1use std::cell::RefCell;
2use std::rc::Rc;
3use std::sync::Arc;
4use std::sync::atomic::AtomicBool;
5
6use anyhow::{Context, Result, anyhow, bail};
7use assistant_slash_command::{
8 ArgumentCompletion, SlashCommand, SlashCommandOutput, SlashCommandOutputSection,
9 SlashCommandResult,
10};
11use futures::AsyncReadExt;
12use gpui::{Task, WeakEntity};
13use html_to_markdown::{TagHandler, convert_html_to_markdown, markdown};
14use http_client::{AsyncBody, HttpClient, HttpClientWithUrl};
15use language::{BufferSnapshot, LspAdapterDelegate};
16use ui::prelude::*;
17use workspace::Workspace;
18
19#[derive(Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Clone, Copy)]
20enum ContentType {
21 Html,
22 Plaintext,
23 Json,
24}
25
26pub struct FetchSlashCommand;
27
28impl FetchSlashCommand {
29 async fn build_message(http_client: Arc<HttpClientWithUrl>, url: &str) -> Result<String> {
30 let mut url = url.to_owned();
31 if !url.starts_with("https://") && !url.starts_with("http://") {
32 url = format!("https://{url}");
33 }
34
35 let mut response = http_client.get(&url, AsyncBody::default(), true).await?;
36
37 let mut body = Vec::new();
38 response
39 .body_mut()
40 .read_to_end(&mut body)
41 .await
42 .context("error reading response body")?;
43
44 if response.status().is_client_error() {
45 let text = String::from_utf8_lossy(body.as_slice());
46 bail!(
47 "status error {}, response: {text:?}",
48 response.status().as_u16()
49 );
50 }
51
52 let Some(content_type) = response.headers().get("content-type") else {
53 bail!("missing Content-Type header");
54 };
55 let content_type = content_type
56 .to_str()
57 .context("invalid Content-Type header")?;
58 let content_type = match content_type {
59 "text/html" => ContentType::Html,
60 "text/plain" => ContentType::Plaintext,
61 "application/json" => ContentType::Json,
62 _ => ContentType::Html,
63 };
64
65 match content_type {
66 ContentType::Html => {
67 let mut handlers: Vec<TagHandler> = vec![
68 Rc::new(RefCell::new(markdown::WebpageChromeRemover)),
69 Rc::new(RefCell::new(markdown::ParagraphHandler)),
70 Rc::new(RefCell::new(markdown::HeadingHandler)),
71 Rc::new(RefCell::new(markdown::ListHandler)),
72 Rc::new(RefCell::new(markdown::TableHandler::new())),
73 Rc::new(RefCell::new(markdown::StyledTextHandler)),
74 ];
75 if url.contains("wikipedia.org") {
76 use html_to_markdown::structure::wikipedia;
77
78 handlers.push(Rc::new(RefCell::new(wikipedia::WikipediaChromeRemover)));
79 handlers.push(Rc::new(RefCell::new(wikipedia::WikipediaInfoboxHandler)));
80 handlers.push(Rc::new(
81 RefCell::new(wikipedia::WikipediaCodeHandler::new()),
82 ));
83 } else {
84 handlers.push(Rc::new(RefCell::new(markdown::CodeHandler)));
85 }
86
87 convert_html_to_markdown(&body[..], &mut handlers)
88 }
89 ContentType::Plaintext => Ok(std::str::from_utf8(&body)?.to_owned()),
90 ContentType::Json => {
91 let json: serde_json::Value = serde_json::from_slice(&body)?;
92
93 Ok(format!(
94 "```json\n{}\n```",
95 serde_json::to_string_pretty(&json)?
96 ))
97 }
98 }
99 }
100}
101
102impl SlashCommand for FetchSlashCommand {
103 fn name(&self) -> String {
104 "fetch".into()
105 }
106
107 fn description(&self) -> String {
108 "Insert fetched URL contents".into()
109 }
110
111 fn icon(&self) -> IconName {
112 IconName::Globe
113 }
114
115 fn menu_text(&self) -> String {
116 self.description()
117 }
118
119 fn requires_argument(&self) -> bool {
120 true
121 }
122
123 fn complete_argument(
124 self: Arc<Self>,
125 _arguments: &[String],
126 _cancel: Arc<AtomicBool>,
127 _workspace: Option<WeakEntity<Workspace>>,
128 _window: &mut Window,
129 _cx: &mut App,
130 ) -> Task<Result<Vec<ArgumentCompletion>>> {
131 Task::ready(Ok(Vec::new()))
132 }
133
134 fn run(
135 self: Arc<Self>,
136 arguments: &[String],
137 _context_slash_command_output_sections: &[SlashCommandOutputSection<language::Anchor>],
138 _context_buffer: BufferSnapshot,
139 workspace: WeakEntity<Workspace>,
140 _delegate: Option<Arc<dyn LspAdapterDelegate>>,
141 _: &mut Window,
142 cx: &mut App,
143 ) -> Task<SlashCommandResult> {
144 let Some(argument) = arguments.first() else {
145 return Task::ready(Err(anyhow!("missing URL")));
146 };
147 let Some(workspace) = workspace.upgrade() else {
148 return Task::ready(Err(anyhow!("workspace was dropped")));
149 };
150
151 let http_client = workspace.read(cx).client().http_client();
152 let url = argument.to_string();
153
154 let text = cx.background_spawn({
155 let url = url.clone();
156 async move { Self::build_message(http_client, &url).await }
157 });
158
159 let url = SharedString::from(url);
160 cx.foreground_executor().spawn(async move {
161 let text = text.await?;
162 if text.trim().is_empty() {
163 bail!("no textual content found");
164 }
165
166 let range = 0..text.len();
167 Ok(SlashCommandOutput {
168 text,
169 sections: vec![SlashCommandOutputSection {
170 range,
171 icon: IconName::Globe,
172 label: format!("fetch {}", url).into(),
173 metadata: None,
174 }],
175 run_commands_in_text: false,
176 }
177 .to_event_stream())
178 })
179 }
180}