]
|
||
fn test_orphan_tool_result_with_text_classified_as_agent() {
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "assistant", "content": "Processing..."},
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "orphan_1", "content": "result data"},
|
||
{"type": "text", "text": "Here's the output from the tool"}
|
||
]}
|
||
]
|
||
});
|
||
// 含有 tool_result → agent（无论是否有 text block）
|
||
let classification = classify_request(&body, false, false, false);
|
||
assert_eq!(classification.initiator, "agent");
|
||
|
||
// sanitize 后 orphan tool_result 变为 text → 纯 text → 分类会变成 user
|
||
// 但正确的执行顺序是先分类再 sanitize，所以这不是问题
|
||
let sanitized = sanitize_orphan_tool_results(body);
|
||
let classification_after = classify_request(&sanitized, false, false, false);
|
||
assert_eq!(classification_after.initiator, "user");
|
||
}
|
||
|
||
#[test]
|
||
fn test_sanitize_orphan_empty_tool_use_id_is_orphan() {
|
||
// tool_use_id 为空或缺失 → 无法匹配任何 tool_use → orphan
|
||
let body = json!({
|
||
"messages": [
|
||
{"role": "assistant", "content": [
|
||
{"type": "tool_use", "id": "tool_1", "name": "read", "input": {}}
|
||
]},
|
||
{"role": "user", "content": [
|
||
{"type": "tool_result", "tool_use_id": "", "content": "empty id"},
|
||
{"type": "tool_result", "content": "missing id field"}
|
||
]}
|
||
]
|
||
});
|
||
let result = sanitize_orphan_tool_results(body);
|
||
let content = result["messages"][1]["content"].as_array().unwrap();
|
||
assert_eq!(content[0]["type"], "text");
|
||
assert_eq!(content[1]["type"], "text");
|
||
}
|
||
|
||
// === strip_thinking_blocks 测试 ===
|
||
|
||
#[test]
|
||
fn test_strip_thinking_removes_assistant_thinking_blocks() {
|
||
let body = serde_json::json!({
|
||
"messages": [
|
||
{"role": "user", "content": [{"type": "text", "text": "hi"}]},
|
||
{"role": "assistant", "content": [
|
||
{"type": "thinking", "thinking": "let me ponder", "signature": "sig"},
|
||
{"type": "redacted_thinking", "data": "opaque"},
|
||
{"type": "text", "text": "hello"},
|
||
{"type": "tool_use", "id": "t1", "name": "read", "input": {}}
|
||
]}
|
||
]
|
||
});
|
||
let result = strip_thinking_blocks(body);
|
||
let content = result["messages"][1]["content"].as_array().unwrap();
|
||
assert_eq!(content.len(), 2);
|
||
assert_eq!(content[0]["type"], "text");
|
||
assert_eq!(content[1]["type"], "tool_use");
|
||
}
|
||
|
||
#[test]
|
||
fn test_strip_thinking_leaves_user_messages_untouched() {
|
||
// 仅处理 assistant，user 的 thinking 块（极少见，但可能）不动
|
||
let body = serde_json::json!({
|
||
"messages": [
|
||
{"role": "user", "content": [
|
||
{"type": "thinking", "thinking": "x"},
|
||
{"type": "text", "text": "hi"}
|
||
]}
|
||
]
|
||
});
|
||
let result = strip_thinking_blocks(body);
|
||
let content = result["messages"][0]["content"].as_array().unwrap();
|
||
assert_eq!(content.len(), 2);
|
||
}
|
||
|
||
#[test]
|
||
fn test_strip_thinking_handles_missing_messages() {
|
||
let body = serde_json::json!({ "model": "claude-3-5-sonnet" });
|
||
let result = strip_thinking_blocks(body.clone());
|
||
assert_eq!(result, body);
|
||
}
|
||
|
||
#[test]
|
||
fn test_strip_thinking_leaves_empty_content_array() {
|
||
// 仅含 thinking 的 assistant 消息剥完后 content 为空——保留上游自处理
|
||
let body = serde_json::json!({
|
||
"messages": [
|
||
{"role": "assistant", "content": [
|
||
{"type": "thinking", "thinking": "solo"}
|
||
]}
|
||
]
|
||
});
|
||
let result = strip_thinking_blocks(body);
|
||
let content = result["messages"][0]["content"].as_array().unwrap();
|
||
assert_eq!(content.len(), 0);
|
||
}
|
||
|
||
#[test]
|
||
fn test_strip_thinking_preserves_signature_on_non_thinking_blocks() {
|
||
// signature 留给 thinking_rectifier 在错误路径处理，此处不动
|
||
let body = serde_json::json!({
|
||
"messages": [
|
||
{"role": "assistant", "content": [
|
||
{"type": "tool_use", "id": "t1", "name": "x", "input": {}, "signature": "s"}
|
||
]}
|
||
]
|
||
});
|
||
let result = strip_thinking_blocks(body);
|
||
let block = &result["messages"][0]["content"][0];
|
||
assert_eq!(block["signature"], "s");
|
||
}
|
||
|
||
#[test]
|
||
fn test_strip_thinking_multiple_assistant_turns() {
|
||
let body = serde_json::json!({
|
||
"messages": [
|
||
{"role": "user", "content": [{"type": "text", "text": "q1"}]},
|
||
{"role": "assistant", "content": [
|
||
{"type": "thinking", "thinking": "a"},
|
||
{"type": "text", "text": "r1"}
|
||
]},
|
||
{"role": "user", "content": [{"type": "text", "text": "q2"}]},
|
||
{"role": "assistant", "content": [
|
||
{"type": "redacted_thinking", "data": "x"},
|
||
{"type": "text", "text": "r2"}
|
||
]}
|
||
]
|
||
});
|
||
let result = strip_thinking_blocks(body);
|
||
let a1 = result["messages"][1]["content"].as_array().unwrap();
|
||
let a2 = result["messages"][3]["content"].as_array().unwrap();
|
||
assert_eq!(a1.len(), 1);
|
||
assert_eq!(a1[0]["text"], "r1");
|
||
assert_eq!(a2.len(), 1);
|
||
assert_eq!(a2[0]["text"], "r2");
|
||
}
|
||
|
||
#[test]
|
||
fn test_strip_thinking_ignores_string_content() {
|
||
// assistant.content 是字符串而非 block 数组 — 历史请求或极简客户端会这样
|
||
// 不应崩溃，也不应转换结构
|
||
let body = serde_json::json!({
|
||
"messages": [
|
||
{"role": "assistant", "content": "plain text response"}
|
||
]
|
||
});
|
||
let result = strip_thinking_blocks(body.clone());
|
||
assert_eq!(result, body);
|
||
}
|
||
|
||
#[test]
|
||
fn test_strip_thinking_preserves_block_order() {
|
||
let body = serde_json::json!({
|
||
"messages": [
|
||
{"role": "assistant", "content": [
|
||
{"type": "thinking", "thinking": "pre"},
|
||
{"type": "text", "text": "A"},
|
||
{"type": "tool_use", "id": "t1", "name": "x", "input": {}},
|
||
{"type": "redacted_thinking", "data": "mid"},
|
||
{"type": "text", "text": "B"}
|
||
]}
|
||
]
|
||
});
|
||
let result = strip_thinking_blocks(body);
|
||
let content = result["messages"][0]["content"].as_array().unwrap();
|
||
assert_eq!(content.len(), 3);
|
||
assert_eq!(content[0]["text"], "A");
|
||
assert_eq!(content[1]["type"], "tool_use");
|
||
assert_eq!(content[2]["text"], "B");
|
||
}
|
||
}
|