Fix: view_img, exec commands

This commit is contained in:
Grail Finder
2026-03-14 16:34:58 +03:00
parent fdcaa6c5e2
commit 26377702d3
3 changed files with 165 additions and 19 deletions

View File

@@ -518,9 +518,18 @@ func runCmd(args map[string]string) []byte {
case "capture_and_view", "screenshot_and_view":
// capture and view screenshot
return captureWindowAndView(args)
case "view_img":
// view_img <file> - view image for multimodal
return []byte(tools.FsViewImg(rest, ""))
case "browser":
// browser <action> [args...] - Playwright browser automation
return runBrowserCommand(rest, args)
case "mkdir", "ls", "cat", "pwd", "cd", "cp", "mv", "rm", "sed", "grep", "head", "tail", "wc", "sort", "uniq", "echo", "time", "stat", "go", "find", "file":
// File operations and shell commands - use ExecChain which has whitelist
return executeCommand(args)
case "git":
// git has its own whitelist in FsGit
return []byte(tools.FsGit(rest, ""))
default:
// Unknown subcommand - return help to remind user of available commands
return []byte(getHelp(nil))
@@ -667,13 +676,13 @@ func getHelp(args []string) string {
# File operations
ls [path] - list files in directory
cat <file> - read file content
see <file> - view image file
view_img <file> - view image file
write <file> - write content to file
stat <file> - get file info
rm <file> - delete file
cp <src> <dst> - copy file
mv <src> <dst> - move/rename file
mkdir <dir> - create directory
mkdir [-p] <dir> - create directory (use full path)
pwd - print working directory
cd <dir> - change directory
sed 's/old/new/[g]' [file] - text replacement
@@ -747,12 +756,12 @@ Use: run "command" to execute.`
Examples:
run "cat readme.md"
run "cat -b image.png" (base64 output)`
case "see":
return `see <image-file>
case "view_img":
return `view_img <image-file>
View an image file for multimodal analysis.
Supports: png, jpg, jpeg, gif, webp, svg
Example:
run "see screenshot.png"`
run "view_img screenshot.png"`
case "write":
return `write <file> [content]
Write content to a file.
@@ -812,6 +821,14 @@ Use: run "command" to execute.`
Print working directory.
Example:
run "pwd"`
case "mkdir":
return `mkdir [-p] <directory>
Create a directory (use full path).
Options:
-p, --parents create parent directories as needed
Examples:
run "mkdir /full/path/myfolder"
run "mkdir -p /full/path/to/nested/folder"`
case "sed":
return `sed 's/old/new/[g]' [file]
Stream editor for text replacement.

View File

@@ -239,7 +239,12 @@ func execBuiltin(name string, args []string, stdin string) string {
}
return ""
}
data, err := os.ReadFile(args[0])
path := args[0]
abs := path
if !filepath.IsAbs(path) {
abs = filepath.Join(fsRootDir, path)
}
data, err := os.ReadFile(abs)
if err != nil {
return fmt.Sprintf("[error] cat: %v", err)
}
@@ -266,6 +271,76 @@ func execBuiltin(name string, args []string, stdin string) string {
}
fsRootDir = abs
return fmt.Sprintf("Changed directory to: %s", fsRootDir)
case "mkdir":
if len(args) == 0 {
return "[error] usage: mkdir [-p] <dir>"
}
createParents := false
var dirPath string
for _, a := range args {
if a == "-p" || a == "--parents" {
createParents = true
} else if dirPath == "" {
dirPath = a
}
}
if dirPath == "" {
return "[error] usage: mkdir [-p] <dir>"
}
abs := dirPath
if !filepath.IsAbs(dirPath) {
abs = filepath.Join(fsRootDir, dirPath)
}
abs = filepath.Clean(abs)
var mkdirFunc func(string, os.FileMode) error
if createParents {
mkdirFunc = os.MkdirAll
} else {
mkdirFunc = os.Mkdir
}
if err := mkdirFunc(abs, 0o755); err != nil {
return fmt.Sprintf("[error] mkdir: %v", err)
}
if createParents {
return fmt.Sprintf("Created %s (with parents)", dirPath)
}
return fmt.Sprintf("Created %s", dirPath)
case "ls":
dir := "."
for _, a := range args {
if !strings.HasPrefix(a, "-") {
dir = a
break
}
}
abs := dir
if !filepath.IsAbs(dir) {
abs = filepath.Join(fsRootDir, dir)
}
entries, err := os.ReadDir(abs)
if err != nil {
return fmt.Sprintf("[error] ls: %v", err)
}
var out strings.Builder
for _, e := range entries {
info, _ := e.Info()
if e.IsDir() {
fmt.Fprintf(&out, "d %-8s %s/\n", "-", e.Name())
} else if info != nil {
size := info.Size()
sizeStr := fmt.Sprintf("%d", size)
if size > 1024 {
sizeStr = fmt.Sprintf("%.1fKB", float64(size)/1024)
}
fmt.Fprintf(&out, "f %-8s %s\n", sizeStr, e.Name())
} else {
fmt.Fprintf(&out, "f %-8s %s\n", "?", e.Name())
}
}
if out.Len() == 0 {
return "(empty directory)"
}
return strings.TrimRight(out.String(), "\n")
case "go":
// Allow all go subcommands
if len(args) == 0 {

View File

@@ -2,7 +2,9 @@ package tools
import (
"encoding/base64"
"encoding/json"
"fmt"
"gf-lt/models"
"os"
"os/exec"
"path/filepath"
@@ -155,27 +157,53 @@ func FsCat(args []string, stdin string) string {
return string(data)
}
func FsSee(args []string, stdin string) string {
func FsViewImg(args []string, stdin string) string {
if len(args) == 0 {
return "[error] usage: see <image-path>"
return "[error] usage: view_img <image-path>"
}
path := args[0]
abs, err := resolvePath(path)
var abs string
if filepath.IsAbs(path) {
abs = path
} else {
var err error
abs, err = resolvePath(path)
if err != nil {
return fmt.Sprintf("[error] %v", err)
}
}
info, err := os.Stat(abs)
if err != nil {
return fmt.Sprintf("[error] see: %v", err)
if _, err := os.Stat(abs); err != nil {
return fmt.Sprintf("[error] view_img: %v", err)
}
if !IsImageFile(path) {
return fmt.Sprintf("[error] not an image file: %s (use cat to read text files)", path)
}
return fmt.Sprintf("Image: %s (%s)\n![image](file://%s)", path, humanSize(info.Size()), abs)
dataURL, err := models.CreateImageURLFromPath(abs)
if err != nil {
return fmt.Sprintf("[error] view_img: %v", err)
}
result := models.MultimodalToolResp{
Type: "multimodal_content",
Parts: []map[string]string{
{"type": "text", "text": "Image: " + path},
{"type": "image_url", "url": dataURL},
},
}
jsonResult, err := json.Marshal(result)
if err != nil {
return fmt.Sprintf("[error] view_img: %v", err)
}
return string(jsonResult)
}
// FsSee is deprecated, use FsViewImg
func FsSee(args []string, stdin string) string {
return FsViewImg(args, stdin)
}
func FsWrite(args []string, stdin string) string {
@@ -352,18 +380,44 @@ func FsMv(args []string, stdin string) string {
func FsMkdir(args []string, stdin string) string {
if len(args) == 0 {
return "[error] usage: mkdir <dir>"
return "[error] usage: mkdir [-p] <dir>"
}
abs, err := resolvePath(args[0])
createParents := false
var dirPath string
for _, a := range args {
if a == "-p" || a == "--parents" {
createParents = true
} else if dirPath == "" {
dirPath = a
}
}
if dirPath == "" {
return "[error] usage: mkdir [-p] <dir>"
}
abs, err := resolvePath(dirPath)
if err != nil {
return fmt.Sprintf("[error] %v", err)
}
if err := os.MkdirAll(abs, 0o755); err != nil {
var mkdirFunc func(string, os.FileMode) error
if createParents {
mkdirFunc = os.MkdirAll
} else {
mkdirFunc = os.Mkdir
}
if err := mkdirFunc(abs, 0o755); err != nil {
return fmt.Sprintf("[error] mkdir: %v", err)
}
return fmt.Sprintf("Created %s", args[0])
if createParents {
return fmt.Sprintf("Created %s (with parents)", dirPath)
}
return fmt.Sprintf("Created %s", dirPath)
}
// Text processing commands