AstroResearch/src/agent/runtime/executor/helpers.rs
Asfmq eaf85707b5 refactor: 全栈架构重构与质量硬化
核心架构重构:
- Config 按职责拆分为 LlmConfig/EmbeddingConfig/VisionConfig/CdsConfig/StorageConfig 五个子结构
- AppState 拆分为 LlmState/DataSourceState/SessionState 三个子结构,消除 50+ 平铺字段
- 新增 ServiceError 结构化错误类型替代 handler 中的 msg.contains() 字符串匹配
- 移除 api::handlers 兼容命名空间,路由直接引用 agent/auth/papers 等模块

认证性能优化:
- login_rate_limiter/upload_rate_limiter 从 Mutex<HashMap> 迁移为 DashMap(无锁)
- 新增 session_last_active: DashMap<String, AtomicU64>,auth 中间件快速路径免写锁
- 会话过期清理改为按间隔触发(300s),避免每次请求全表扫描
- MAX_SESSIONS 1000→10000,SSE 广播通道 256→1024

Agent 工具增强:
- AgentTool trait 新增 is_internal()/display_name(),SSE 事件携带工具元数据
- 新增 GET /chat/tools 端点暴露注册工具列表
- pending_questions/pending_permissions 增加 created_at 时间戳,自动清理过期条目(10min TTL)
- Agent 超时现在正确 abort 后台任务并设置取消令牌

观测数据源修复:
- FITS 解析: APOGEE/DESI 改用 read_image+切片替代 read_rows(修复 fitsio panic)
- ZTF: CIRCLE 参数分隔符 +→空格(修复 IRSA 400),半径自动裁剪至硬上限
- MAST TESS: parse_tic_json 兼容数组/对象两种 API 响应格式
- 统一检索: per_target_limit 默认 50→1,sources 支持 per-source release/version
- Gaia 测光从 VizieR 镜像切换至官方 TAP 服务

RAG 并发优化:
- 向量化降级从串行改为并发 5 条/批(buffer_unordered)
- 混合检索 RRF 合并从借用改为 owned RetrievalResult

安全加固:
- PDF 中间件: URL 解码 %2F/%2E 后判扩展名;文件名过滤非 ASCII + 禁 \ 防头注入
- chat_agent 日志截断问题内容至 50 字符;list_sessions 强制 limit clamp

前端双主题:
- 设计令牌三层架构: primitive→semantic→component,浅色暖纸张学术/暗色 Night Indigo
- useTheme hook + ThemeToggle 侧边栏组件 + main.tsx 防 FOUC 初始化
- 全组件从硬编码 slate 色迁移至语义令牌(bg-surface/text-content/border-subtle 等)
- 新增 ToastContainer 非阻塞通知系统

部署优化:
- deploy.sh 引入 SSH ControlMaster 单次密码复用
2026-07-11 14:57:40 +08:00

268 lines
8.7 KiB
Rust
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

// src/agent/runtime/executor/helpers.rs
//
// 执行器辅助类型与函数PreparedCall, ToolResultMessage, ToolExecutionResult,
// execute_single_tool, process_single_result, save_tool_message_sync。
use sqlx::SqlitePool;
use std::sync::atomic::{AtomicBool, Ordering};
use std::sync::Arc;
use tokio::sync::mpsc;
use tracing::warn;
use crate::clients::llm::ChatMessage;
use super::AgentStreamEvent;
use crate::agent::hooks::{
event_label, HookRegistry, PostToolUseContext, PostToolUseFailureContext,
};
#[derive(Debug, Clone)]
pub struct PreparedCall {
pub tool_call_id: String,
pub tool_name: String,
pub args: serde_json::Value,
}
/// 单次工具执行后的消息 + 元数据
pub struct ToolResultMessage {
pub chat_message: ChatMessage,
pub was_error: bool,
}
/// 工具执行结果摘要
pub struct ToolExecutionResult {
/// 每条工具调用对应的 tool_result 消息(供调用方 push 到 messages
pub tool_messages: Vec<ToolResultMessage>,
pub was_cancelled: bool,
pub had_duplicate: bool,
/// Hook 注入的附加上下文PreToolUse + PostToolUse需注入 LLM 消息列表
pub hook_contexts: Vec<String>,
/// Hook 的阻塞错误详情(用于日志和诊断)
pub blocking_errors: Vec<String>,
}
/// 执行单个工具调用(含超时和取消检测)。
///
/// 从原 `execute_parallel` 的闭包提取,供分区后的批次执行复用。
pub(super) async fn execute_single_tool(
tool_opt: Option<&dyn crate::agent::tools::AgentTool>,
args: serde_json::Value,
tool_ctx: &crate::agent::tools::ToolContext,
cancelled: &Arc<AtomicBool>,
timeout_dur: std::time::Duration,
tool_name: &str,
) -> crate::agent::tools::ToolOutput {
let tool = match tool_opt {
Some(t) => t,
None => return crate::agent::tools::ToolOutput::error(format!("未知工具: {}", tool_name)),
};
let interrupt_behavior = tool.interrupt_behavior();
let is_blocking = interrupt_behavior == crate::agent::tools::InterruptBehavior::Block;
let tool_fut = tool.execute(args, tool_ctx);
let cancelled = cancelled.clone();
let cancel_fut = async {
loop {
tokio::time::sleep(std::time::Duration::from_millis(250)).await;
if !is_blocking && cancelled.load(Ordering::SeqCst) {
return;
}
}
};
tokio::select! {
res = tokio::time::timeout(timeout_dur, tool_fut) => {
match res {
Ok(output) => output,
Err(_) => crate::agent::tools::ToolOutput::error(format!(
"工具 {} 执行超时({}秒)",
tool_name,
timeout_dur.as_secs()
)),
}
}
_ = cancel_fut => {
crate::agent::tools::ToolOutput::error("执行已被用户取消")
}
}
}
/// 处理单个工具执行结果SSE 事件、PostToolUse hooks、持久化
///
/// 从原 `execute_parallel` 的结果处理循环提取。
#[allow(clippy::too_many_arguments)]
pub(super) async fn process_single_result(
tool_call_id: &str,
tool_name: &str,
tool_args: &serde_json::Value,
output: &crate::agent::tools::ToolOutput,
cancelled_flag: bool,
exec_start: std::time::Instant,
tx: &mpsc::UnboundedSender<AgentStreamEvent>,
hook_registry: &HookRegistry,
library_dir: &std::path::Path,
sid: &str,
agent_name: &str,
step: usize,
max_output_chars: usize,
tool_messages: &mut Vec<ToolResultMessage>,
additional_contexts: &mut Vec<String>,
db: &SqlitePool,
turn_index: i32,
tool_registry: &crate::agent::tools::ToolRegistry,
) {
use crate::agent::tools::persist::maybe_persist_tool_result;
let elapsed_ms = exec_start.elapsed().as_millis() as u64;
let (is_internal, display_name) = if let Some(tool) = tool_registry.get(tool_name) {
(tool.is_internal(), tool.display_name().to_string())
} else {
(false, tool_name.to_string())
};
// SSE 事件 — 立即推送到前端
let _ = tx.send(AgentStreamEvent::ToolResult {
tool_call_id: tool_call_id.to_string(),
name: tool_name.to_string(),
display_name,
output: output.content.clone(),
is_error: output.is_error,
metadata: output.metadata.clone(),
step,
is_internal,
});
// 输出处理:小结果直接传递,大结果持久化到磁盘并返回 stub
// 但对于已从磁盘读取内容的工具(如 read_file跳过持久化以防止级联
let tool_results_dir = library_dir.join(".agent").join("tool-results");
let (processed_content, _persisted_path) = if output.skip_persist {
(output.content.clone(), None)
} else {
maybe_persist_tool_result(
&output.content,
tool_call_id,
max_output_chars,
&tool_results_dir,
)
.await
};
// PostToolUse hook
let post_ctx = PostToolUseContext {
session_id: sid.to_string(),
agent_name: agent_name.to_string(),
tool_name: tool_name.to_string(),
tool_args: tool_args.clone(),
output_content: processed_content.clone(),
is_error: output.is_error,
step,
elapsed_ms,
};
let post_result = hook_registry.run_post_tool_use(&post_ctx).await;
let final_content = post_result.final_content;
// 非可信内容包裹(间接 prompt 注入防御)
let llm_content =
crate::agent::runtime::untrusted::wrap_untrusted_content(tool_name, &final_content);
// 收集 PostToolUse hook 注入的上下文
if !post_result.tagged_contexts.is_empty() {
for tc in &post_result.tagged_contexts {
additional_contexts.push(format!(
"[Hook: {} | {}] {}",
tc.hook_name,
event_label(tc.source_event),
tc.content,
));
}
} else {
for ctx in &post_result.additional_contexts {
additional_contexts.push(ctx.clone());
}
}
// 收集 PostToolUse 的警告
for warning in &post_result.warnings {
additional_contexts.push(format!("[Hook Warning] {}", warning));
}
// 事后权限请求audit trail
for (perm_tool, perm) in &post_result.post_permission_requests {
warn!(
"[Executor] Hook 事后请求权限: tool={} permission={}",
perm_tool, perm
);
}
// PostToolUseFailure hook
if output.is_error {
let failure_ctx = PostToolUseFailureContext {
session_id: sid.to_string(),
agent_name: agent_name.to_string(),
tool_name: tool_name.to_string(),
tool_args: tool_args.clone(),
error_message: output.content.clone(),
is_interrupt: cancelled_flag,
step,
elapsed_ms,
};
hook_registry
.run_on_post_tool_use_failure(&failure_ctx)
.await;
}
// 发送给 LLM 使用包裹后的内容(安全防御)
let chat_message = ChatMessage::tool_result(tool_call_id, &llm_content);
// 持久化到数据库fire-and-forget
save_tool_message_sync(db, sid, turn_index, step, &chat_message);
tool_messages.push(ToolResultMessage {
chat_message,
was_error: output.is_error,
});
}
/// 同步保存 tool 角色消息到数据库。
pub(super) fn save_tool_message_sync(
db: &SqlitePool,
session_id: &str,
turn_index: i32,
step_index: usize,
msg: &ChatMessage,
) {
let db_clone = db.clone();
let session_id = session_id.to_string();
let content = msg.text().unwrap_or("").to_string();
let tool_call_id = msg.tool_call_id.clone();
// 提前序列化,避免闭包内的生命周期问题
let metadata_str =
serde_json::to_string(&serde_json::json!({ "role": "tool" })).unwrap_or_default();
let raw_json = serde_json::to_string(&msg).unwrap_or_default();
// fire-and-forget: tool 消息保存失败不影响主流程
tokio::spawn(async move {
let token_count = content.len() as i32 / 4;
if let Err(e) = sqlx::query(
"INSERT INTO agent_messages (session_id, turn_index, step_index, role, content, tool_call_id, token_count, metadata, raw_json, agent_name) \
VALUES (?, ?, ?, 'tool', ?, ?, ?, ?, ?, ?)",
)
.bind(&session_id)
.bind(turn_index)
.bind(step_index as i32)
.bind(&content)
.bind(&tool_call_id)
.bind(token_count)
.bind(&metadata_str)
.bind(&raw_json)
.bind("lead")
.execute(&db_clone)
.await
{
warn!("[Executor] 保存 tool 消息失败(非致命): {}", e);
}
});
}