1package tools
2
3import (
4 "context"
5 "encoding/json"
6 "fmt"
7 "io"
8 "net/http"
9 "strings"
10 "time"
11 "unicode/utf8"
12
13 md "github.com/JohannesKaufmann/html-to-markdown"
14 "github.com/PuerkitoBio/goquery"
15 "github.com/charmbracelet/crush/internal/permission"
16)
17
18type FetchParams struct {
19 URL string `json:"url"`
20 Format string `json:"format"`
21 Timeout int `json:"timeout,omitempty"`
22}
23
24type FetchPermissionsParams struct {
25 URL string `json:"url"`
26 Format string `json:"format"`
27 Timeout int `json:"timeout,omitempty"`
28}
29
30type fetchTool struct {
31 client *http.Client
32 permissions permission.Service
33 workingDir string
34}
35
36const (
37 FetchToolName = "fetch"
38 fetchToolDescription = `Fetches content from a URL and returns it in the specified format.
39
40WHEN TO USE THIS TOOL:
41- Use when you need to download content from a URL
42- Helpful for retrieving documentation, API responses, or web content
43- Useful for getting external information to assist with tasks
44
45HOW TO USE:
46- Provide the URL to fetch content from
47- Specify the desired output format (text, markdown, or html)
48- Optionally set a timeout for the request
49
50FEATURES:
51- Supports three output formats: text, markdown, and html
52- Automatically handles HTTP redirects
53- Sets reasonable timeouts to prevent hanging
54- Validates input parameters before making requests
55
56LIMITATIONS:
57- Maximum response size is 5MB
58- Only supports HTTP and HTTPS protocols
59- Cannot handle authentication or cookies
60- Some websites may block automated requests
61
62TIPS:
63- Use text format for plain text content or simple API responses
64- Use markdown format for content that should be rendered with formatting
65- Use html format when you need the raw HTML structure
66- Set appropriate timeouts for potentially slow websites`
67)
68
69func NewFetchTool(permissions permission.Service, workingDir string) BaseTool {
70 return &fetchTool{
71 client: &http.Client{
72 Timeout: 30 * time.Second,
73 Transport: &http.Transport{
74 MaxIdleConns: 100,
75 MaxIdleConnsPerHost: 10,
76 IdleConnTimeout: 90 * time.Second,
77 },
78 },
79 permissions: permissions,
80 workingDir: workingDir,
81 }
82}
83
84func (t *fetchTool) Name() string {
85 return FetchToolName
86}
87
88func (t *fetchTool) Info() ToolInfo {
89 return ToolInfo{
90 Name: FetchToolName,
91 Description: fetchToolDescription,
92 Parameters: map[string]any{
93 "url": map[string]any{
94 "type": "string",
95 "description": "The URL to fetch content from",
96 },
97 "format": map[string]any{
98 "type": "string",
99 "description": "The format to return the content in (text, markdown, or html)",
100 "enum": []string{"text", "markdown", "html"},
101 },
102 "timeout": map[string]any{
103 "type": "number",
104 "description": "Optional timeout in seconds (max 120)",
105 },
106 },
107 Required: []string{"url", "format"},
108 }
109}
110
111func (t *fetchTool) Run(ctx context.Context, call ToolCall) (ToolResponse, error) {
112 var params FetchParams
113 if err := json.Unmarshal([]byte(call.Input), ¶ms); err != nil {
114 return NewTextErrorResponse("Failed to parse fetch parameters: " + err.Error()), nil
115 }
116
117 if params.URL == "" {
118 return NewTextErrorResponse("URL parameter is required"), nil
119 }
120
121 format := strings.ToLower(params.Format)
122 if format != "text" && format != "markdown" && format != "html" {
123 return NewTextErrorResponse("Format must be one of: text, markdown, html"), nil
124 }
125
126 if !strings.HasPrefix(params.URL, "http://") && !strings.HasPrefix(params.URL, "https://") {
127 return NewTextErrorResponse("URL must start with http:// or https://"), nil
128 }
129
130 sessionID, messageID := GetContextValues(ctx)
131 if sessionID == "" || messageID == "" {
132 return ToolResponse{}, fmt.Errorf("session ID and message ID are required for creating a new file")
133 }
134
135 p := t.permissions.Request(
136 permission.CreatePermissionRequest{
137 SessionID: sessionID,
138 Path: t.workingDir,
139 ToolName: FetchToolName,
140 Action: "fetch",
141 Description: fmt.Sprintf("Fetch content from URL: %s", params.URL),
142 Params: FetchPermissionsParams(params),
143 },
144 )
145
146 if !p {
147 return ToolResponse{}, permission.ErrorPermissionDenied
148 }
149
150 // Handle timeout with context
151 requestCtx := ctx
152 if params.Timeout > 0 {
153 maxTimeout := 120 // 2 minutes
154 if params.Timeout > maxTimeout {
155 params.Timeout = maxTimeout
156 }
157 var cancel context.CancelFunc
158 requestCtx, cancel = context.WithTimeout(ctx, time.Duration(params.Timeout)*time.Second)
159 defer cancel()
160 }
161
162 req, err := http.NewRequestWithContext(requestCtx, "GET", params.URL, nil)
163 if err != nil {
164 return ToolResponse{}, fmt.Errorf("failed to create request: %w", err)
165 }
166
167 req.Header.Set("User-Agent", "crush/1.0")
168
169 resp, err := t.client.Do(req)
170 if err != nil {
171 return ToolResponse{}, fmt.Errorf("failed to fetch URL: %w", err)
172 }
173 defer resp.Body.Close()
174
175 if resp.StatusCode != http.StatusOK {
176 return NewTextErrorResponse(fmt.Sprintf("Request failed with status code: %d", resp.StatusCode)), nil
177 }
178
179 maxSize := int64(5 * 1024 * 1024) // 5MB
180 body, err := io.ReadAll(io.LimitReader(resp.Body, maxSize))
181 if err != nil {
182 return NewTextErrorResponse("Failed to read response body: " + err.Error()), nil
183 }
184
185 content := string(body)
186
187 isValidUt8 := utf8.ValidString(content)
188 if !isValidUt8 {
189 return NewTextErrorResponse("Response content is not valid UTF-8"), nil
190 }
191 contentType := resp.Header.Get("Content-Type")
192
193 switch format {
194 case "text":
195 if strings.Contains(contentType, "text/html") {
196 text, err := extractTextFromHTML(content)
197 if err != nil {
198 return NewTextErrorResponse("Failed to extract text from HTML: " + err.Error()), nil
199 }
200 content = text
201 }
202
203 case "markdown":
204 if strings.Contains(contentType, "text/html") {
205 markdown, err := convertHTMLToMarkdown(content)
206 if err != nil {
207 return NewTextErrorResponse("Failed to convert HTML to Markdown: " + err.Error()), nil
208 }
209 content = markdown
210 }
211
212 content = "```\n" + content + "\n```"
213
214 case "html":
215 // return only the body of the HTML document
216 if strings.Contains(contentType, "text/html") {
217 doc, err := goquery.NewDocumentFromReader(strings.NewReader(content))
218 if err != nil {
219 return NewTextErrorResponse("Failed to parse HTML: " + err.Error()), nil
220 }
221 body, err := doc.Find("body").Html()
222 if err != nil {
223 return NewTextErrorResponse("Failed to extract body from HTML: " + err.Error()), nil
224 }
225 if body == "" {
226 return NewTextErrorResponse("No body content found in HTML"), nil
227 }
228 content = "<html>\n<body>\n" + body + "\n</body>\n</html>"
229 }
230 }
231 // calculate byte size of content
232 contentSize := int64(len(content))
233 if contentSize > MaxReadSize {
234 content = content[:MaxReadSize]
235 content += fmt.Sprintf("\n\n[Content truncated to %d bytes]", MaxReadSize)
236 }
237
238 return NewTextResponse(content), nil
239}
240
241func extractTextFromHTML(html string) (string, error) {
242 doc, err := goquery.NewDocumentFromReader(strings.NewReader(html))
243 if err != nil {
244 return "", err
245 }
246
247 text := doc.Find("body").Text()
248 text = strings.Join(strings.Fields(text), " ")
249
250 return text, nil
251}
252
253func convertHTMLToMarkdown(html string) (string, error) {
254 converter := md.NewConverter("", true, nil)
255
256 markdown, err := converter.ConvertString(html)
257 if err != nil {
258 return "", err
259 }
260
261 return markdown, nil
262}