feature/acp-session-list-cursor
claude/github-issue-40-0au5ua
claude/linked-issue-acp-tui-u4vvpa
claude/minor-release-cut-vkugow
claude/sigit-acp-local-chat-cx6380
claude/sigit-cloud-agent-expansion-reox0a
claude/tool-permission-system
claude/zen-feynman-0u78dk
development
feature/acp-cancel-request
feature/acp-session-list-cursor
feature/acp-v1-conformance
feature/agent-tools-multiedit-glob-todos-remember
feature/cargo-workspace
feature/hooks
feature/load-local-model-explicitly
feature/sigit-agent-profile
feature/sigit-code-cloud
feature/xcode-mcp-server
fix/acp-replay-loaded-session-history
fix/acp-replay-reasoning-and-attachments
fix/dsml-doubled-bar-markers
main
release/v1.5.10
release/v1.5.6
release/v1.6.1
| 1 | //! The stop reason a prompt turn ends with (issue #140). |
| 2 | //! |
| 3 | //! ACP defines five and the agent only ever sent `end_turn` and `cancelled`, |
| 4 | //! so a client could not tell a finished answer from one cut off at the token |
| 5 | //! limit, withheld by the endpoint, or stopped at the tool-round cap. The |
| 6 | //! `cancelled` case is covered in `acp_permissions.rs`; this file drives the |
| 7 | //! other three through a scripted endpoint. |
| 8 | |
| 9 | use std::collections::VecDeque; |
| 10 | use std::io::{BufRead, BufReader, Read, Write}; |
| 11 | use std::net::TcpListener; |
| 12 | use std::process::{Child, ChildStdin, Command, Stdio}; |
| 13 | use std::sync::mpsc::{Receiver, channel}; |
| 14 | use std::sync::{Arc, Mutex}; |
| 15 | use std::time::{Duration, Instant}; |
| 16 | |
| 17 | use serde_json::{Value, json}; |
| 18 | |
| 19 | const TIMEOUT: Duration = Duration::from_secs(60); |
| 20 | |
| 21 | /// The built-in tool-round cap (`MAX_TOOL_ROUNDS` in `main.rs`). |
| 22 | const TOOL_ROUND_CAP: usize = 24; |
| 23 | |
| 24 | /// A streamed text reply that ends with `finish_reason`. |
| 25 | fn sse_text(text: &str, finish_reason: &str) -> String { |
| 26 | format!( |
| 27 | "data: {}\n\ndata: {}\n\ndata: [DONE]\n\n", |
| 28 | json!({"choices": [{"delta": {"content": text}}]}), |
| 29 | json!({"choices": [{"delta": {}, "finish_reason": finish_reason}]}), |
| 30 | ) |
| 31 | } |
| 32 | |
| 33 | fn sse_tool_call(id: &str, name: &str, arguments: &str) -> String { |
| 34 | format!( |
| 35 | "data: {}\n\ndata: {}\n\ndata: [DONE]\n\n", |
| 36 | json!({ |
| 37 | "choices": [{"delta": {"tool_calls": [{ |
| 38 | "index": 0, |
| 39 | "id": id, |
| 40 | "function": {"name": name, "arguments": arguments}, |
| 41 | }]}}] |
| 42 | }), |
| 43 | json!({"choices": [{"delta": {}, "finish_reason": "tool_calls"}]}), |
| 44 | ) |
| 45 | } |
| 46 | |
| 47 | /// A scripted endpoint: one SSE body per request, in order, then empty |
| 48 | /// replies. Every request body it receives is kept for the test to read. |
| 49 | struct FakeEndpoint { |
| 50 | port: u16, |
| 51 | requests: Arc<Mutex<Vec<Value>>>, |
| 52 | } |
| 53 | |
| 54 | fn start_fake_endpoint(replies: Vec<String>) -> FakeEndpoint { |
| 55 | let listener = TcpListener::bind("127.0.0.1:0").expect("bind fake endpoint"); |
| 56 | let port = listener.local_addr().unwrap().port(); |
| 57 | let queue = Mutex::new(VecDeque::from(replies)); |
| 58 | let requests = Arc::new(Mutex::new(Vec::new())); |
| 59 | let seen = Arc::clone(&requests); |
| 60 | |
| 61 | std::thread::spawn(move || { |
| 62 | for stream in listener.incoming() { |
| 63 | let Ok(mut stream) = stream else { continue }; |
| 64 | let mut reader = BufReader::new(match stream.try_clone() { |
| 65 | Ok(clone) => clone, |
| 66 | Err(_) => continue, |
| 67 | }); |
| 68 | let mut content_length = 0usize; |
| 69 | loop { |
| 70 | let mut line = String::new(); |
| 71 | if reader.read_line(&mut line).unwrap_or(0) == 0 { |
| 72 | break; |
| 73 | } |
| 74 | let line = line.trim(); |
| 75 | if line.is_empty() { |
| 76 | break; |
| 77 | } |
| 78 | if let Some(length) = line.to_ascii_lowercase().strip_prefix("content-length:") { |
| 79 | content_length = length.trim().parse().unwrap_or(0); |
| 80 | } |
| 81 | } |
| 82 | let mut body = vec![0u8; content_length]; |
| 83 | if reader.read_exact(&mut body).is_err() { |
| 84 | continue; |
| 85 | } |
| 86 | if let Ok(request) = serde_json::from_slice::<Value>(&body) { |
| 87 | seen.lock().unwrap().push(request); |
| 88 | } |
| 89 | let reply = queue |
| 90 | .lock() |
| 91 | .unwrap() |
| 92 | .pop_front() |
| 93 | .unwrap_or_else(|| "data: [DONE]\n\n".to_string()); |
| 94 | let response = format!( |
| 95 | "HTTP/1.1 200 OK\r\ncontent-type: text/event-stream\r\ncontent-length: {}\r\n\ |
| 96 | connection: close\r\n\r\n{}", |
| 97 | reply.len(), |
| 98 | reply |
| 99 | ); |
| 100 | let _ = stream.write_all(response.as_bytes()); |
| 101 | let _ = stream.flush(); |
| 102 | } |
| 103 | }); |
| 104 | |
| 105 | FakeEndpoint { port, requests } |
| 106 | } |
| 107 | |
| 108 | struct AgentUnderTest { |
| 109 | child: Child, |
| 110 | stdin: ChildStdin, |
| 111 | incoming: Receiver<Value>, |
| 112 | next_id: u64, |
| 113 | } |
| 114 | |
| 115 | fn spawn_agent(port: u16, config_dir: &std::path::Path) -> AgentUnderTest { |
| 116 | spawn_agent_with_env(port, config_dir, &[]) |
| 117 | } |
| 118 | |
| 119 | fn spawn_agent_with_env( |
| 120 | port: u16, |
| 121 | config_dir: &std::path::Path, |
| 122 | env: &[(&str, &str)], |
| 123 | ) -> AgentUnderTest { |
| 124 | let mut child = Command::new(env!("CARGO_BIN_EXE_sigit")) |
| 125 | .env("OPENAI_BASE_URL", format!("http://127.0.0.1:{port}")) |
| 126 | .env("OPENAI_API_KEY", "test-key") |
| 127 | .env("SIGIT_MODEL", "scripted-model") |
| 128 | .env("SIGIT_CONFIG_DIR", config_dir) |
| 129 | .env("SIGIT_MCP", "off") |
| 130 | .env("SIGIT_PERMISSIONS", "allow") |
| 131 | .env_remove("SIGIT_LOCAL_INFERENCE") |
| 132 | .env_remove("SIGIT_MAX_TOOL_ROUNDS") |
| 133 | .envs(env.iter().copied()) |
| 134 | .stdin(Stdio::piped()) |
| 135 | .stdout(Stdio::piped()) |
| 136 | .stderr(Stdio::null()) |
| 137 | .spawn() |
| 138 | .expect("spawn sigit in ACP mode"); |
| 139 | |
| 140 | let stdout = child.stdout.take().unwrap(); |
| 141 | let (message_tx, incoming) = channel(); |
| 142 | std::thread::spawn(move || { |
| 143 | for line in BufReader::new(stdout).lines() { |
| 144 | let Ok(line) = line else { break }; |
| 145 | if let Ok(message) = serde_json::from_str::<Value>(&line) |
| 146 | && message_tx.send(message).is_err() |
| 147 | { |
| 148 | break; |
| 149 | } |
| 150 | } |
| 151 | }); |
| 152 | |
| 153 | let stdin = child.stdin.take().unwrap(); |
| 154 | AgentUnderTest { |
| 155 | child, |
| 156 | stdin, |
| 157 | incoming, |
| 158 | next_id: 0, |
| 159 | } |
| 160 | } |
| 161 | |
| 162 | impl AgentUnderTest { |
| 163 | fn request(&mut self, method: &str, params: Value) -> u64 { |
| 164 | self.next_id += 1; |
| 165 | let id = self.next_id; |
| 166 | let mut line = |
| 167 | json!({"jsonrpc": "2.0", "id": id, "method": method, "params": params}).to_string(); |
| 168 | line.push('\n'); |
| 169 | self.stdin.write_all(line.as_bytes()).expect("write stdin"); |
| 170 | self.stdin.flush().expect("flush stdin"); |
| 171 | id |
| 172 | } |
| 173 | |
| 174 | fn wait_for_response(&mut self, id: u64) -> Value { |
| 175 | let deadline = Instant::now() + TIMEOUT; |
| 176 | loop { |
| 177 | let remaining = deadline.saturating_duration_since(Instant::now()); |
| 178 | let Ok(message) = self.incoming.recv_timeout(remaining) else { |
| 179 | panic!("timed out waiting for the response to request {id}"); |
| 180 | }; |
| 181 | if message["id"] == id && message.get("method").is_none() { |
| 182 | assert!( |
| 183 | message.get("error").is_none(), |
| 184 | "request {id} failed: {message}" |
| 185 | ); |
| 186 | return message; |
| 187 | } |
| 188 | } |
| 189 | } |
| 190 | |
| 191 | fn open_session(&mut self, cwd: &std::path::Path) -> String { |
| 192 | let id = self.request( |
| 193 | "initialize", |
| 194 | json!({"protocolVersion": 1, "clientCapabilities": {}}), |
| 195 | ); |
| 196 | self.wait_for_response(id); |
| 197 | |
| 198 | let id = self.request("session/new", json!({"cwd": cwd, "mcpServers": []})); |
| 199 | self.wait_for_response(id)["result"]["sessionId"] |
| 200 | .as_str() |
| 201 | .expect("session id") |
| 202 | .to_string() |
| 203 | } |
| 204 | |
| 205 | /// Send one text prompt and return the stop reason the turn ended with. |
| 206 | fn prompt(&mut self, session_id: &str, text: &str) -> String { |
| 207 | self.prompt_with_reply(session_id, text).0 |
| 208 | } |
| 209 | |
| 210 | /// Like [`Self::prompt`], also returning the assistant text the client |
| 211 | /// was sent on the way, joined the way a client renders it. |
| 212 | fn prompt_with_reply(&mut self, session_id: &str, text: &str) -> (String, String) { |
| 213 | let id = self.request( |
| 214 | "session/prompt", |
| 215 | json!({"sessionId": session_id, "prompt": [{"type": "text", "text": text}]}), |
| 216 | ); |
| 217 | let mut rendered = String::new(); |
| 218 | let deadline = Instant::now() + TIMEOUT; |
| 219 | loop { |
| 220 | let remaining = deadline.saturating_duration_since(Instant::now()); |
| 221 | let Ok(message) = self.incoming.recv_timeout(remaining) else { |
| 222 | panic!("timed out waiting for the response to request {id}"); |
| 223 | }; |
| 224 | let update = &message["params"]["update"]; |
| 225 | if message["method"] == "session/update" |
| 226 | && update["sessionUpdate"] == "agent_message_chunk" |
| 227 | && let Some(chunk) = update["content"]["text"].as_str() |
| 228 | { |
| 229 | rendered.push_str(chunk); |
| 230 | } |
| 231 | if message["id"] == id && message.get("method").is_none() { |
| 232 | assert!(message.get("error").is_none(), "prompt failed: {message}"); |
| 233 | let stop_reason = message["result"]["stopReason"] |
| 234 | .as_str() |
| 235 | .expect("stop reason") |
| 236 | .to_string(); |
| 237 | return (stop_reason, rendered); |
| 238 | } |
| 239 | } |
| 240 | } |
| 241 | } |
| 242 | |
| 243 | impl Drop for AgentUnderTest { |
| 244 | fn drop(&mut self) { |
| 245 | let _ = self.child.kill(); |
| 246 | let _ = self.child.wait(); |
| 247 | } |
| 248 | } |
| 249 | |
| 250 | fn scratch(name: &str) -> std::path::PathBuf { |
| 251 | let dir = std::env::temp_dir().join(format!("sigit_acp_stop_{name}_{}", std::process::id())); |
| 252 | std::fs::create_dir_all(dir.join("config")).unwrap(); |
| 253 | std::fs::create_dir_all(dir.join("work")).unwrap(); |
| 254 | dir |
| 255 | } |
| 256 | |
| 257 | /// Whether any message of a chat-completion request mentions `needle`. |
| 258 | fn request_mentions(request: &Value, needle: &str) -> bool { |
| 259 | request["messages"] |
| 260 | .as_array() |
| 261 | .is_some_and(|messages| messages.iter().any(|m| m.to_string().contains(needle))) |
| 262 | } |
| 263 | |
| 264 | #[test] |
| 265 | fn a_reply_cut_off_at_the_token_limit_ends_with_max_tokens() { |
| 266 | let dir = scratch("length"); |
| 267 | let endpoint = start_fake_endpoint(vec![sse_text("The answer is", "length")]); |
| 268 | let mut agent = spawn_agent(endpoint.port, &dir.join("config")); |
| 269 | let session_id = agent.open_session(&dir.join("work")); |
| 270 | |
| 271 | assert_eq!(agent.prompt(&session_id, "explain"), "max_tokens"); |
| 272 | |
| 273 | std::fs::remove_dir_all(&dir).ok(); |
| 274 | } |
| 275 | |
| 276 | #[test] |
| 277 | fn a_finished_reply_still_ends_with_end_turn() { |
| 278 | let dir = scratch("stop"); |
| 279 | let endpoint = start_fake_endpoint(vec![sse_text("Done.", "stop")]); |
| 280 | let mut agent = spawn_agent(endpoint.port, &dir.join("config")); |
| 281 | let session_id = agent.open_session(&dir.join("work")); |
| 282 | |
| 283 | assert_eq!(agent.prompt(&session_id, "explain"), "end_turn"); |
| 284 | |
| 285 | std::fs::remove_dir_all(&dir).ok(); |
| 286 | } |
| 287 | |
| 288 | /// A refusal is more than a label: ACP says the refused prompt and what |
| 289 | /// followed it are not part of the next prompt, and the client drops them from |
| 290 | /// its view on that promise. The agent has to drop them from history too. |
| 291 | #[test] |
| 292 | fn a_withheld_reply_ends_with_refusal_and_leaves_history() { |
| 293 | let dir = scratch("refusal"); |
| 294 | let endpoint = start_fake_endpoint(vec![ |
| 295 | sse_text("Sure.", "stop"), |
| 296 | sse_text("", "content_filter"), |
| 297 | sse_text("Here you go.", "stop"), |
| 298 | ]); |
| 299 | let mut agent = spawn_agent(endpoint.port, &dir.join("config")); |
| 300 | let session_id = agent.open_session(&dir.join("work")); |
| 301 | |
| 302 | assert_eq!(agent.prompt(&session_id, "first-kept-prompt"), "end_turn"); |
| 303 | assert_eq!( |
| 304 | agent.prompt(&session_id, "second-refused-prompt"), |
| 305 | "refusal" |
| 306 | ); |
| 307 | assert_eq!(agent.prompt(&session_id, "third-prompt"), "end_turn"); |
| 308 | |
| 309 | let requests = endpoint.requests.lock().unwrap(); |
| 310 | let last = requests |
| 311 | .last() |
| 312 | .expect("the third prompt reached the endpoint"); |
| 313 | assert!( |
| 314 | request_mentions(last, "first-kept-prompt"), |
| 315 | "the turn before the refusal stays in history: {last}" |
| 316 | ); |
| 317 | assert!( |
| 318 | !request_mentions(last, "second-refused-prompt"), |
| 319 | "the refused prompt must not be replayed: {last}" |
| 320 | ); |
| 321 | |
| 322 | std::fs::remove_dir_all(&dir).ok(); |
| 323 | } |
| 324 | |
| 325 | /// The case the reason exists for: the model keeps asking for tools until the |
| 326 | /// agent stops offering them. |
| 327 | #[test] |
| 328 | fn a_turn_stopped_at_the_tool_round_cap_ends_with_max_turn_requests() { |
| 329 | let dir = scratch("cap"); |
| 330 | // Each call differs, or the repeated-call guard ends the turn first. The |
| 331 | // commands stay quote-free: `cmd /C` mangles quotes on Windows. |
| 332 | let replies = (1..=TOOL_ROUND_CAP) |
| 333 | .map(|round| { |
| 334 | sse_tool_call( |
| 335 | &format!("call_{round}"), |
| 336 | "run_command", |
| 337 | &json!({"command": format!("echo round-{round}")}).to_string(), |
| 338 | ) |
| 339 | }) |
| 340 | .chain([sse_text("Out of rounds.", "stop")]) |
| 341 | .collect(); |
| 342 | let endpoint = start_fake_endpoint(replies); |
| 343 | let mut agent = spawn_agent(endpoint.port, &dir.join("config")); |
| 344 | let session_id = agent.open_session(&dir.join("work")); |
| 345 | |
| 346 | assert_eq!(agent.prompt(&session_id, "keep going"), "max_turn_requests"); |
| 347 | assert_eq!( |
| 348 | endpoint.requests.lock().unwrap().len(), |
| 349 | TOOL_ROUND_CAP + 1, |
| 350 | "one request per tool round, plus the forced closing reply" |
| 351 | ); |
| 352 | |
| 353 | std::fs::remove_dir_all(&dir).ok(); |
| 354 | } |
| 355 | |
| 356 | /// What issue #120 looked like from the editor: the turn reached the cap, the |
| 357 | /// model used its last reply to announce the next step, the call behind it was |
| 358 | /// dropped, and the thread stopped on "Now commit:" with nothing after it. |
| 359 | #[test] |
| 360 | fn a_turn_stopped_at_the_cap_says_so_to_the_user_and_the_model() { |
| 361 | let dir = scratch("cap_notice"); |
| 362 | let endpoint = start_fake_endpoint(vec![ |
| 363 | sse_tool_call("call_1", "run_command", r#"{"command":"echo round-1"}"#), |
| 364 | sse_tool_call("call_2", "run_command", r#"{"command":"echo round-2"}"#), |
| 365 | sse_text("Now commit:", "stop"), |
| 366 | ]); |
| 367 | let mut agent = spawn_agent_with_env( |
| 368 | endpoint.port, |
| 369 | &dir.join("config"), |
| 370 | &[("SIGIT_MAX_TOOL_ROUNDS", "2")], |
| 371 | ); |
| 372 | let session_id = agent.open_session(&dir.join("work")); |
| 373 | |
| 374 | let (stop_reason, reply) = agent.prompt_with_reply(&session_id, "ship it"); |
| 375 | |
| 376 | assert_eq!(stop_reason, "max_turn_requests"); |
| 377 | assert!( |
| 378 | reply.contains("Now commit:") && reply.contains("all 2 of its tool rounds"), |
| 379 | "the user must be told why the turn stopped: {reply:?}" |
| 380 | ); |
| 381 | |
| 382 | // The model is told before its last reply, on the result of the last |
| 383 | // call, and only there. |
| 384 | let requests = endpoint.requests.lock().unwrap(); |
| 385 | assert_eq!(requests.len(), 3); |
| 386 | let tool_results = |request: &Value| -> Vec<String> { |
| 387 | request["messages"] |
| 388 | .as_array() |
| 389 | .expect("messages") |
| 390 | .iter() |
| 391 | .filter(|message| message["role"] == "tool") |
| 392 | .map(|message| message["content"].as_str().unwrap_or_default().to_string()) |
| 393 | .collect() |
| 394 | }; |
| 395 | let before_cap = tool_results(&requests[1]); |
| 396 | assert!( |
| 397 | before_cap |
| 398 | .iter() |
| 399 | .all(|content| !content.contains("[siGit Code]")), |
| 400 | "no note while rounds remain: {before_cap:?}" |
| 401 | ); |
| 402 | let at_cap = tool_results(&requests[2]); |
| 403 | assert!( |
| 404 | at_cap.last().is_some_and(|content| { |
| 405 | content.contains("round-2") && content.contains("no more tools can run") |
| 406 | }), |
| 407 | "the last tool result carries the note: {at_cap:?}" |
| 408 | ); |
| 409 | assert!( |
| 410 | requests[2].get("tools").is_none(), |
| 411 | "the closing round offers no tools: {}", |
| 412 | requests[2] |
| 413 | ); |
| 414 | |
| 415 | std::fs::remove_dir_all(&dir).ok(); |
| 416 | } |
| 417 | |
| 418 | /// A turn that finishes inside its rounds gets no notice. |
| 419 | #[test] |
| 420 | fn a_turn_that_finishes_inside_the_cap_gets_no_notice() { |
| 421 | let dir = scratch("under_cap"); |
| 422 | let endpoint = start_fake_endpoint(vec![ |
| 423 | sse_tool_call("call_1", "run_command", r#"{"command":"echo round-1"}"#), |
| 424 | sse_text("Committed.", "stop"), |
| 425 | ]); |
| 426 | let mut agent = spawn_agent_with_env( |
| 427 | endpoint.port, |
| 428 | &dir.join("config"), |
| 429 | &[("SIGIT_MAX_TOOL_ROUNDS", "2")], |
| 430 | ); |
| 431 | let session_id = agent.open_session(&dir.join("work")); |
| 432 | |
| 433 | let (stop_reason, reply) = agent.prompt_with_reply(&session_id, "ship it"); |
| 434 | |
| 435 | assert_eq!(stop_reason, "end_turn"); |
| 436 | assert_eq!(reply.trim(), "Committed."); |
| 437 | |
| 438 | std::fs::remove_dir_all(&dir).ok(); |
| 439 | } |