teabot/internal/docker/docker.go

package docker

import (
	"bytes"
	"context"
	"errors"
	"fmt"
	"io"
	"os"
	"os/exec"
	"path/filepath"
	"time"
)

// jobScript is the entrypoint executed inside the container. It configures the
// git identity and credentials, clones the target repo, and runs Claude in
// non-interactive print mode reading the prompt from a mounted file. All inputs
// arrive via environment variables so the script itself is static.
const jobScript = `#!/usr/bin/env bash
set -euo pipefail

export HOME="${TEABOT_HOME}"
mkdir -p "$HOME/.config"

git config --global user.name "${TEABOT_GIT_NAME}"
git config --global user.email "${TEABOT_GIT_EMAIL}"
git config --global credential.helper store
git config --global init.defaultBranch main
umask 077
printf 'https://%s:%s@%s\n' "${TEABOT_GIT_USER}" "${TEABOT_TOKEN}" "${TEABOT_GIT_HOST}" > "$HOME/.git-credentials"

WORK="$HOME/work"
mkdir -p "$WORK"
cd "$WORK"
echo "teabot: cloning ${TEABOT_CLONE_URL}"
git clone --quiet "${TEABOT_CLONE_URL}" repo
cd repo

echo "teabot: starting claude session"
claude --print --dangerously-skip-permissions < /teabot/prompt.txt
`

// DockerRunner runs jobs with the local docker CLI.
type DockerRunner struct {
	// DockerPath is the docker binary (default "docker").
	DockerPath string
	// SELinuxLabel is the volume relabel suffix. On Fedora this must be "z"
	// (shared) or "Z" (private) so bind mounts are accessible under SELinux.
	SELinuxLabel string
	// WorkRoot is where per-job scratch directories are created
	// (default os.TempDir()).
	WorkRoot string
	// Stdout receives streamed container output (nil discards the stream; the
	// captured output is always returned in Result regardless).
	Stdout io.Writer
}

// NewDockerRunner builds a runner with sensible defaults for this host.
func NewDockerRunner() *DockerRunner {
	return &DockerRunner{DockerPath: "docker", SELinuxLabel: "z"}
}

// Run implements Runner.
func (r *DockerRunner) Run(ctx context.Context, job Job) (Result, error) {
	if job.Image == "" {
		return Result{}, errors.New("job image is empty")
	}
	jobDir, err := r.prepareJobDir(job)
	if err != nil {
		return Result{}, err
	}
	defer func() { _ = os.RemoveAll(jobDir) }()

	if job.Timeout > 0 {
		var cancel context.CancelFunc
		ctx, cancel = context.WithTimeout(ctx, job.Timeout)
		defer cancel()
	}

	args := r.buildArgs(job, jobDir)
	start := time.Now()
	cmd := exec.CommandContext(ctx, r.DockerPath, args...)
	var buf bytes.Buffer
	if r.Stdout != nil {
		cmd.Stdout = io.MultiWriter(&buf, r.Stdout)
		cmd.Stderr = io.MultiWriter(&buf, r.Stdout)
	} else {
		cmd.Stdout = &buf
		cmd.Stderr = &buf
	}

	runErr := cmd.Run()
	res := Result{Output: buf.String(), Duration: time.Since(start)}
	if ctx.Err() == context.DeadlineExceeded {
		res.TimedOut = true
		res.ExitCode = -1
		return res, fmt.Errorf("job %q timed out after %s", job.Label, job.Timeout)
	}
	var exitErr *exec.ExitError
	if errors.As(runErr, &exitErr) {
		res.ExitCode = exitErr.ExitCode()
		return res, nil
	}
	if runErr != nil {
		return res, fmt.Errorf("running docker: %w", runErr)
	}
	res.ExitCode = 0
	return res, nil
}

// prepareJobDir materialises the mounted files for a job: the prompt, the job
// script, a private copy of the Claude config (so token refreshes never touch
// the host's real config), and a copy of the personality's tea config.
func (r *DockerRunner) prepareJobDir(job Job) (string, error) {
	root := r.WorkRoot
	if root == "" {
		root = os.TempDir()
	}
	if err := os.MkdirAll(root, 0o700); err != nil {
		return "", err
	}
	jobDir, err := os.MkdirTemp(root, "teabot-job-")
	if err != nil {
		return "", err
	}
	if err := os.WriteFile(filepath.Join(jobDir, "prompt.txt"), []byte(job.Prompt), 0o600); err != nil {
		return "", err
	}
	if err := os.WriteFile(filepath.Join(jobDir, "job.sh"), []byte(jobScript), 0o700); err != nil {
		return "", err
	}
	// Copy the Claude config dir so the container can refresh subscription
	// tokens without mutating the host's ~/.claude.
	if job.ClaudeConfigDir != "" {
		if _, statErr := os.Stat(job.ClaudeConfigDir); statErr == nil {
			if err := copyTree(job.ClaudeConfigDir, filepath.Join(jobDir, "claude")); err != nil {
				return "", fmt.Errorf("copying claude config: %w", err)
			}
		}
	}
	// Copy the personality's tea config to the mounted XDG location.
	if job.TeaConfigPath != "" {
		teaDir := filepath.Join(jobDir, "tea")
		if err := os.MkdirAll(teaDir, 0o700); err != nil {
			return "", err
		}
		if err := copyFile(job.TeaConfigPath, filepath.Join(teaDir, "config.yml")); err != nil {
			return "", fmt.Errorf("copying tea config: %w", err)
		}
	}
	return jobDir, nil
}

// buildArgs assembles the full `docker run` argument list for a job. It is pure
// (given jobDir) so it can be unit-tested without invoking docker.
func (r *DockerRunner) buildArgs(job Job, jobDir string) []string {
	home := job.ContainerHome
	if home == "" {
		home = "/home/agent"
	}
	label := r.SELinuxLabel
	mount := func(host, container string, ro bool) string {
		spec := host + ":" + container
		if ro {
			spec += ":ro"
			if label != "" {
				spec += "," + label
			}
		} else if label != "" {
			spec += ":" + label
		}
		return spec
	}

	args := []string{"run", "--rm", "--entrypoint", "/bin/bash"}

	// Mount the job scratch (prompt + script) read-only.
	args = append(args, "-v", mount(filepath.Join(jobDir, "prompt.txt"), "/teabot/prompt.txt", true))
	args = append(args, "-v", mount(filepath.Join(jobDir, "job.sh"), "/teabot/job.sh", true))

	// Mount the private Claude config copy read-write (token refresh).
	if job.ClaudeConfigDir != "" {
		args = append(args, "-v", mount(filepath.Join(jobDir, "claude"), home+"/.claude", false))
	}
	// Mount the tea config read-only at the XDG path.
	if job.TeaConfigPath != "" {
		args = append(args, "-v", mount(filepath.Join(jobDir, "tea", "config.yml"), home+"/.config/tea/config.yml", true))
	}

	// Environment consumed by job.sh.
	env := map[string]string{
		"TEABOT_HOME":      home,
		"TEABOT_GIT_NAME":  job.GitName,
		"TEABOT_GIT_EMAIL": job.GitEmail,
		"TEABOT_GIT_USER":  job.GitUser,
		"TEABOT_TOKEN":     job.Token,
		"TEABOT_GIT_HOST":  job.GitHost,
		"TEABOT_CLONE_URL": job.CloneURL,
		"XDG_CONFIG_HOME":  home + "/.config",
	}
	if job.AnthropicAPIKey != "" {
		env["ANTHROPIC_API_KEY"] = job.AnthropicAPIKey
	}
	if job.AnthropicBaseURL != "" {
		env["ANTHROPIC_BASE_URL"] = job.AnthropicBaseURL
	}
	for _, k := range sortedKeys(env) {
		args = append(args, "-e", k+"="+env[k])
	}

	args = append(args, job.Image, "/teabot/job.sh")
	return args
}

// sortedKeys returns map keys in deterministic order (stable docker args ease
// testing and logging).
func sortedKeys(m map[string]string) []string {
	keys := make([]string, 0, len(m))
	for k := range m {
		keys = append(keys, k)
	}
	// simple insertion sort avoids importing sort for a tiny map
	for i := 1; i < len(keys); i++ {
		for j := i; j > 0 && keys[j-1] > keys[j]; j-- {
			keys[j-1], keys[j] = keys[j], keys[j-1]
		}
	}
	return keys
}

func copyFile(src, dst string) error {
	in, err := os.Open(src)
	if err != nil {
		return err
	}
	defer func() { _ = in.Close() }()
	if err := os.MkdirAll(filepath.Dir(dst), 0o700); err != nil {
		return err
	}
	out, err := os.OpenFile(dst, os.O_CREATE|os.O_WRONLY|os.O_TRUNC, 0o600)
	if err != nil {
		return err
	}
	if _, err := io.Copy(out, in); err != nil {
		_ = out.Close()
		return err
	}
	// Close explicitly (not deferred) so a flush error surfaces to the caller.
	return out.Close()
}

// copyTree recursively copies a directory tree (regular files, dirs, and
// symlink targets are dereferenced by copyFile via Open).
func copyTree(src, dst string) error {
	return filepath.Walk(src, func(path string, info os.FileInfo, err error) error {
		if err != nil {
			return err
		}
		rel, err := filepath.Rel(src, path)
		if err != nil {
			return err
		}
		target := filepath.Join(dst, rel)
		if info.IsDir() {
			return os.MkdirAll(target, 0o700)
		}
		if !info.Mode().IsRegular() {
			return nil // skip sockets/devices; symlinks are followed by Walk's lstat -> handle below
		}
		return copyFile(path, target)
	})
}