1package tools
  2
  3import (
  4	"context"
  5	"encoding/json"
  6	"fmt"
  7	"io"
  8	"net/http"
  9	"strings"
 10	"time"
 11
 12	md "github.com/JohannesKaufmann/html-to-markdown"
 13	"github.com/PuerkitoBio/goquery"
 14	"github.com/charmbracelet/crush/internal/config"
 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}
 34
 35const (
 36	FetchToolName        = "fetch"
 37	fetchToolDescription = `Fetches content from a URL and returns it in the specified format.
 38
 39WHEN TO USE THIS TOOL:
 40- Use when you need to download content from a URL
 41- Helpful for retrieving documentation, API responses, or web content
 42- Useful for getting external information to assist with tasks
 43
 44HOW TO USE:
 45- Provide the URL to fetch content from
 46- Specify the desired output format (text, markdown, or html)
 47- Optionally set a timeout for the request
 48
 49FEATURES:
 50- Supports three output formats: text, markdown, and html
 51- Automatically handles HTTP redirects
 52- Sets reasonable timeouts to prevent hanging
 53- Validates input parameters before making requests
 54
 55LIMITATIONS:
 56- Maximum response size is 5MB
 57- Only supports HTTP and HTTPS protocols
 58- Cannot handle authentication or cookies
 59- Some websites may block automated requests
 60
 61TIPS:
 62- Use text format for plain text content or simple API responses
 63- Use markdown format for content that should be rendered with formatting
 64- Use html format when you need the raw HTML structure
 65- Set appropriate timeouts for potentially slow websites`
 66)
 67
 68func NewFetchTool(permissions permission.Service) BaseTool {
 69	return &fetchTool{
 70		client: &http.Client{
 71			Timeout: 30 * time.Second,
 72			Transport: &http.Transport{
 73				MaxIdleConns:        100,
 74				MaxIdleConnsPerHost: 10,
 75				IdleConnTimeout:     90 * time.Second,
 76			},
 77		},
 78		permissions: permissions,
 79	}
 80}
 81
 82func (t *fetchTool) Name() string {
 83	return FetchToolName
 84}
 85
 86func (t *fetchTool) Info() ToolInfo {
 87	return ToolInfo{
 88		Name:        FetchToolName,
 89		Description: fetchToolDescription,
 90		Parameters: map[string]any{
 91			"url": map[string]any{
 92				"type":        "string",
 93				"description": "The URL to fetch content from",
 94			},
 95			"format": map[string]any{
 96				"type":        "string",
 97				"description": "The format to return the content in (text, markdown, or html)",
 98				"enum":        []string{"text", "markdown", "html"},
 99			},
100			"timeout": map[string]any{
101				"type":        "number",
102				"description": "Optional timeout in seconds (max 120)",
103			},
104		},
105		Required: []string{"url", "format"},
106	}
107}
108
109func (t *fetchTool) Run(ctx context.Context, call ToolCall) (ToolResponse, error) {
110	var params FetchParams
111	if err := json.Unmarshal([]byte(call.Input), ¶ms); err != nil {
112		return NewTextErrorResponse("Failed to parse fetch parameters: " + err.Error()), nil
113	}
114
115	if params.URL == "" {
116		return NewTextErrorResponse("URL parameter is required"), nil
117	}
118
119	format := strings.ToLower(params.Format)
120	if format != "text" && format != "markdown" && format != "html" {
121		return NewTextErrorResponse("Format must be one of: text, markdown, html"), nil
122	}
123
124	if !strings.HasPrefix(params.URL, "http://") && !strings.HasPrefix(params.URL, "https://") {
125		return NewTextErrorResponse("URL must start with http:// or https://"), nil
126	}
127
128	sessionID, messageID := GetContextValues(ctx)
129	if sessionID == "" || messageID == "" {
130		return ToolResponse{}, fmt.Errorf("session ID and message ID are required for creating a new file")
131	}
132
133	p := t.permissions.Request(
134		permission.CreatePermissionRequest{
135			SessionID:   sessionID,
136			Path:        config.WorkingDirectory(),
137			ToolName:    FetchToolName,
138			Action:      "fetch",
139			Description: fmt.Sprintf("Fetch content from URL: %s", params.URL),
140			Params:      FetchPermissionsParams(params),
141		},
142	)
143
144	if !p {
145		return ToolResponse{}, permission.ErrorPermissionDenied
146	}
147
148	// Handle timeout with context
149	requestCtx := ctx
150	if params.Timeout > 0 {
151		maxTimeout := 120 // 2 minutes
152		if params.Timeout > maxTimeout {
153			params.Timeout = maxTimeout
154		}
155		var cancel context.CancelFunc
156		requestCtx, cancel = context.WithTimeout(ctx, time.Duration(params.Timeout)*time.Second)
157		defer cancel()
158	}
159
160	req, err := http.NewRequestWithContext(requestCtx, "GET", params.URL, nil)
161	if err != nil {
162		return ToolResponse{}, fmt.Errorf("failed to create request: %w", err)
163	}
164
165	req.Header.Set("User-Agent", "crush/1.0")
166
167	resp, err := t.client.Do(req)
168	if err != nil {
169		return ToolResponse{}, fmt.Errorf("failed to fetch URL: %w", err)
170	}
171	defer resp.Body.Close()
172
173	if resp.StatusCode != http.StatusOK {
174		return NewTextErrorResponse(fmt.Sprintf("Request failed with status code: %d", resp.StatusCode)), nil
175	}
176
177	maxSize := int64(5 * 1024 * 1024) // 5MB
178	body, err := io.ReadAll(io.LimitReader(resp.Body, maxSize))
179	if err != nil {
180		return NewTextErrorResponse("Failed to read response body: " + err.Error()), nil
181	}
182
183	content := string(body)
184	contentType := resp.Header.Get("Content-Type")
185
186	switch format {
187	case "text":
188		if strings.Contains(contentType, "text/html") {
189			text, err := extractTextFromHTML(content)
190			if err != nil {
191				return NewTextErrorResponse("Failed to extract text from HTML: " + err.Error()), nil
192			}
193			return NewTextResponse(text), nil
194		}
195		return NewTextResponse(content), nil
196
197	case "markdown":
198		if strings.Contains(contentType, "text/html") {
199			markdown, err := convertHTMLToMarkdown(content)
200			if err != nil {
201				return NewTextErrorResponse("Failed to convert HTML to Markdown: " + err.Error()), nil
202			}
203			return NewTextResponse(markdown), nil
204		}
205
206		return NewTextResponse("```\n" + content + "\n```"), nil
207
208	case "html":
209		return NewTextResponse(content), nil
210
211	default:
212		return NewTextResponse(content), nil
213	}
214}
215
216func extractTextFromHTML(html string) (string, error) {
217	doc, err := goquery.NewDocumentFromReader(strings.NewReader(html))
218	if err != nil {
219		return "", err
220	}
221
222	text := doc.Text()
223	text = strings.Join(strings.Fields(text), " ")
224
225	return text, nil
226}
227
228func convertHTMLToMarkdown(html string) (string, error) {
229	converter := md.NewConverter("", true, nil)
230
231	markdown, err := converter.ConvertString(html)
232	if err != nil {
233		return "", err
234	}
235
236	return markdown, nil
237}