agent-file-tools 0.18.1

Agent File Tools — tree-sitter powered code analysis for AI agents
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
use std::collections::HashMap;
use std::fs;
use std::io::Read;
use std::path::{Path, PathBuf};
use std::process::{Command, Stdio};
use std::sync::mpsc;
use std::thread;
use std::time::{Duration, Instant, SystemTime, UNIX_EPOCH};

use serde::Deserialize;
use serde_json::json;

#[cfg(unix)]
use std::os::unix::process::CommandExt;

use crate::bash_background::process::terminate_process;
use crate::context::AppContext;
use crate::protocol::{
    ProgressFrame, ProgressKind, RawRequest, Response, ERROR_PERMISSION_REQUIRED,
};

const DEFAULT_TIMEOUT_MS: u64 = 120_000;
const INLINE_OUTPUT_LIMIT: usize = 30 * 1024;

#[derive(Debug, Deserialize)]
struct BashParams {
    command: String,
    #[serde(default)]
    timeout: Option<u64>,
    #[serde(default)]
    workdir: Option<PathBuf>,
    #[serde(default)]
    description: Option<String>,
    #[serde(default)]
    background: bool,
    #[serde(default = "default_compressed")]
    compressed: bool,
    #[serde(default)]
    permissions_granted: Vec<String>,
    #[serde(default)]
    permissions_requested: bool,
    #[serde(default)]
    env: HashMap<String, String>,
}

struct ExecutionResult {
    output: String,
    exit_code: i32,
    duration_ms: u64,
    timed_out: bool,
}

enum OutputEvent {
    Stdout(Vec<u8>),
    Stderr(Vec<u8>),
}

pub fn handle(req: &RawRequest, ctx: &AppContext) -> Response {
    let raw_params = req
        .params
        .get("params")
        .cloned()
        .unwrap_or_else(|| req.params.clone());
    let params = match serde_json::from_value::<BashParams>(raw_params) {
        Ok(params) => params,
        Err(e) => {
            return Response::error(
                &req.id,
                "invalid_request",
                format!("bash: invalid params: {e}"),
            );
        }
    };

    if let Some(description) = params.description.as_deref() {
        log::debug!("bash description: {description}");
    }

    let workdir = params
        .workdir
        .clone()
        .unwrap_or_else(|| default_workdir(ctx));
    let permission_asks = if params.permissions_requested || ctx.config().bash_permissions {
        crate::bash_permissions::scan::scan_with_cwd(&params.command, ctx, &workdir)
    } else {
        Vec::new()
    };
    if !permission_asks.is_empty()
        && !permissions_granted_cover(&permission_asks, &params.permissions_granted)
    {
        return Response::error_with_data(
            &req.id,
            ERROR_PERMISSION_REQUIRED,
            "bash command requires permission",
            json!({ "asks": permission_asks }),
        );
    }

    if params.background {
        let workdir = params.workdir.clone();
        let env = (!params.env.is_empty()).then_some(params.env.clone());
        return crate::bash_background::spawn(
            &req.id,
            req.session(),
            &params.command,
            workdir,
            env,
            params.timeout,
            ctx,
        );
    }

    if let Some(mut response) =
        crate::bash_rewrite::try_rewrite(&params.command, req.session_id.as_deref(), ctx)
    {
        // Rewriter rules build their own internal request with a placeholder id
        // (e.g. "bash_rewrite") to call into read/grep/glob handlers. Stamp the
        // original bash request id back onto the response so the bridge correlates
        // it with the in-flight `send()` instead of timing out.
        response.id = req.id.clone();
        return response;
    }

    let timeout = params.timeout.map(Duration::from_millis);
    let mut result = match spawn_command(
        &req.id,
        &params.command,
        &workdir,
        timeout,
        &params.env,
        ctx,
    ) {
        Ok(result) => result,
        Err(message) => return Response::error(&req.id, "execution_failed", message),
    };

    if params.compressed {
        result.output = crate::compress::compress(&params.command, result.output, ctx);
    }

    let (output, truncated, output_path) = match maybe_truncate(&result.output, ctx) {
        Ok(truncated) => truncated,
        Err(message) => return Response::error(&req.id, "output_write_failed", message),
    };

    Response::success(
        &req.id,
        json!({
            "output": output,
            "exit_code": result.exit_code,
            "duration_ms": result.duration_ms,
            "truncated": truncated,
            "output_path": output_path,
            "timed_out": result.timed_out,
        }),
    )
}

fn permissions_granted_cover(
    asks: &[crate::bash_permissions::PermissionAsk],
    granted: &[String],
) -> bool {
    if asks.is_empty() {
        return true;
    }
    if granted.is_empty() {
        return false;
    }

    asks.iter().all(|ask| {
        ask.patterns
            .iter()
            .chain(ask.always.iter())
            .any(|pattern| granted.iter().any(|grant| grant == pattern))
    })
}

fn default_compressed() -> bool {
    true
}

fn default_workdir(ctx: &AppContext) -> PathBuf {
    // Prefer the configured project root so bash commands run against the
    // user's project rather than the (often unrelated) cwd of the long-lived
    // aft worker process. Falls back to process cwd only when no project root
    // is configured (e.g. direct CLI usage).
    if let Some(root) = ctx.config().project_root.clone() {
        return root;
    }
    std::env::current_dir().unwrap_or_else(|_| PathBuf::from("."))
}

fn spawn_command(
    request_id: &str,
    command: &str,
    workdir: &Path,
    timeout: Option<Duration>,
    env: &HashMap<String, String>,
    ctx: &AppContext,
) -> Result<ExecutionResult, String> {
    let started = Instant::now();
    let mut child = shell_command(command)
        .current_dir(workdir)
        .envs(env)
        .stdout(Stdio::piped())
        .stderr(Stdio::piped())
        .spawn()
        .map_err(|e| format!("failed to spawn bash command: {e}"))?;

    let stdout = child.stdout.take().ok_or("failed to capture stdout")?;
    let stderr = child.stderr.take().ok_or("failed to capture stderr")?;
    let (tx, rx) = mpsc::channel::<OutputEvent>();
    let stdout_reader = spawn_reader(stdout, tx.clone(), true);
    let stderr_reader = spawn_reader(stderr, tx, false);

    let effective_timeout = timeout.unwrap_or(Duration::from_millis(DEFAULT_TIMEOUT_MS));
    let mut timed_out = false;
    let mut output = String::new();
    let status = loop {
        match child.try_wait() {
            Ok(Some(status)) => break status,
            Ok(None) => {
                drain_output_events(request_id, ctx, &rx, &mut output);
                if started.elapsed() >= effective_timeout {
                    timed_out = true;
                    terminate_process(&mut child);
                    break child
                        .wait()
                        .map_err(|e| format!("failed to wait after timeout: {e}"))?;
                }
                thread::sleep(Duration::from_millis(10));
            }
            Err(e) => return Err(format!("failed to poll bash command: {e}")),
        }
    };

    let _ = stdout_reader.join();
    let _ = stderr_reader.join();
    drain_output_events(request_id, ctx, &rx, &mut output);

    Ok(ExecutionResult {
        output,
        exit_code: status.code().unwrap_or(if timed_out { 124 } else { -1 }),
        duration_ms: started.elapsed().as_millis().try_into().unwrap_or(u64::MAX),
        timed_out,
    })
}

#[cfg(windows)]
fn shell_command(command: &str) -> Command {
    let mut cmd = Command::new("powershell.exe");
    cmd.args([
        "-NoLogo",
        "-NoProfile",
        "-ExecutionPolicy",
        "Bypass",
        "-Command",
        command,
    ]);
    cmd
}

#[cfg(not(windows))]
fn shell_command(command: &str) -> Command {
    let mut cmd = Command::new("/bin/sh");
    cmd.args(["-c", command]);
    cmd.process_group(0);
    cmd
}

fn spawn_reader<R>(
    mut reader: R,
    tx: mpsc::Sender<OutputEvent>,
    is_stdout: bool,
) -> thread::JoinHandle<()>
where
    R: Read + Send + 'static,
{
    let handle = thread::spawn(move || {
        let mut buffer = [0_u8; 8192];
        loop {
            match reader.read(&mut buffer) {
                Ok(0) => break,
                Ok(n) => {
                    let bytes = buffer[..n].to_vec();
                    let event = if is_stdout {
                        OutputEvent::Stdout(bytes)
                    } else {
                        OutputEvent::Stderr(bytes)
                    };
                    if tx.send(event).is_err() {
                        break;
                    }
                }
                Err(_) => break,
            }
        }
    });
    handle
}

fn drain_output_events(
    request_id: &str,
    ctx: &AppContext,
    rx: &mpsc::Receiver<OutputEvent>,
    output: &mut String,
) {
    while let Ok(event) = rx.try_recv() {
        let (kind, bytes) = match event {
            OutputEvent::Stdout(bytes) => (ProgressKind::Stdout, bytes),
            OutputEvent::Stderr(bytes) => (ProgressKind::Stderr, bytes),
        };
        let chunk = String::from_utf8_lossy(&bytes).into_owned();
        output.push_str(&chunk);
        ctx.emit_progress(ProgressFrame::new(request_id, kind, chunk));
    }
}

fn maybe_truncate(
    output: &str,
    ctx: &AppContext,
) -> Result<(String, bool, Option<String>), String> {
    if output.len() <= INLINE_OUTPUT_LIMIT {
        return Ok((output.to_string(), false, None));
    }

    let dir = bash_output_dir(ctx);
    fs::create_dir_all(&dir).map_err(|e| {
        format!(
            "failed to create bash output directory {}: {e}",
            dir.display()
        )
    })?;
    let path = dir.join(format!("{}.txt", random_id()));
    fs::write(&path, output)
        .map_err(|e| format!("failed to write bash output {}: {e}", path.display()))?;

    let start = inline_output_suffix_start(output);
    Ok((
        output[start..].to_string(),
        true,
        Some(path.display().to_string()),
    ))
}

/// Compute the byte index where the last `INLINE_OUTPUT_LIMIT` bytes of
/// `output` start, snapped forward to a UTF-8 character boundary so we
/// never split a multi-byte char.
///
/// The earlier implementation walked `char_indices().rev().find_map(...)`,
/// which returned the LAST char's start index on the very first iteration
/// (because `output.len() - idx == 1 <= INLINE_OUTPUT_LIMIT`). That bug
/// made the inline preview a single character for any output above the
/// limit. This helper computes the suffix start by byte arithmetic and
/// keeps approximately `INLINE_OUTPUT_LIMIT` trailing bytes intact.
fn inline_output_suffix_start(output: &str) -> usize {
    let mut start = output.len().saturating_sub(INLINE_OUTPUT_LIMIT);
    while start < output.len() && !output.is_char_boundary(start) {
        start += 1;
    }
    start
}

fn bash_output_dir(ctx: &AppContext) -> PathBuf {
    if let Some(dir) = std::env::var_os("AFT_CACHE_DIR") {
        return PathBuf::from(dir).join("aft").join("bash-output");
    }
    if let Some(dir) = ctx.config().storage_dir.clone() {
        return dir.join("bash-output");
    }
    let home = std::env::var_os("HOME")
        .map(PathBuf::from)
        .unwrap_or_else(|| PathBuf::from("."));
    home.join(".cache").join("aft").join("bash-output")
}

fn random_id() -> String {
    let nanos = SystemTime::now()
        .duration_since(UNIX_EPOCH)
        .map(|duration| duration.as_nanos())
        .unwrap_or(0);
    format!("{}-{}", std::process::id(), nanos)
}

#[cfg(test)]
mod tests {
    use super::*;

    /// Regression: prior reverse `char_indices` logic returned only the LAST
    /// character of `output` because the first reverse-iteration index already
    /// satisfied `output.len() - idx == 1 <= INLINE_OUTPUT_LIMIT`. The new
    /// implementation must keep approximately `INLINE_OUTPUT_LIMIT` trailing
    /// bytes intact for ASCII input.
    #[test]
    fn inline_output_suffix_keeps_full_limit_for_ascii() {
        let total = INLINE_OUTPUT_LIMIT * 2;
        let output: String = "x".repeat(total);
        let start = inline_output_suffix_start(&output);
        let suffix_len = output.len() - start;
        assert!(
            suffix_len > INLINE_OUTPUT_LIMIT / 2,
            "ascii suffix too short: got {suffix_len} bytes (limit={INLINE_OUTPUT_LIMIT})"
        );
        assert!(
            suffix_len <= INLINE_OUTPUT_LIMIT,
            "ascii suffix exceeded limit: got {suffix_len} bytes (limit={INLINE_OUTPUT_LIMIT})"
        );
        // Guard against a regression to the 1-char bug.
        assert!(suffix_len > 1, "suffix collapsed to a single character");
    }

    /// The suffix-start index must always land on a UTF-8 char boundary so
    /// `output[start..]` is a valid `&str`. Multi-byte chars (like 4-byte
    /// emoji) require boundary snapping when the raw byte split lands inside
    /// a code point.
    #[test]
    fn inline_output_suffix_respects_utf8_boundaries() {
        // Each crab is 4 bytes. 20_000 of them = 80_000 bytes, well over the
        // inline limit. The byte index `len - INLINE_OUTPUT_LIMIT` is unlikely
        // to be a 4-byte boundary.
        let output: String = "🦀".repeat(20_000);
        let start = inline_output_suffix_start(&output);
        assert!(
            output.is_char_boundary(start),
            "suffix split a multi-byte char"
        );
        // Slicing must succeed without panic.
        let suffix = &output[start..];
        let suffix_bytes = suffix.len();
        assert!(
            suffix_bytes <= INLINE_OUTPUT_LIMIT + 4,
            "utf8 suffix far above limit: got {suffix_bytes} bytes (limit={INLINE_OUTPUT_LIMIT})"
        );
        assert!(
            suffix_bytes > INLINE_OUTPUT_LIMIT / 2,
            "utf8 suffix too short: got {suffix_bytes} bytes (limit={INLINE_OUTPUT_LIMIT})"
        );
    }

    /// Output below the inline limit is returned by `maybe_truncate` directly,
    /// but the helper must still return `0` so callers slicing `output[start..]`
    /// get the full string.
    #[test]
    fn inline_output_suffix_returns_zero_for_short_input() {
        let output = "small";
        assert_eq!(inline_output_suffix_start(output), 0);
    }
}