summaryrefslogtreecommitdiff
path: root/tools
diff options
context:
space:
mode:
Diffstat (limited to 'tools')
-rw-r--r--tools/chain.go42
-rw-r--r--tools/fs.go82
-rw-r--r--tools/pw.go83
-rw-r--r--tools/tools.go1201
4 files changed, 558 insertions, 850 deletions
diff --git a/tools/chain.go b/tools/chain.go
index 4afb7e5..b94d2f8 100644
--- a/tools/chain.go
+++ b/tools/chain.go
@@ -32,10 +32,8 @@ func ParseChain(input string) []Segment {
var current strings.Builder
runes := []rune(input)
n := len(runes)
-
for i := 0; i < n; i++ {
ch := runes[i]
-
// handle quotes
if ch == '\'' || ch == '"' {
quote := ch
@@ -50,7 +48,6 @@ func ParseChain(input string) []Segment {
}
continue
}
-
// &&
if ch == '&' && i+1 < n && runes[i+1] == '&' {
segments = append(segments, Segment{
@@ -61,7 +58,6 @@ func ParseChain(input string) []Segment {
i++ // skip second &
continue
}
-
// ;
if ch == ';' {
segments = append(segments, Segment{
@@ -71,7 +67,6 @@ func ParseChain(input string) []Segment {
current.Reset()
continue
}
-
// ||
if ch == '|' && i+1 < n && runes[i+1] == '|' {
segments = append(segments, Segment{
@@ -82,7 +77,6 @@ func ParseChain(input string) []Segment {
i++ // skip second |
continue
}
-
// | (single pipe)
if ch == '|' {
segments = append(segments, Segment{
@@ -92,16 +86,13 @@ func ParseChain(input string) []Segment {
current.Reset()
continue
}
-
current.WriteRune(ch)
}
-
// last segment
last := strings.TrimSpace(current.String())
if last != "" {
segments = append(segments, Segment{Raw: last, Op: OpNone})
}
-
return segments
}
@@ -112,12 +103,10 @@ func ExecChain(command string) string {
if len(segments) == 0 {
return "[error] empty command"
}
-
var collected []string
var lastOutput string
var lastErr error
pipeInput := ""
-
for i, seg := range segments {
if i > 0 {
prevOp := segments[i-1].Op
@@ -130,7 +119,6 @@ func ExecChain(command string) string {
continue
}
}
-
// determine stdin for this segment
segStdin := ""
if i == 0 {
@@ -138,9 +126,7 @@ func ExecChain(command string) string {
} else if segments[i-1].Op == OpPipe {
segStdin = lastOutput
}
-
lastOutput, lastErr = execSingle(seg.Raw, segStdin)
-
// pipe: output flows to next command's stdin
// && or ;: collect output
if i < len(segments)-1 && seg.Op == OpPipe {
@@ -150,7 +136,6 @@ func ExecChain(command string) string {
collected = append(collected, lastOutput)
}
}
-
return strings.Join(collected, "\n")
}
@@ -160,15 +145,12 @@ func execSingle(command, stdin string) (string, error) {
if len(parts) == 0 {
return "", fmt.Errorf("empty command")
}
-
name := parts[0]
args := parts[1:]
-
// Check if it's a built-in Go command
if result := execBuiltin(name, args, stdin); result != "" {
return result, nil
}
-
// Otherwise execute as system command
cmd := exec.Command(name, args...)
if stdin != "" {
@@ -187,7 +169,6 @@ func tokenize(input string) []string {
var current strings.Builder
inQuote := false
var quoteChar rune
-
for _, ch := range input {
if inQuote {
if ch == quoteChar {
@@ -197,13 +178,11 @@ func tokenize(input string) []string {
}
continue
}
-
if ch == '\'' || ch == '"' {
inQuote = true
quoteChar = ch
continue
}
-
if ch == ' ' || ch == '\t' {
if current.Len() > 0 {
tokens = append(tokens, current.String())
@@ -211,14 +190,11 @@ func tokenize(input string) []string {
}
continue
}
-
current.WriteRune(ch)
}
-
if current.Len() > 0 {
tokens = append(tokens, current.String())
}
-
return tokens
}
@@ -242,7 +218,7 @@ func execBuiltin(name string, args []string, stdin string) string {
path := args[0]
abs := path
if !filepath.IsAbs(path) {
- abs = filepath.Join(fsRootDir, path)
+ abs = filepath.Join(cfg.FilePickerDir, path)
}
data, err := os.ReadFile(abs)
if err != nil {
@@ -250,16 +226,16 @@ func execBuiltin(name string, args []string, stdin string) string {
}
return string(data)
case "pwd":
- return fsRootDir
+ return cfg.FilePickerDir
case "cd":
if len(args) == 0 {
return "[error] usage: cd <dir>"
}
dir := args[0]
- // Resolve relative to fsRootDir
+ // Resolve relative to cfg.FilePickerDir
abs := dir
if !filepath.IsAbs(dir) {
- abs = filepath.Join(fsRootDir, dir)
+ abs = filepath.Join(cfg.FilePickerDir, dir)
}
abs = filepath.Clean(abs)
info, err := os.Stat(abs)
@@ -269,8 +245,8 @@ func execBuiltin(name string, args []string, stdin string) string {
if !info.IsDir() {
return fmt.Sprintf("[error] cd: not a directory: %s", dir)
}
- fsRootDir = abs
- return fmt.Sprintf("Changed directory to: %s", fsRootDir)
+ cfg.FilePickerDir = abs
+ return fmt.Sprintf("Changed directory to: %s", cfg.FilePickerDir)
case "mkdir":
if len(args) == 0 {
return "[error] usage: mkdir [-p] <dir>"
@@ -289,7 +265,7 @@ func execBuiltin(name string, args []string, stdin string) string {
}
abs := dirPath
if !filepath.IsAbs(dirPath) {
- abs = filepath.Join(fsRootDir, dirPath)
+ abs = filepath.Join(cfg.FilePickerDir, dirPath)
}
abs = filepath.Clean(abs)
var mkdirFunc func(string, os.FileMode) error
@@ -315,7 +291,7 @@ func execBuiltin(name string, args []string, stdin string) string {
}
abs := dir
if !filepath.IsAbs(dir) {
- abs = filepath.Join(fsRootDir, dir)
+ abs = filepath.Join(cfg.FilePickerDir, dir)
}
entries, err := os.ReadDir(abs)
if err != nil {
@@ -347,7 +323,7 @@ func execBuiltin(name string, args []string, stdin string) string {
return "[error] usage: go <subcommand> [options]"
}
cmd := exec.Command("go", args...)
- cmd.Dir = fsRootDir
+ cmd.Dir = cfg.FilePickerDir
output, err := cmd.CombinedOutput()
if err != nil {
return fmt.Sprintf("[error] go %s: %v\n%s", args[0], err, string(output))
diff --git a/tools/fs.go b/tools/fs.go
index f0368f6..7b33a3c 100644
--- a/tools/fs.go
+++ b/tools/fs.go
@@ -14,7 +14,6 @@ import (
"time"
)
-var fsRootDir string
var memoryStore MemoryStore
var agentRole string
@@ -31,11 +30,11 @@ func SetMemoryStore(store MemoryStore, role string) {
}
func SetFSRoot(dir string) {
- fsRootDir = dir
+ cfg.FilePickerDir = dir
}
func GetFSRoot() string {
- return fsRootDir
+ return cfg.FilePickerDir
}
func SetFSCwd(dir string) error {
@@ -50,26 +49,24 @@ func SetFSCwd(dir string) error {
if !info.IsDir() {
return fmt.Errorf("not a directory: %s", dir)
}
- fsRootDir = abs
+ cfg.FilePickerDir = abs
return nil
}
func resolvePath(rel string) (string, error) {
- if fsRootDir == "" {
+ if cfg.FilePickerDir == "" {
return "", fmt.Errorf("fs root not set")
}
-
if filepath.IsAbs(rel) {
abs := filepath.Clean(rel)
- if !strings.HasPrefix(abs, fsRootDir+string(os.PathSeparator)) && abs != fsRootDir {
+ if !strings.HasPrefix(abs, cfg.FilePickerDir+string(os.PathSeparator)) && abs != cfg.FilePickerDir {
return "", fmt.Errorf("path escapes fs root: %s", rel)
}
return abs, nil
}
-
- abs := filepath.Join(fsRootDir, rel)
+ abs := filepath.Join(cfg.FilePickerDir, rel)
abs = filepath.Clean(abs)
- if !strings.HasPrefix(abs, fsRootDir+string(os.PathSeparator)) && abs != fsRootDir {
+ if !strings.HasPrefix(abs, cfg.FilePickerDir+string(os.PathSeparator)) && abs != cfg.FilePickerDir {
return "", fmt.Errorf("path escapes fs root: %s", rel)
}
return abs, nil
@@ -100,12 +97,10 @@ func FsLs(args []string, stdin string) string {
if err != nil {
return fmt.Sprintf("[error] %v", err)
}
-
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()
@@ -136,17 +131,14 @@ func FsCat(args []string, stdin string) string {
if path == "" {
return "[error] usage: cat <path>"
}
-
abs, err := resolvePath(path)
if err != nil {
return fmt.Sprintf("[error] %v", err)
}
-
data, err := os.ReadFile(abs)
if err != nil {
return fmt.Sprintf("[error] cat: %v", err)
}
-
if b64 {
result := base64.StdEncoding.EncodeToString(data)
if IsImageFile(path) {
@@ -162,7 +154,6 @@ func FsViewImg(args []string, stdin string) string {
return "[error] usage: view_img <image-path>"
}
path := args[0]
-
var abs string
if filepath.IsAbs(path) {
abs = path
@@ -173,20 +164,16 @@ func FsViewImg(args []string, stdin string) string {
return fmt.Sprintf("[error] %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)
}
-
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{
@@ -222,16 +209,13 @@ func FsWrite(args []string, stdin string) string {
if path == "" {
return "[error] usage: write <path> [content] or pipe stdin"
}
-
abs, err := resolvePath(path)
if err != nil {
return fmt.Sprintf("[error] %v", err)
}
-
if err := os.MkdirAll(filepath.Dir(abs), 0o755); err != nil {
return fmt.Sprintf("[error] mkdir: %v", err)
}
-
var data []byte
if b64 {
src := stdin
@@ -251,18 +235,14 @@ func FsWrite(args []string, stdin string) string {
data = []byte(stdin)
}
}
-
if err := os.WriteFile(abs, data, 0o644); err != nil {
return fmt.Sprintf("[error] write: %v", err)
}
-
size := humanSize(int64(len(data)))
result := fmt.Sprintf("Written %s → %s", size, path)
-
if IsImageFile(path) {
result += fmt.Sprintf("\n![image](file://%s)", abs)
}
-
return result
}
@@ -270,17 +250,14 @@ func FsStat(args []string, stdin string) string {
if len(args) == 0 {
return "[error] usage: stat <path>"
}
-
abs, err := resolvePath(args[0])
if err != nil {
return fmt.Sprintf("[error] %v", err)
}
-
info, err := os.Stat(abs)
if err != nil {
return fmt.Sprintf("[error] stat: %v", err)
}
-
mime := "application/octet-stream"
if IsImageFile(args[0]) {
ext := strings.ToLower(filepath.Ext(args[0]))
@@ -297,7 +274,6 @@ func FsStat(args []string, stdin string) string {
mime = "image/svg+xml"
}
}
-
var out strings.Builder
fmt.Fprintf(&out, "File: %s\n", args[0])
fmt.Fprintf(&out, "Size: %s (%d bytes)\n", humanSize(info.Size()), info.Size())
@@ -313,12 +289,10 @@ func FsRm(args []string, stdin string) string {
if len(args) == 0 {
return "[error] usage: rm <path>"
}
-
abs, err := resolvePath(args[0])
if err != nil {
return fmt.Sprintf("[error] %v", err)
}
-
if err := os.RemoveAll(abs); err != nil {
return fmt.Sprintf("[error] rm: %v", err)
}
@@ -329,7 +303,6 @@ func FsCp(args []string, stdin string) string {
if len(args) < 2 {
return "[error] usage: cp <src> <dst>"
}
-
srcAbs, err := resolvePath(args[0])
if err != nil {
return fmt.Sprintf("[error] %v", err)
@@ -338,16 +311,13 @@ func FsCp(args []string, stdin string) string {
if err != nil {
return fmt.Sprintf("[error] %v", err)
}
-
data, err := os.ReadFile(srcAbs)
if err != nil {
return fmt.Sprintf("[error] cp read: %v", err)
}
-
if err := os.MkdirAll(filepath.Dir(dstAbs), 0o755); err != nil {
return fmt.Sprintf("[error] cp mkdir: %v", err)
}
-
if err := os.WriteFile(dstAbs, data, 0o644); err != nil {
return fmt.Sprintf("[error] cp write: %v", err)
}
@@ -358,7 +328,6 @@ func FsMv(args []string, stdin string) string {
if len(args) < 2 {
return "[error] usage: mv <src> <dst>"
}
-
srcAbs, err := resolvePath(args[0])
if err != nil {
return fmt.Sprintf("[error] %v", err)
@@ -367,11 +336,9 @@ func FsMv(args []string, stdin string) string {
if err != nil {
return fmt.Sprintf("[error] %v", err)
}
-
if err := os.MkdirAll(filepath.Dir(dstAbs), 0o755); err != nil {
return fmt.Sprintf("[error] mv mkdir: %v", err)
}
-
if err := os.Rename(srcAbs, dstAbs); err != nil {
return fmt.Sprintf("[error] mv: %v", err)
}
@@ -382,10 +349,8 @@ func FsMkdir(args []string, stdin string) string {
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
@@ -393,27 +358,22 @@ func FsMkdir(args []string, stdin string) string {
dirPath = a
}
}
-
if dirPath == "" {
return "[error] usage: mkdir [-p] <dir>"
}
-
abs, err := resolvePath(dirPath)
if err != nil {
return fmt.Sprintf("[error] %v", err)
}
-
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)
}
@@ -459,7 +419,6 @@ func FsGrep(args []string, stdin string) string {
if ignoreCase {
pattern = strings.ToLower(pattern)
}
-
lines := strings.Split(stdin, "\n")
var matched []string
for _, line := range lines {
@@ -549,7 +508,6 @@ func FsSort(args []string, stdin string) string {
numeric = true
}
}
-
sortFunc := func(i, j int) bool {
if numeric {
ni, _ := strconv.Atoi(lines[i])
@@ -564,7 +522,6 @@ func FsSort(args []string, stdin string) string {
}
return lines[i] < lines[j]
}
-
sort.Slice(lines, sortFunc)
return strings.Join(lines, "\n")
}
@@ -577,7 +534,6 @@ func FsUniq(args []string, stdin string) string {
showCount = true
}
}
-
var result []string
var prev string
first := true
@@ -623,17 +579,14 @@ func FsGit(args []string, stdin string) string {
if len(args) == 0 {
return "[error] usage: git <subcommand> [options]"
}
-
subcmd := args[0]
if !allowedGitSubcommands[subcmd] {
return fmt.Sprintf("[error] git: '%s' is not an allowed git command. Allowed: status, log, diff, show, branch, reflog, rev-parse, shortlog, describe, rev-list", subcmd)
}
-
abs, err := resolvePath(".")
if err != nil {
return fmt.Sprintf("[error] git: %v", err)
}
-
// Pass all args to git (first arg is subcommand, rest are options)
cmd := exec.Command("git", args...)
cmd.Dir = abs
@@ -645,7 +598,7 @@ func FsGit(args []string, stdin string) string {
}
func FsPwd(args []string, stdin string) string {
- return fsRootDir
+ return cfg.FilePickerDir
}
func FsCd(args []string, stdin string) string {
@@ -664,19 +617,17 @@ func FsCd(args []string, stdin string) string {
if !info.IsDir() {
return fmt.Sprintf("[error] cd: not a directory: %s", dir)
}
- fsRootDir = abs
- return fmt.Sprintf("Changed directory to: %s", fsRootDir)
+ cfg.FilePickerDir = abs
+ return fmt.Sprintf("Changed directory to: %s", cfg.FilePickerDir)
}
func FsSed(args []string, stdin string) string {
if len(args) == 0 {
return "[error] usage: sed 's/old/new/[g]' [file]"
}
-
inPlace := false
var filePath string
var pattern string
-
for _, a := range args {
if a == "-i" || a == "--in-place" {
inPlace = true
@@ -687,21 +638,17 @@ func FsSed(args []string, stdin string) string {
filePath = a
}
}
-
if pattern == "" {
return "[error] usage: sed 's/old/new/[g]' [file]"
}
-
// Parse pattern: s/old/new/flags
parts := strings.Split(pattern[1:], "/")
if len(parts) < 2 {
return "[error] invalid sed pattern. Use: s/old/new/[g]"
}
-
oldStr := parts[0]
newStr := parts[1]
global := len(parts) >= 3 && strings.Contains(parts[2], "g")
-
var content string
if filePath != "" && stdin == "" {
// Read from file
@@ -720,14 +667,12 @@ func FsSed(args []string, stdin string) string {
} else {
return "[error] sed: no input (use file path or pipe from stdin)"
}
-
// Apply sed replacement
if global {
content = strings.ReplaceAll(content, oldStr, newStr)
} else {
content = strings.Replace(content, oldStr, newStr, 1)
}
-
if inPlace && filePath != "" {
abs, err := resolvePath(filePath)
if err != nil {
@@ -738,7 +683,6 @@ func FsSed(args []string, stdin string) string {
}
return fmt.Sprintf("Modified %s", filePath)
}
-
return content
}
@@ -746,11 +690,9 @@ func FsMemory(args []string, stdin string) string {
if len(args) == 0 {
return "[error] usage: memory store <topic> <data> | memory get <topic> | memory list | memory forget <topic>"
}
-
if memoryStore == nil {
return "[error] memory store not initialized"
}
-
switch args[0] {
case "store":
if len(args) < 3 && stdin == "" {
@@ -768,7 +710,6 @@ func FsMemory(args []string, stdin string) string {
return fmt.Sprintf("[error] failed to store: %v", err)
}
return fmt.Sprintf("Stored under topic: %s", topic)
-
case "get":
if len(args) < 2 {
return "[error] usage: memory get <topic>"
@@ -779,7 +720,6 @@ func FsMemory(args []string, stdin string) string {
return fmt.Sprintf("[error] failed to recall: %v", err)
}
return fmt.Sprintf("Topic: %s\n%s", topic, data)
-
case "list", "topics":
topics, err := memoryStore.RecallTopics(agentRole)
if err != nil {
@@ -789,7 +729,6 @@ func FsMemory(args []string, stdin string) string {
return "No topics stored."
}
return "Topics: " + strings.Join(topics, ", ")
-
case "forget", "delete":
if len(args) < 2 {
return "[error] usage: memory forget <topic>"
@@ -800,7 +739,6 @@ func FsMemory(args []string, stdin string) string {
return fmt.Sprintf("[error] failed to forget: %v", err)
}
return fmt.Sprintf("Deleted topic: %s", topic)
-
default:
return fmt.Sprintf("[error] unknown subcommand: %s. Use: store, get, list, topics, forget, delete", args[0])
}
diff --git a/tools/pw.go b/tools/pw.go
index c21e8fe..936da1a 100644
--- a/tools/pw.go
+++ b/tools/pw.go
@@ -12,87 +12,6 @@ import (
"github.com/playwright-community/playwright-go"
)
-var browserToolSysMsg = `
-Additional browser automation tools (Playwright):
-[
-{
- "name": "pw_start",
- "args": [],
- "when_to_use": "start a browser instance before doing any browser automation. Must be called first."
-},
-{
- "name": "pw_stop",
- "args": [],
- "when_to_use": "stop the browser instance when done with automation."
-},
-{
- "name": "pw_is_running",
- "args": [],
- "when_to_use": "check if browser is currently running."
-},
-{
- "name": "pw_navigate",
- "args": ["url"],
- "when_to_use": "open a specific URL in the web browser."
-},
-{
- "name": "pw_click",
- "args": ["selector", "index"],
- "when_to_use": "click on an element on the current webpage. Use 'index' for multiple matches (default 0)."
-},
-{
- "name": "pw_fill",
- "args": ["selector", "text", "index"],
- "when_to_use": "type text into an input field. Use 'index' for multiple matches (default 0)."
-},
-{
- "name": "pw_extract_text",
- "args": ["selector"],
- "when_to_use": "extract text content from the page or specific elements. Use selector 'body' for all page text."
-},
-{
- "name": "pw_screenshot",
- "args": ["selector", "full_page"],
- "when_to_use": "take a screenshot of the page or a specific element. Returns a file path to the image. Use to verify actions or inspect visual state."
-},
-{
- "name": "pw_screenshot_and_view",
- "args": ["selector", "full_page"],
- "when_to_use": "take a screenshot and return the image for viewing. Use to visually verify page state."
-},
-{
- "name": "pw_wait_for_selector",
- "args": ["selector", "timeout"],
- "when_to_use": "wait for an element to appear on the page before proceeding with further actions."
-},
-{
- "name": "pw_drag",
- "args": ["x1", "y1", "x2", "y2"],
- "when_to_use": "drag the mouse from point (x1,y1) to (x2,y2)."
-},
-{
- "name": "pw_click_at",
- "args": ["x", "y"],
- "when_to_use": "click at specific X,Y coordinates on the page. Use when you know the exact position."
-},
-{
- "name": "pw_get_html",
- "args": ["selector"],
- "when_to_use": "get the HTML content of the page or a specific element. Use to understand page structure or extract raw HTML."
-},
-{
- "name": "pw_get_dom",
- "args": ["selector"],
- "when_to_use": "get a structured DOM representation with tag, attributes, text, and children. Use to inspect element hierarchy and properties."
-},
-{
- "name": "pw_search_elements",
- "args": ["text", "selector"],
- "when_to_use": "search for elements by text content or CSS selector. Returns matching elements with their tags, text, and HTML."
-}
-]
-`
-
var (
pw *playwright.Playwright
browser playwright.Browser
@@ -532,6 +451,7 @@ func pwDragBySelector(args map[string]string) []byte {
return []byte(fmt.Sprintf(`{"success": true, "message": "%s"}`, msg))
}
+// nolint:unused
func pwClickAt(args map[string]string) []byte {
x, ok := args["x"]
if !ok {
@@ -682,6 +602,7 @@ func pwGetDOM(args map[string]string) []byte {
return []byte(fmt.Sprintf(`{"dom": %s}`, string(data)))
}
+// nolint:unused
func pwSearchElements(args map[string]string) []byte {
text := args["text"]
selector := args["selector"]
diff --git a/tools/tools.go b/tools/tools.go
index ee20a2d..bf6976f 100644
--- a/tools/tools.go
+++ b/tools/tools.go
@@ -8,11 +8,9 @@ import (
"gf-lt/config"
"gf-lt/models"
"gf-lt/storage"
- "io"
"log/slog"
"os"
"os/exec"
- "path/filepath"
"regexp"
"strconv"
"strings"
@@ -25,20 +23,6 @@ import (
)
var (
- ToolCallRE = regexp.MustCompile(`__tool_call__\s*([\s\S]*?)__tool_call__`)
- QuotesRE = regexp.MustCompile(`(".*?")`)
- StarRE = regexp.MustCompile(`(\*.*?\*)`)
- ThinkRE = regexp.MustCompile(`(?s)<think>.*?</think>`)
- toolCallRE = ToolCallRE
- quotesRE = QuotesRE
- starRE = StarRE
- thinkRE = ThinkRE
- CodeBlockRE = regexp.MustCompile(`(?s)\x60{3}(?:.*?)\n(.*?)\n\s*\x60{3}\s*`)
- SingleBacktickRE = regexp.MustCompile(`\x60([^\x60]*)\x60`)
- codeBlockRE = CodeBlockRE
- singleBacktickRE = SingleBacktickRE
- RoleRE = regexp.MustCompile(`^(\w+):`)
- SysLabels = []string{"assistant"}
RpDefenitionSysMsg = `
For this roleplay immersion is at most importance.
Every character thinks and acts based on their personality and setting of the roleplay.
@@ -115,32 +99,8 @@ After that you are free to respond to the user.
ragSearchSysPrompt = `Synthesize the document search results, extracting key information and presenting a concise answer. Provide sources and document IDs where relevant.`
readURLSysPrompt = `Extract and summarize the content from the webpage. Provide key information, main points, and any relevant details.`
summarySysPrompt = `Please provide a concise summary of the following conversation. Focus on key points, decisions, and actions. Provide only the summary, no additional commentary.`
- webAgentClient *agent.AgentClient
- webAgentClientOnce sync.Once
- webAgentsOnce sync.Once
)
-var windowToolSysMsg = `
-Additional window tools (available only if xdotool and maim are installed):
-[
-{
-"name":"list_windows",
-"args": [],
-"when_to_use": "when asked to list visible windows; returns map of window ID to window name"
-},
-{
-"name":"capture_window",
-"args": ["window"],
-"when_to_use": "when asked to take a screenshot of a specific window; saves to /tmp; window can be ID or name substring; returns file path"
-},
-{
-"name":"capture_window_and_view",
-"args": ["window"],
-"when_to_use": "when asked to take a screenshot of a specific window and show it; saves to /tmp and returns image for viewing; window can be ID or name substring"
-}
-]
-`
-
var WebSearcher searcher.WebSurfer
var (
@@ -156,9 +116,22 @@ type Tools struct {
logger *slog.Logger
store storage.FullRepo
WindowToolsAvailable bool
- getTokenFunc func() string
- webAgentClient *agent.AgentClient
- webAgentClientOnce sync.Once
+ // getTokenFunc func() string
+ webAgentClient *agent.AgentClient
+ webAgentClientOnce sync.Once
+ webSearchAgent agent.AgenterB
+}
+
+func (t *Tools) initAgentsB() {
+ t.GetWebAgentClient()
+ t.webSearchAgent = agent.NewWebAgentB(t.webAgentClient, webSearchSysPrompt)
+ agent.RegisterB("rag_search", agent.NewWebAgentB(t.webAgentClient, ragSearchSysPrompt))
+ // Register websearch agent
+ agent.RegisterB("websearch", agent.NewWebAgentB(t.webAgentClient, webSearchSysPrompt))
+ // Register read_url agent
+ agent.RegisterB("read_url", agent.NewWebAgentB(t.webAgentClient, readURLSysPrompt))
+ // Register summarize_chat agent
+ agent.RegisterB("summarize_chat", agent.NewWebAgentB(t.webAgentClient, summarySysPrompt))
}
func InitTools(cfg *config.Config, logger *slog.Logger, store storage.FullRepo) *Tools {
@@ -201,6 +174,7 @@ func InitTools(cfg *config.Config, logger *slog.Logger, store storage.FullRepo)
store: store,
}
t.checkWindowTools()
+ t.initAgentsB()
return t
}
@@ -224,17 +198,17 @@ func SetTokenFunc(fn func() string) {
getTokenFunc = fn
}
-func getWebAgentClient() *agent.AgentClient {
- webAgentClientOnce.Do(func() {
+func (t *Tools) GetWebAgentClient() *agent.AgentClient {
+ t.webAgentClientOnce.Do(func() {
getToken := func() string {
if getTokenFunc != nil {
return getTokenFunc()
}
return ""
}
- webAgentClient = agent.NewAgentClient(cfg, logger, getToken)
+ t.webAgentClient = agent.NewAgentClient(cfg, logger, getToken)
})
- return webAgentClient
+ return t.webAgentClient
}
func RegisterWindowTools(modelHasVision bool) {
@@ -242,27 +216,13 @@ func RegisterWindowTools(modelHasVision bool) {
// Window tools registration happens here if needed
}
-func RegisterPlaywrightTools() {
- removePlaywrightToolsFromBaseTools()
- if cfg != nil && cfg.PlaywrightEnabled {
- // Playwright tools are registered here
- }
-}
-
-// webAgentsOnce.Do(func() {
-// client := getWebAgentClient()
-// // Register rag_search agent
-// agent.RegisterB("rag_search", agent.NewWebAgentB(client, ragSearchSysPrompt))
-// // Register websearch agent
-// agent.RegisterB("websearch", agent.NewWebAgentB(client, webSearchSysPrompt))
-// // Register read_url agent
-// agent.RegisterB("read_url", agent.NewWebAgentB(client, readURLSysPrompt))
-// // Register summarize_chat agent
-// agent.RegisterB("summarize_chat", agent.NewWebAgentB(client, summarySysPrompt))
-// })
+// func RegisterPlaywrightTools() {
+// removePlaywrightToolsFromBaseTools()
+// if cfg != nil && cfg.PlaywrightEnabled {
+// // Playwright tools are registered here
+// }
// }
-// web search (depends on extra server)
func websearch(args map[string]string) []byte {
// make http request return bytes
query, ok := args["query"]
@@ -407,89 +367,6 @@ func readURLRaw(args map[string]string) []byte {
return []byte(fmt.Sprintf("%+v", resp))
}
-// // Helper functions for file operations
-// func resolvePath(p string) string {
-// if filepath.IsAbs(p) {
-// return p
-// }
-// return filepath.Join(cfg.FilePickerDir, p)
-// }
-
-func readStringFromFile(filename string) (string, error) {
- data, err := os.ReadFile(filename)
- if err != nil {
- return "", err
- }
- return string(data), nil
-}
-
-func writeStringToFile(filename string, data string) error {
- return os.WriteFile(filename, []byte(data), 0644)
-}
-
-func appendStringToFile(filename string, data string) error {
- file, err := os.OpenFile(filename, os.O_APPEND|os.O_CREATE|os.O_WRONLY, 0644)
- if err != nil {
- return err
- }
- defer file.Close()
- _, err = file.WriteString(data)
- return err
-}
-
-func removeFile(filename string) error {
- return os.Remove(filename)
-}
-
-func moveFile(src, dst string) error {
- // First try with os.Rename (works within same filesystem)
- if err := os.Rename(src, dst); err == nil {
- return nil
- }
- // If that fails (e.g., cross-filesystem), copy and delete
- return copyAndRemove(src, dst)
-}
-
-func copyFile(src, dst string) error {
- srcFile, err := os.Open(src)
- if err != nil {
- return err
- }
- defer srcFile.Close()
- dstFile, err := os.Create(dst)
- if err != nil {
- return err
- }
- defer dstFile.Close()
- _, err = io.Copy(dstFile, srcFile)
- return err
-}
-
-func copyAndRemove(src, dst string) error {
- // Copy the file
- if err := copyFile(src, dst); err != nil {
- return err
- }
- // Remove the source file
- return os.Remove(src)
-}
-
-func listDirectory(path string) ([]string, error) {
- entries, err := os.ReadDir(path)
- if err != nil {
- return nil, err
- }
- var files []string
- for _, entry := range entries {
- if entry.IsDir() {
- files = append(files, entry.Name()+"/") // Add "/" to indicate directory
- } else {
- files = append(files, entry.Name())
- }
- }
- return files, nil
-}
-
// Unified run command - single entry point for shell, memory, and todo
func runCmd(args map[string]string) []byte {
commandStr := args["command"]
@@ -498,16 +375,13 @@ func runCmd(args map[string]string) []byte {
logger.Error(msg)
return []byte(msg)
}
-
// Parse the command - first word is subcommand
parts := strings.Fields(commandStr)
if len(parts) == 0 {
return []byte("[error] empty command")
}
-
subcmd := parts[0]
rest := parts[1:]
-
// Route to appropriate handler
switch subcmd {
case "help":
@@ -566,10 +440,8 @@ Actions:
wait <selector> - wait for element
drag <from> <to> - drag element`)
}
-
action := args[0]
rest := args[1:]
-
switch action {
case "start":
return pwStart(originalArgs)
@@ -916,9 +788,7 @@ func handleTodoSubcommand(args []string, originalArgs map[string]string) []byte
if len(args) == 0 {
return []byte("usage: todo create|read|update|delete")
}
-
subcmd := args[0]
-
switch subcmd {
case "create":
task := strings.Join(args[1:], " ")
@@ -929,26 +799,22 @@ func handleTodoSubcommand(args []string, originalArgs map[string]string) []byte
return []byte("usage: todo create <task>")
}
return todoCreate(map[string]string{"task": task})
-
case "read":
id := ""
if len(args) > 1 {
id = args[1]
}
return todoRead(map[string]string{"id": id})
-
case "update":
if len(args) < 2 {
return []byte("usage: todo update <id> <status>")
}
return todoUpdate(map[string]string{"id": args[1], "status": args[2]})
-
case "delete":
if len(args) < 2 {
return []byte("usage: todo delete <id>")
}
return todoDelete(map[string]string{"id": args[1]})
-
default:
return []byte(fmt.Sprintf("unknown todo subcommand: %s", subcmd))
}
@@ -962,55 +828,53 @@ func executeCommand(args map[string]string) []byte {
logger.Error(msg)
return []byte(msg)
}
-
// Use chain execution for pipe/chaining support
result := ExecChain(commandStr)
return []byte(result)
}
-// handleCdCommand handles the cd command to update FilePickerDir
-func handleCdCommand(args []string) []byte {
- var targetDir string
- if len(args) == 0 {
- // cd with no args goes to home directory
- homeDir, err := os.UserHomeDir()
- if err != nil {
- msg := "cd: cannot determine home directory: " + err.Error()
- logger.Error(msg)
- return []byte(msg)
- }
- targetDir = homeDir
- } else {
- targetDir = args[0]
- }
- // Resolve relative paths against current FilePickerDir
- if !filepath.IsAbs(targetDir) {
- targetDir = filepath.Join(cfg.FilePickerDir, targetDir)
- }
- // Verify the directory exists
- info, err := os.Stat(targetDir)
- if err != nil {
- msg := "cd: " + targetDir + ": " + err.Error()
- logger.Error(msg)
- return []byte(msg)
- }
- if !info.IsDir() {
- msg := "cd: " + targetDir + ": not a directory"
- logger.Error(msg)
- return []byte(msg)
- }
-
- // Update FilePickerDir
- absDir, err := filepath.Abs(targetDir)
- if err != nil {
- msg := "cd: failed to resolve path: " + err.Error()
- logger.Error(msg)
- return []byte(msg)
- }
- cfg.FilePickerDir = absDir
- msg := "FilePickerDir changed to: " + absDir
- return []byte(msg)
-}
+// // handleCdCommand handles the cd command to update FilePickerDir
+// func handleCdCommand(args []string) []byte {
+// var targetDir string
+// if len(args) == 0 {
+// // cd with no args goes to home directory
+// homeDir, err := os.UserHomeDir()
+// if err != nil {
+// msg := "cd: cannot determine home directory: " + err.Error()
+// logger.Error(msg)
+// return []byte(msg)
+// }
+// targetDir = homeDir
+// } else {
+// targetDir = args[0]
+// }
+// // Resolve relative paths against current FilePickerDir
+// if !filepath.IsAbs(targetDir) {
+// targetDir = filepath.Join(cfg.FilePickerDir, targetDir)
+// }
+// // Verify the directory exists
+// info, err := os.Stat(targetDir)
+// if err != nil {
+// msg := "cd: " + targetDir + ": " + err.Error()
+// logger.Error(msg)
+// return []byte(msg)
+// }
+// if !info.IsDir() {
+// msg := "cd: " + targetDir + ": not a directory"
+// logger.Error(msg)
+// return []byte(msg)
+// }
+// // Update FilePickerDir
+// absDir, err := filepath.Abs(targetDir)
+// if err != nil {
+// msg := "cd: failed to resolve path: " + err.Error()
+// logger.Error(msg)
+// return []byte(msg)
+// }
+// cfg.FilePickerDir = absDir
+// msg := "FilePickerDir changed to: " + absDir
+// return []byte(msg)
+// }
// Helper functions for command execution
// Todo structure
@@ -1405,8 +1269,8 @@ var FnMap = map[string]fnSig{
"view_img": viewImgTool,
"help": helpTool,
// Unified run command
- "run": runCmd,
- // "summarize_chat": summarizeChat,
+ "run": runCmd,
+ "summarize_chat": summarizeChat,
}
func removeWindowToolsFromBaseTools() {
@@ -1427,476 +1291,485 @@ func removeWindowToolsFromBaseTools() {
delete(FnMap, "capture_window_and_view")
}
-func removePlaywrightToolsFromBaseTools() {
- playwrightToolNames := map[string]bool{
- "pw_start": true,
- "pw_stop": true,
- "pw_is_running": true,
- "pw_navigate": true,
- "pw_click": true,
- "pw_click_at": true,
- "pw_fill": true,
- "pw_extract_text": true,
- "pw_screenshot": true,
- "pw_screenshot_and_view": true,
- "pw_wait_for_selector": true,
- "pw_drag": true,
- }
- var filtered []models.Tool
- for _, tool := range BaseTools {
- if !playwrightToolNames[tool.Function.Name] {
- filtered = append(filtered, tool)
- }
+func summarizeChat(args map[string]string) []byte {
+ data, err := json.Marshal(args)
+ if err != nil {
+ return []byte("error: failed to marshal arguments")
}
- BaseTools = filtered
- delete(FnMap, "pw_start")
- delete(FnMap, "pw_stop")
- delete(FnMap, "pw_is_running")
- delete(FnMap, "pw_navigate")
- delete(FnMap, "pw_click")
- delete(FnMap, "pw_click_at")
- delete(FnMap, "pw_fill")
- delete(FnMap, "pw_extract_text")
- delete(FnMap, "pw_screenshot")
- delete(FnMap, "pw_screenshot_and_view")
- delete(FnMap, "pw_wait_for_selector")
- delete(FnMap, "pw_drag")
+ return []byte(data)
}
-func (t *Tools) RegisterWindowTools(modelHasVision bool) {
- removeWindowToolsFromBaseTools()
- if t.WindowToolsAvailable {
- FnMap["list_windows"] = listWindows
- FnMap["capture_window"] = captureWindow
- windowTools := []models.Tool{
- {
- Type: "function",
- Function: models.ToolFunc{
- Name: "list_windows",
- Description: "List all visible windows with their IDs and names. Returns a map of window ID to window name.",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{},
- Properties: map[string]models.ToolArgProps{},
- },
- },
- },
- {
- Type: "function",
- Function: models.ToolFunc{
- Name: "capture_window",
- Description: "Capture a screenshot of a specific window and save it to /tmp. Requires window parameter (window ID or name substring).",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{"window"},
- Properties: map[string]models.ToolArgProps{
- "window": models.ToolArgProps{
- Type: "string",
- Description: "window ID or window name (partial match)",
- },
- },
- },
- },
- },
- }
- if modelHasVision {
- FnMap["capture_window_and_view"] = captureWindowAndView
- windowTools = append(windowTools, models.Tool{
- Type: "function",
- Function: models.ToolFunc{
- Name: "capture_window_and_view",
- Description: "Capture a screenshot of a specific window, save it to /tmp, and return the image for viewing. Requires window parameter (window ID or name substring).",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{"window"},
- Properties: map[string]models.ToolArgProps{
- "window": models.ToolArgProps{
- Type: "string",
- Description: "window ID or window name (partial match)",
- },
- },
- },
- },
- })
- }
- BaseTools = append(BaseTools, windowTools...)
- ToolSysMsg += windowToolSysMsg
- }
-}
+// func removePlaywrightToolsFromBaseTools() {
+// playwrightToolNames := map[string]bool{
+// "pw_start": true,
+// "pw_stop": true,
+// "pw_is_running": true,
+// "pw_navigate": true,
+// "pw_click": true,
+// "pw_click_at": true,
+// "pw_fill": true,
+// "pw_extract_text": true,
+// "pw_screenshot": true,
+// "pw_screenshot_and_view": true,
+// "pw_wait_for_selector": true,
+// "pw_drag": true,
+// }
+// var filtered []models.Tool
+// for _, tool := range BaseTools {
+// if !playwrightToolNames[tool.Function.Name] {
+// filtered = append(filtered, tool)
+// }
+// }
+// BaseTools = filtered
+// delete(FnMap, "pw_start")
+// delete(FnMap, "pw_stop")
+// delete(FnMap, "pw_is_running")
+// delete(FnMap, "pw_navigate")
+// delete(FnMap, "pw_click")
+// delete(FnMap, "pw_click_at")
+// delete(FnMap, "pw_fill")
+// delete(FnMap, "pw_extract_text")
+// delete(FnMap, "pw_screenshot")
+// delete(FnMap, "pw_screenshot_and_view")
+// delete(FnMap, "pw_wait_for_selector")
+// delete(FnMap, "pw_drag")
+// }
-var browserAgentSysPrompt = `You are an autonomous browser automation agent. Your goal is to complete the user's task by intelligently using browser automation
-
-Important: The browser may already be running from a previous task! Always check pw_is_running first before starting a new browser.
-
-Available tools:
-- pw_start: Start browser (only if not already running)
-- pw_stop: Stop browser (only when you're truly done and browser is no longer needed)
-- pw_is_running: Check if browser is running
-- pw_navigate: Go to a URL
-- pw_click: Click an element by CSS selector
-- pw_fill: Type text into an input
-- pw_extract_text: Get text from page/element
-- pw_screenshot: Take a screenshot (returns file path)
-- pw_screenshot_and_view: Take screenshot with image for viewing
-- pw_wait_for_selector: Wait for element to appear
-- pw_drag: Drag mouse from one point to another
-- pw_click_at: Click at X,Y coordinates
-- pw_get_html: Get HTML content
-- pw_get_dom: Get structured DOM tree
-- pw_search_elements: Search for elements by text or selector
-
-Workflow:
-1. First, check if browser is already running (pw_is_running)
-2. Only start browser if not already running (pw_start)
-3. Navigate to required pages (pw_navigate)
-4. Interact with elements as needed (click, fill, etc.)
-5. Extract information or take screenshots as requested
-6. IMPORTANT: Do NOT stop the browser when done! Leave it running so the user can continue interacting with the page in subsequent requests.
-
-Always provide clear feedback about what you're doing and what you found.`
-
-func runBrowserAgent(args map[string]string) []byte {
- task, ok := args["task"]
- if !ok || task == "" {
- return []byte(`{"error": "task argument is required"}`)
- }
- client := getWebAgentClient()
- pwAgent := agent.NewPWAgent(client, browserAgentSysPrompt)
- pwAgent.SetTools(agent.GetPWTools())
- return pwAgent.ProcessTask(task)
-}
+// func (t *Tools) RegisterWindowTools(modelHasVision bool) {
+// removeWindowToolsFromBaseTools()
+// if t.WindowToolsAvailable {
+// FnMap["list_windows"] = listWindows
+// FnMap["capture_window"] = captureWindow
+// windowTools := []models.Tool{
+// {
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "list_windows",
+// Description: "List all visible windows with their IDs and names. Returns a map of window ID to window name.",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{},
+// Properties: map[string]models.ToolArgProps{},
+// },
+// },
+// },
+// {
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "capture_window",
+// Description: "Capture a screenshot of a specific window and save it to /tmp. Requires window parameter (window ID or name substring).",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{"window"},
+// Properties: map[string]models.ToolArgProps{
+// "window": models.ToolArgProps{
+// Type: "string",
+// Description: "window ID or window name (partial match)",
+// },
+// },
+// },
+// },
+// },
+// }
+// if modelHasVision {
+// FnMap["capture_window_and_view"] = captureWindowAndView
+// windowTools = append(windowTools, models.Tool{
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "capture_window_and_view",
+// Description: "Capture a screenshot of a specific window, save it to /tmp, and return the image for viewing. Requires window parameter (window ID or name substring).",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{"window"},
+// Properties: map[string]models.ToolArgProps{
+// "window": models.ToolArgProps{
+// Type: "string",
+// Description: "window ID or window name (partial match)",
+// },
+// },
+// },
+// },
+// })
+// }
+// BaseTools = append(BaseTools, windowTools...)
+// ToolSysMsg += windowToolSysMsg
+// }
+// }
-func registerPlaywrightTools() {
- removePlaywrightToolsFromBaseTools()
- if cfg != nil && cfg.PlaywrightEnabled {
- FnMap["pw_start"] = pwStart
- FnMap["pw_stop"] = pwStop
- FnMap["pw_is_running"] = pwIsRunning
- FnMap["pw_navigate"] = pwNavigate
- FnMap["pw_click"] = pwClick
- FnMap["pw_click_at"] = pwClickAt
- FnMap["pw_fill"] = pwFill
- FnMap["pw_extract_text"] = pwExtractText
- FnMap["pw_screenshot"] = pwScreenshot
- FnMap["pw_screenshot_and_view"] = pwScreenshotAndView
- FnMap["pw_wait_for_selector"] = pwWaitForSelector
- FnMap["pw_drag"] = pwDrag
- FnMap["pw_get_html"] = pwGetHTML
- FnMap["pw_get_dom"] = pwGetDOM
- FnMap["pw_search_elements"] = pwSearchElements
- playwrightTools := []models.Tool{
- {
- Type: "function",
- Function: models.ToolFunc{
- Name: "pw_start",
- Description: "Start a Playwright browser instance. Call this first before using other pw_ Uses headless mode by default (set PlaywrightHeadless=false in config for GUI).",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{},
- Properties: map[string]models.ToolArgProps{},
- },
- },
- },
- {
- Type: "function",
- Function: models.ToolFunc{
- Name: "pw_stop",
- Description: "Stop the Playwright browser instance. Call when done with browser automation.",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{},
- Properties: map[string]models.ToolArgProps{},
- },
- },
- },
- {
- Type: "function",
- Function: models.ToolFunc{
- Name: "pw_is_running",
- Description: "Check if Playwright browser is currently running.",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{},
- Properties: map[string]models.ToolArgProps{},
- },
- },
- },
- {
- Type: "function",
- Function: models.ToolFunc{
- Name: "pw_navigate",
- Description: "Navigate to a URL in the browser.",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{"url"},
- Properties: map[string]models.ToolArgProps{
- "url": models.ToolArgProps{
- Type: "string",
- Description: "URL to navigate to",
- },
- },
- },
- },
- },
- {
- Type: "function",
- Function: models.ToolFunc{
- Name: "pw_click",
- Description: "Click on an element using CSS selector. Use 'index' for multiple matches (default 0).",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{"selector"},
- Properties: map[string]models.ToolArgProps{
- "selector": models.ToolArgProps{
- Type: "string",
- Description: "CSS selector for the element to click",
- },
- "index": models.ToolArgProps{
- Type: "string",
- Description: "optional index for multiple matches (default 0)",
- },
- },
- },
- },
- },
- {
- Type: "function",
- Function: models.ToolFunc{
- Name: "pw_fill",
- Description: "Fill an input field with text using CSS selector.",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{"selector", "text"},
- Properties: map[string]models.ToolArgProps{
- "selector": models.ToolArgProps{
- Type: "string",
- Description: "CSS selector for the input element",
- },
- "text": models.ToolArgProps{
- Type: "string",
- Description: "text to fill into the input",
- },
- "index": models.ToolArgProps{
- Type: "string",
- Description: "optional index for multiple matches (default 0)",
- },
- },
- },
- },
- },
- {
- Type: "function",
- Function: models.ToolFunc{
- Name: "pw_extract_text",
- Description: "Extract text content from the page or specific elements using CSS selector. Use 'body' for all page text.",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{"selector"},
- Properties: map[string]models.ToolArgProps{
- "selector": models.ToolArgProps{
- Type: "string",
- Description: "CSS selector (use 'body' for all page text)",
- },
- },
- },
- },
- },
- {
- Type: "function",
- Function: models.ToolFunc{
- Name: "pw_screenshot",
- Description: "Take a screenshot of the page or a specific element. Returns file path to saved image.",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{},
- Properties: map[string]models.ToolArgProps{
- "selector": models.ToolArgProps{
- Type: "string",
- Description: "optional CSS selector for element to screenshot",
- },
- "full_page": models.ToolArgProps{
- Type: "string",
- Description: "optional: 'true' to capture full page (default false)",
- },
- },
- },
- },
- },
- {
- Type: "function",
- Function: models.ToolFunc{
- Name: "pw_screenshot_and_view",
- Description: "Take a screenshot and return the image for viewing. Use when model needs to see the screenshot.",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{},
- Properties: map[string]models.ToolArgProps{
- "selector": models.ToolArgProps{
- Type: "string",
- Description: "optional CSS selector for element to screenshot",
- },
- "full_page": models.ToolArgProps{
- Type: "string",
- Description: "optional: 'true' to capture full page (default false)",
- },
- },
- },
- },
- },
- {
- Type: "function",
- Function: models.ToolFunc{
- Name: "pw_wait_for_selector",
- Description: "Wait for an element to appear on the page.",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{"selector"},
- Properties: map[string]models.ToolArgProps{
- "selector": models.ToolArgProps{
- Type: "string",
- Description: "CSS selector to wait for",
- },
- "timeout": models.ToolArgProps{
- Type: "string",
- Description: "optional timeout in ms (default 30000)",
- },
- },
- },
- },
- },
- {
- Type: "function",
- Function: models.ToolFunc{
- Name: "pw_drag",
- Description: "Drag the mouse from one point to another.",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{"x1", "y1", "x2", "y2"},
- Properties: map[string]models.ToolArgProps{
- "x1": models.ToolArgProps{
- Type: "string",
- Description: "starting X coordinate",
- },
- "y1": models.ToolArgProps{
- Type: "string",
- Description: "starting Y coordinate",
- },
- "x2": models.ToolArgProps{
- Type: "string",
- Description: "ending X coordinate",
- },
- "y2": models.ToolArgProps{
- Type: "string",
- Description: "ending Y coordinate",
- },
- },
- },
- },
- },
- {
- Type: "function",
- Function: models.ToolFunc{
- Name: "pw_get_html",
- Description: "Get the HTML content of the page or a specific element.",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{},
- Properties: map[string]models.ToolArgProps{
- "selector": models.ToolArgProps{
- Type: "string",
- Description: "optional CSS selector (default: body)",
- },
- },
- },
- },
- },
- {
- Type: "function",
- Function: models.ToolFunc{
- Name: "pw_get_dom",
- Description: "Get a structured DOM representation of an element with tag, attributes, text, and children.",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{},
- Properties: map[string]models.ToolArgProps{
- "selector": models.ToolArgProps{
- Type: "string",
- Description: "optional CSS selector (default: body)",
- },
- },
- },
- },
- },
- {
- Type: "function",
- Function: models.ToolFunc{
- Name: "pw_search_elements",
- Description: "Search for elements by text content or CSS selector. Returns matching elements with their tags, text, and HTML.",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{},
- Properties: map[string]models.ToolArgProps{
- "text": models.ToolArgProps{
- Type: "string",
- Description: "text to search for in elements",
- },
- "selector": models.ToolArgProps{
- Type: "string",
- Description: "CSS selector to search for",
- },
- },
- },
- },
- },
- }
- BaseTools = append(BaseTools, playwrightTools...)
- ToolSysMsg += browserToolSysMsg
- agent.RegisterPWTool("pw_start", pwStart)
- agent.RegisterPWTool("pw_stop", pwStop)
- agent.RegisterPWTool("pw_is_running", pwIsRunning)
- agent.RegisterPWTool("pw_navigate", pwNavigate)
- agent.RegisterPWTool("pw_click", pwClick)
- agent.RegisterPWTool("pw_click_at", pwClickAt)
- agent.RegisterPWTool("pw_fill", pwFill)
- agent.RegisterPWTool("pw_extract_text", pwExtractText)
- agent.RegisterPWTool("pw_screenshot", pwScreenshot)
- agent.RegisterPWTool("pw_screenshot_and_view", pwScreenshotAndView)
- agent.RegisterPWTool("pw_wait_for_selector", pwWaitForSelector)
- agent.RegisterPWTool("pw_drag", pwDrag)
- agent.RegisterPWTool("pw_get_html", pwGetHTML)
- agent.RegisterPWTool("pw_get_dom", pwGetDOM)
- agent.RegisterPWTool("pw_search_elements", pwSearchElements)
- browserAgentTool := []models.Tool{
- {
- Type: "function",
- Function: models.ToolFunc{
- Name: "browser_agent",
- Description: "Autonomous browser automation agent. Use for complex multi-step browser tasks like 'go to website, login, and take screenshot'. The agent will plan and execute steps automatically using browser ",
- Parameters: models.ToolFuncParams{
- Type: "object",
- Required: []string{"task"},
- Properties: map[string]models.ToolArgProps{
- "task": {Type: "string", Description: "The task to accomplish, e.g., 'go to github.com and take a screenshot of the homepage'"},
- },
- },
- },
- },
- }
- BaseTools = append(BaseTools, browserAgentTool...)
- FnMap["browser_agent"] = runBrowserAgent
- }
-}
+// for pw agentA
+// var browserAgentSysPrompt = `You are an autonomous browser automation agent. Your goal is to complete the user's task by intelligently using browser automation
+
+// Important: The browser may already be running from a previous task! Always check pw_is_running first before starting a new browser.
+
+// Available tools:
+// - pw_start: Start browser (only if not already running)
+// - pw_stop: Stop browser (only when you're truly done and browser is no longer needed)
+// - pw_is_running: Check if browser is running
+// - pw_navigate: Go to a URL
+// - pw_click: Click an element by CSS selector
+// - pw_fill: Type text into an input
+// - pw_extract_text: Get text from page/element
+// - pw_screenshot: Take a screenshot (returns file path)
+// - pw_screenshot_and_view: Take screenshot with image for viewing
+// - pw_wait_for_selector: Wait for element to appear
+// - pw_drag: Drag mouse from one point to another
+// - pw_click_at: Click at X,Y coordinates
+// - pw_get_html: Get HTML content
+// - pw_get_dom: Get structured DOM tree
+// - pw_search_elements: Search for elements by text or selector
+
+// Workflow:
+// 1. First, check if browser is already running (pw_is_running)
+// 2. Only start browser if not already running (pw_start)
+// 3. Navigate to required pages (pw_navigate)
+// 4. Interact with elements as needed (click, fill, etc.)
+// 5. Extract information or take screenshots as requested
+// 6. IMPORTANT: Do NOT stop the browser when done! Leave it running so the user can continue interacting with the page in subsequent requests.
+
+// Always provide clear feedback about what you're doing and what you found.`
+
+// func (t *Tools) runBrowserAgent(args map[string]string) []byte {
+// task, ok := args["task"]
+// if !ok || task == "" {
+// return []byte(`{"error": "task argument is required"}`)
+// }
+// client := t.GetWebAgentClient()
+// pwAgent := agent.NewPWAgent(client, browserAgentSysPrompt)
+// pwAgent.SetTools(agent.GetPWTools())
+// return pwAgent.ProcessTask(task)
+// }
+
+// func registerPlaywrightTools() {
+// removePlaywrightToolsFromBaseTools()
+// if cfg != nil && cfg.PlaywrightEnabled {
+// FnMap["pw_start"] = pwStart
+// FnMap["pw_stop"] = pwStop
+// FnMap["pw_is_running"] = pwIsRunning
+// FnMap["pw_navigate"] = pwNavigate
+// FnMap["pw_click"] = pwClick
+// FnMap["pw_click_at"] = pwClickAt
+// FnMap["pw_fill"] = pwFill
+// FnMap["pw_extract_text"] = pwExtractText
+// FnMap["pw_screenshot"] = pwScreenshot
+// FnMap["pw_screenshot_and_view"] = pwScreenshotAndView
+// FnMap["pw_wait_for_selector"] = pwWaitForSelector
+// FnMap["pw_drag"] = pwDrag
+// FnMap["pw_get_html"] = pwGetHTML
+// FnMap["pw_get_dom"] = pwGetDOM
+// FnMap["pw_search_elements"] = pwSearchElements
+// playwrightTools := []models.Tool{
+// {
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "pw_start",
+// Description: "Start a Playwright browser instance. Call this first before using other pw_ Uses headless mode by default (set PlaywrightHeadless=false in config for GUI).",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{},
+// Properties: map[string]models.ToolArgProps{},
+// },
+// },
+// },
+// {
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "pw_stop",
+// Description: "Stop the Playwright browser instance. Call when done with browser automation.",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{},
+// Properties: map[string]models.ToolArgProps{},
+// },
+// },
+// },
+// {
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "pw_is_running",
+// Description: "Check if Playwright browser is currently running.",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{},
+// Properties: map[string]models.ToolArgProps{},
+// },
+// },
+// },
+// {
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "pw_navigate",
+// Description: "Navigate to a URL in the browser.",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{"url"},
+// Properties: map[string]models.ToolArgProps{
+// "url": models.ToolArgProps{
+// Type: "string",
+// Description: "URL to navigate to",
+// },
+// },
+// },
+// },
+// },
+// {
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "pw_click",
+// Description: "Click on an element using CSS selector. Use 'index' for multiple matches (default 0).",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{"selector"},
+// Properties: map[string]models.ToolArgProps{
+// "selector": models.ToolArgProps{
+// Type: "string",
+// Description: "CSS selector for the element to click",
+// },
+// "index": models.ToolArgProps{
+// Type: "string",
+// Description: "optional index for multiple matches (default 0)",
+// },
+// },
+// },
+// },
+// },
+// {
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "pw_fill",
+// Description: "Fill an input field with text using CSS selector.",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{"selector", "text"},
+// Properties: map[string]models.ToolArgProps{
+// "selector": models.ToolArgProps{
+// Type: "string",
+// Description: "CSS selector for the input element",
+// },
+// "text": models.ToolArgProps{
+// Type: "string",
+// Description: "text to fill into the input",
+// },
+// "index": models.ToolArgProps{
+// Type: "string",
+// Description: "optional index for multiple matches (default 0)",
+// },
+// },
+// },
+// },
+// },
+// {
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "pw_extract_text",
+// Description: "Extract text content from the page or specific elements using CSS selector. Use 'body' for all page text.",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{"selector"},
+// Properties: map[string]models.ToolArgProps{
+// "selector": models.ToolArgProps{
+// Type: "string",
+// Description: "CSS selector (use 'body' for all page text)",
+// },
+// },
+// },
+// },
+// },
+// {
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "pw_screenshot",
+// Description: "Take a screenshot of the page or a specific element. Returns file path to saved image.",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{},
+// Properties: map[string]models.ToolArgProps{
+// "selector": models.ToolArgProps{
+// Type: "string",
+// Description: "optional CSS selector for element to screenshot",
+// },
+// "full_page": models.ToolArgProps{
+// Type: "string",
+// Description: "optional: 'true' to capture full page (default false)",
+// },
+// },
+// },
+// },
+// },
+// {
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "pw_screenshot_and_view",
+// Description: "Take a screenshot and return the image for viewing. Use when model needs to see the screenshot.",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{},
+// Properties: map[string]models.ToolArgProps{
+// "selector": models.ToolArgProps{
+// Type: "string",
+// Description: "optional CSS selector for element to screenshot",
+// },
+// "full_page": models.ToolArgProps{
+// Type: "string",
+// Description: "optional: 'true' to capture full page (default false)",
+// },
+// },
+// },
+// },
+// },
+// {
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "pw_wait_for_selector",
+// Description: "Wait for an element to appear on the page.",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{"selector"},
+// Properties: map[string]models.ToolArgProps{
+// "selector": models.ToolArgProps{
+// Type: "string",
+// Description: "CSS selector to wait for",
+// },
+// "timeout": models.ToolArgProps{
+// Type: "string",
+// Description: "optional timeout in ms (default 30000)",
+// },
+// },
+// },
+// },
+// },
+// {
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "pw_drag",
+// Description: "Drag the mouse from one point to another.",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{"x1", "y1", "x2", "y2"},
+// Properties: map[string]models.ToolArgProps{
+// "x1": models.ToolArgProps{
+// Type: "string",
+// Description: "starting X coordinate",
+// },
+// "y1": models.ToolArgProps{
+// Type: "string",
+// Description: "starting Y coordinate",
+// },
+// "x2": models.ToolArgProps{
+// Type: "string",
+// Description: "ending X coordinate",
+// },
+// "y2": models.ToolArgProps{
+// Type: "string",
+// Description: "ending Y coordinate",
+// },
+// },
+// },
+// },
+// },
+// {
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "pw_get_html",
+// Description: "Get the HTML content of the page or a specific element.",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{},
+// Properties: map[string]models.ToolArgProps{
+// "selector": models.ToolArgProps{
+// Type: "string",
+// Description: "optional CSS selector (default: body)",
+// },
+// },
+// },
+// },
+// },
+// {
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "pw_get_dom",
+// Description: "Get a structured DOM representation of an element with tag, attributes, text, and children.",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{},
+// Properties: map[string]models.ToolArgProps{
+// "selector": models.ToolArgProps{
+// Type: "string",
+// Description: "optional CSS selector (default: body)",
+// },
+// },
+// },
+// },
+// },
+// {
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "pw_search_elements",
+// Description: "Search for elements by text content or CSS selector. Returns matching elements with their tags, text, and HTML.",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{},
+// Properties: map[string]models.ToolArgProps{
+// "text": models.ToolArgProps{
+// Type: "string",
+// Description: "text to search for in elements",
+// },
+// "selector": models.ToolArgProps{
+// Type: "string",
+// Description: "CSS selector to search for",
+// },
+// },
+// },
+// },
+// },
+// }
+// BaseTools = append(BaseTools, playwrightTools...)
+// ToolSysMsg += browserToolSysMsg
+// agent.RegisterPWTool("pw_start", pwStart)
+// agent.RegisterPWTool("pw_stop", pwStop)
+// agent.RegisterPWTool("pw_is_running", pwIsRunning)
+// agent.RegisterPWTool("pw_navigate", pwNavigate)
+// agent.RegisterPWTool("pw_click", pwClick)
+// agent.RegisterPWTool("pw_click_at", pwClickAt)
+// agent.RegisterPWTool("pw_fill", pwFill)
+// agent.RegisterPWTool("pw_extract_text", pwExtractText)
+// agent.RegisterPWTool("pw_screenshot", pwScreenshot)
+// agent.RegisterPWTool("pw_screenshot_and_view", pwScreenshotAndView)
+// agent.RegisterPWTool("pw_wait_for_selector", pwWaitForSelector)
+// agent.RegisterPWTool("pw_drag", pwDrag)
+// agent.RegisterPWTool("pw_get_html", pwGetHTML)
+// agent.RegisterPWTool("pw_get_dom", pwGetDOM)
+// agent.RegisterPWTool("pw_search_elements", pwSearchElements)
+// browserAgentTool := []models.Tool{
+// {
+// Type: "function",
+// Function: models.ToolFunc{
+// Name: "browser_agent",
+// Description: "Autonomous browser automation agent. Use for complex multi-step browser tasks like 'go to website, login, and take screenshot'. The agent will plan and execute steps automatically using browser ",
+// Parameters: models.ToolFuncParams{
+// Type: "object",
+// Required: []string{"task"},
+// Properties: map[string]models.ToolArgProps{
+// "task": {Type: "string", Description: "The task to accomplish, e.g., 'go to github.com and take a screenshot of the homepage'"},
+// },
+// },
+// },
+// },
+// }
+// BaseTools = append(BaseTools, browserAgentTool...)
+// FnMap["browser_agent"] = tooler.runBrowserAgent
+// }
+// }
-func CallToolWithAgent(name string, args map[string]string) []byte {
+func CallToolWithAgent(name string, args map[string]string) ([]byte, bool) {
f, ok := FnMap[name]
if !ok {
- return []byte(fmt.Sprintf("tool %s not found", name))
+ return []byte(fmt.Sprintf("tool %s not found", name)), false
}
raw := f(args)
if a := agent.Get(name); a != nil {
- return a.Process(args, raw)
+ return a.Process(args, raw), true
}
- return raw
+ return raw, true
}
// openai style def
generated by cgit v1.2.3 (git 2.25.1) at 2026-03-19 09:38:50 +0000