Files
Agent-System-Backend/internal/tools/exec_interpreter.go
T
weichen 4df81ef1da
ci / vet-and-test (pull_request) Successful in 43s
Enable parallel tool execution, fix delegate truncation, sandbox deps layer
Why: Real runs were fully serial: the direct path never set the kernel's
parallel tool mode, delegate_task handed the parent only the child's
160-char summary (the full text was dropped), MiniMax streamed its
reasoning inside content <think> blocks, and the EXT-16 sandbox hid
~/.local so python-docx/openpyxl/pptx could not even import. All four
were measured defects (F1/F2/F3/F5) in the delegation/routing proposal.

Why: Serial-only mutual exclusion is not enough for order-sensitive
tools - the card-stream test failed with out-of-order appends - so the
ordering moved into the kernel (SerialTools partition, Agent-Native
#44); process events pair by CallID (#42) and tool results write back
through the injected call context (#43).

Evidence: go test ./... green, including new regression anchors:
TestDelegateToolReturnsFullText, TestToolEventsPairByCallID,
TestToolExecModeAndSerialSet, TestOpenLLMClientReasoningSplitAuto,
TestSandboxArgvDepsLayer, TestExecEnvJoinsPythonPath.

Impact: Direct runs now execute independent same-turn calls
(searches, reads, delegate_task, review_findings) concurrently by
default (runtime.parallelTools=false restores sequential); card flow,
plans, workflows, exec primitives and file edits keep model-declared
order. delegate_task returns the capped full child text. MiniMax-*
models get reasoning_split=true so thinking streams via
reasoning_content. --exec-deps-dir/AGENT_SYSTEM_EXEC_DEPS_DIR mounts a
read-only dependency layer into the sandbox and merges PYTHONPATH.
Bumps deps/agent-native to c348441.

Refs: proposal 2026-10-10-agent-delegation-routing-sandbox.md (E1,
F1-F5); Agent-Native #42/#43/#44; Dev task feature/agent-e1-parallel-fixes.
2026-10-10 09:02:10 +00:00

340 lines
11 KiB
Go
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.
// run_shell / run_python 解释器执行器(与 register_artifact 同族)。
//
// 语义:模型提交的脚本原样落盘到会话工作区 .agent/ 下(字节精确、执行后保留以便审计),
// 再以该文件为唯一参数启动白名单解释器(bash / python3 的绝对路径)——模型输入不进入
// 进程参数。工作目录锁定会话沙箱;超时取请求 context 与 timeout_sec(默认 600s,无硬上限)
// 的先到者,超时杀掉整个进程组;输出 64KB 头尾截断。执行中的输出经可选回调流出。
// 非零退出、超时与取消作为文本反馈返回
// (模型据此自纠),仅启动失败算工具错误。
package tools
import (
"bytes"
"context"
"encoding/json"
"errors"
"fmt"
"io"
"os"
"os/exec"
"path/filepath"
"strings"
"syscall"
"time"
core "agent-system-core"
agenttools "agent-native/runtime/tools"
)
// RegisterInterpreterTools 注册解释器执行原语(run_shell / run_python)。
func RegisterInterpreterTools(r *Registry) error {
if err := r.Register(core.ToolDesc{
Name: "run_shell", Source: core.SourceBuiltin, Risk: "medium",
Description: "在会话工作区目录内执行 bash 脚本:检查环境、安装依赖(pip install --user …)、处理文件等。脚本原样落盘执行;返回退出码与合并输出(stdout+stderr,超限头尾截断)。以当前用户身份运行。",
Parameters: json.RawMessage(`{"type":"object","properties":{` +
`"script":{"type":"string","description":"bash 脚本(可多行)"},` +
`"timeout_sec":{"type":"integer","description":"超时秒数(默认 600,可自行设更长)"}}` +
`,"required":["script"]}`),
}, execShell); err != nil {
return err
}
return r.Register(core.ToolDesc{
Name: "run_python", Source: core.SourceBuiltin, Risk: "medium",
Description: "执行 Python 脚本(python3,工作目录为会话工作区):生成文档/表格/演示稿/数据处理的首选路径。脚本原样落盘执行;stdout/stderr 合并返回,超限头尾截断。python-docx/openpyxl/python-pptx 缺失时可先 pip install --user 安装。交付物用 import housestyle(技能目录在环境变量 AGENT_SKILLS_DIR,house-style 已在 PYTHONPATH)。",
Parameters: json.RawMessage(`{"type":"object","properties":{` +
`"script":{"type":"string","description":"完整 Python 脚本(utf-8)"},` +
`"timeout_sec":{"type":"integer","description":"超时秒数(默认 600,可自行设更长)"}}` +
`,"required":["script"]}`),
}, execPython)
}
// skillsRoot 由组合根在启动时注入(<data>/skills)。空 = 不注入技能路径。
var skillsRoot string
// SetSkillsRoot 让 run_python / run_shell 能 import housestyle,并看到技能目录。
func SetSkillsRoot(dir string) { skillsRoot = dir }
// execDepsDir 是 sandbox 档的只读依赖层(site-packages 等);组合根启动时注入。
var execDepsDir string
// SetExecDepsDir 配置 sandbox 档依赖层目录(空 = 不挂载,现状)。
func SetExecDepsDir(dir string) { execDepsDir = dir }
// execEnv 最小环境:PATH/HOME/LANG/TZ/TERM 继承(pip --user、字体、中文 locale),
// 其余变量不带入子进程。技能根配置后追加 AGENT_SKILLS_DIR 与 PYTHONPATH。
func execEnv() []string {
var out []string
for _, kv := range os.Environ() {
for _, keep := range []string{"PATH=", "HOME=", "LANG=", "LC_", "TZ=", "TERM="} {
if strings.HasPrefix(kv, keep) {
out = append(out, kv)
break
}
}
}
out = append(out, "PYTHONDONTWRITEBYTECODE=1")
var pypaths []string
if execDepsDir != "" {
pypaths = append(pypaths, execDepsDir)
}
if skillsRoot != "" {
out = append(out, "AGENT_SKILLS_DIR="+skillsRoot)
pypaths = append(pypaths, filepath.Join(skillsRoot, "house-style"))
}
if len(pypaths) > 0 { // 合成一条,避免重复键互相覆盖
out = append(out, "PYTHONPATH="+strings.Join(pypaths, string(os.PathListSeparator)))
}
return out
}
// truncateOutput 头 2/3 + 尾 1/3 保留,中部标记省略。
func truncateOutput(b []byte) string {
if len(b) <= execMaxOutput {
return string(b)
}
head := execMaxOutput * 2 / 3
tail := execMaxOutput / 3
return string(b[:head]) + fmt.Sprintf("\n…(输出超限,省略 %d 字节)…\n", len(b)-head-tail) + string(b[len(b)-tail:])
}
// writeScript 把模型脚本落盘到工作区 .agent/ 下(字节精确,执行后保留),返回绝对路径。
func writeScript(rc ToolExecCtx, sandbox, prefix, ext, content string) (string, error) {
dir := filepath.Join(sandbox, execScriptDir)
if err := os.MkdirAll(dir, 0o755); err != nil {
return "", err
}
p := filepath.Join(dir, fmt.Sprintf("%s-%s-%d%s", prefix, rc.ConversationID(), time.Now().UnixNano(), ext))
if err := os.WriteFile(p, []byte(content), 0o600); err != nil {
return "", err
}
return p, nil
}
type execContexter interface {
Context() context.Context
}
func execParentCtx(rc ToolExecCtx) context.Context {
if c, ok := rc.(execContexter); ok {
if ctx := c.Context(); ctx != nil {
return ctx
}
}
return context.Background()
}
// resolveInterpreter 把白名单解释器名钉成绝对路径,避免 PATH 在子进程里被换掉。
func resolveInterpreter(name string) (string, error) {
switch name {
case "bash", "python3":
default:
return "", fmt.Errorf("interpreter %s is not allowlisted", name)
}
p, err := exec.LookPath(name)
if err != nil {
return "", err
}
if !filepath.IsAbs(p) {
return "", fmt.Errorf("%s: resolved path is not absolute: %s", name, p)
}
return p, nil
}
// runInterpreter 启动白名单解释器执行落盘脚本(脚本路径为唯一参数)。
// 进程启动经 EXT-16 执行后端(direct=现状直启 / sandbox=bwrap 隔离档)。
// 非零退出、超时与请求取消都是正常反馈,以文本返回;仅启动失败算错误。
func runInterpreter(parent context.Context, interpreter, scriptPath, dir string, timeoutS int, sink func(stream, text string)) (string, error) {
if parent == nil {
parent = context.Background()
}
if timeoutS <= 0 {
timeoutS = execDefaultTimeoutS
}
bin, err := resolveInterpreter(interpreter)
if err != nil {
return "", fmt.Errorf("%s: %w", interpreter, err)
}
ctx, cancel := context.WithTimeout(parent, time.Duration(timeoutS)*time.Second)
defer cancel()
spec := ExecSpec{
Bin: bin, Args: []string{scriptPath}, Dir: dir, Env: execEnv(),
SkillsDir: skillsRoot, DepsDir: execDepsDir, Timeout: time.Duration(timeoutS) * time.Second,
}
pr, pw := io.Pipe()
var buf bytes.Buffer
readDone := make(chan struct{})
go func() {
defer close(readDone)
tmp := make([]byte, 4096)
for {
n, rerr := pr.Read(tmp)
if n > 0 {
if buf.Len() < 1<<20 {
buf.Write(tmp[:n])
}
if sink != nil {
sink("stdout", string(tmp[:n]))
}
}
if rerr != nil {
return
}
}
}()
proc, err := currentExecBackend().Start(ctx, spec, pw)
if err != nil {
_ = pw.Close()
<-readDone
return "", fmt.Errorf("%s: %w", interpreter, err)
}
err = proc.Wait()
_ = pw.Close()
<-readDone
text := truncateOutput(buf.Bytes())
if errors.Is(ctx.Err(), context.DeadlineExceeded) {
return fmt.Sprintf("exit=timeout(%ds)\n%s", timeoutS, text), nil
}
if errors.Is(parent.Err(), context.Canceled) {
return fmt.Sprintf("exit=canceled\n%s", text), nil
}
if err != nil {
var ee *exec.ExitError
if errors.As(err, &ee) {
return fmt.Sprintf("exit=%d\n%s", ee.ExitCode(), text), nil
}
return "", fmt.Errorf("%s: %v:%s", interpreter, err, text)
}
return "exit=0\n" + text, nil
}
func execShell(rc ToolExecCtx, args json.RawMessage) (string, error) {
if err := execGate(rc); err != nil {
return "", err
}
var a struct {
Script string `json:"script"`
Command string `json:"command"`
TimeoutS int `json:"timeout_sec"`
TimeoutSec int `json:"timeoutSec"`
}
if len(args) > 0 {
if err := json.Unmarshal(args, &a); err != nil {
return "", fmt.Errorf("run_shell: bad args: %w", err)
}
}
script := a.Script
if strings.TrimSpace(script) == "" {
script = a.Command
}
if strings.TrimSpace(script) == "" {
return "", fmt.Errorf("run_shell: script required")
}
sandbox, err := execSandbox(rc)
if err != nil {
return "", err
}
p, err := writeScript(rc, sandbox, "sh", ".sh", script)
if err != nil {
return "", err
}
return runInterpreter(execParentCtx(rc), "bash", p, sandbox, firstPos(a.TimeoutS, a.TimeoutSec), toolOutputSink(rc))
}
// ScriptCommand 准备一条以当前用户身份、工作目录为会话沙箱的 python3 命令。
// 调用方在 Start 之前设置 ExtraFiles、Stdout 和 Stderr。取消 ctx 会杀掉整个进程组。
func ScriptCommand(ctx context.Context, sandbox, path string) (*exec.Cmd, error) {
if ctx == nil {
ctx = context.Background()
}
bin, err := resolveInterpreter("python3")
if err != nil {
return nil, err
}
cmd := exec.CommandContext(ctx, bin, path)
cmd.Dir = sandbox
cmd.Env = execEnv()
cmd.SysProcAttr = &syscall.SysProcAttr{Setpgid: true}
cmd.WaitDelay = 2 * time.Second
cmd.Cancel = func() error {
if cmd.Process == nil {
return os.ErrProcessDone
}
err := syscall.Kill(-cmd.Process.Pid, syscall.SIGKILL)
if errors.Is(err, syscall.ESRCH) {
return nil
}
return err
}
return cmd, nil
}
func execPython(rc ToolExecCtx, args json.RawMessage) (string, error) {
if err := execGate(rc); err != nil {
return "", err
}
var a struct {
Script string `json:"script"`
TimeoutS int `json:"timeout_sec"`
TimeoutSec int `json:"timeoutSec"`
}
if len(args) > 0 {
if err := json.Unmarshal(args, &a); err != nil {
return "", fmt.Errorf("run_python: bad args: %w", err)
}
}
if strings.TrimSpace(a.Script) == "" {
return "", fmt.Errorf("run_python: script required")
}
sandbox, err := execSandbox(rc)
if err != nil {
return "", err
}
p, err := writeScript(rc, sandbox, "py", ".py", a.Script)
if err != nil {
return "", err
}
return runInterpreter(execParentCtx(rc), "python3", p, sandbox, firstPos(a.TimeoutS, a.TimeoutSec), toolOutputSink(rc))
}
// toolOutputSink 在执行上下文实现了 EmitToolOutput 时,把 stdout 增量交出去。
func toolOutputSink(rc ToolExecCtx) func(stream, text string) {
type emitter interface {
EmitToolOutput(stream, text string)
}
e, ok := rc.(emitter)
if !ok {
return nil
}
return e.EmitToolOutput
}
func firstPos(vals ...int) int {
for _, v := range vals {
if v > 0 {
return v
}
}
return 0
}
// InterpreterExecConfig 把产品解释器白名单换算为 agent-native 内核的
// ExecConfig(squad 角色经 agent-native 后端执行时复用同一套约束与环境)。
// 两个解释器都缺席或启动级硬关闭时返回 nil(内核侧不注册执行工具,fail-closed)。
func InterpreterExecConfig() *agenttools.ExecConfig {
if ExecDisabled() {
return nil
}
cfg := &agenttools.ExecConfig{Env: execEnv()}
if p, err := exec.LookPath("bash"); err == nil {
cfg.Bash = p
}
if p, err := exec.LookPath("python3"); err == nil {
cfg.Python = p
}
if cfg.Bash == "" && cfg.Python == "" {
return nil
}
return cfg
}