1use std::sync::Arc;
2
3use anyhow::Result;
4use assistant_tool::ToolWorkingSet;
5use chrono::{DateTime, Utc};
6use collections::{BTreeMap, HashMap, HashSet};
7use futures::StreamExt as _;
8use gpui::{App, Context, Entity, EventEmitter, SharedString, Task};
9use language_model::{
10 LanguageModel, LanguageModelCompletionEvent, LanguageModelRegistry, LanguageModelRequest,
11 LanguageModelRequestMessage, LanguageModelRequestTool, LanguageModelToolResult,
12 LanguageModelToolUseId, MaxMonthlySpendReachedError, MessageContent, PaymentRequiredError,
13 Role, StopReason,
14};
15use project::Project;
16use serde::{Deserialize, Serialize};
17use util::{post_inc, TryFutureExt as _};
18use uuid::Uuid;
19
20use crate::context::{attach_context_to_message, ContextId, ContextSnapshot};
21use crate::thread_store::SavedThread;
22use crate::tool_use::{PendingToolUse, ToolUse, ToolUseState};
23
24#[derive(Debug, Clone, Copy)]
25pub enum RequestKind {
26 Chat,
27 /// Used when summarizing a thread.
28 Summarize,
29}
30
31#[derive(Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Clone, Serialize, Deserialize)]
32pub struct ThreadId(Arc<str>);
33
34impl ThreadId {
35 pub fn new() -> Self {
36 Self(Uuid::new_v4().to_string().into())
37 }
38}
39
40impl std::fmt::Display for ThreadId {
41 fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
42 write!(f, "{}", self.0)
43 }
44}
45
46#[derive(Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Clone, Copy, Serialize, Deserialize)]
47pub struct MessageId(pub(crate) usize);
48
49impl MessageId {
50 fn post_inc(&mut self) -> Self {
51 Self(post_inc(&mut self.0))
52 }
53}
54
55/// A message in a [`Thread`].
56#[derive(Debug, Clone)]
57pub struct Message {
58 pub id: MessageId,
59 pub role: Role,
60 pub text: String,
61}
62
63/// A thread of conversation with the LLM.
64pub struct Thread {
65 id: ThreadId,
66 updated_at: DateTime<Utc>,
67 summary: Option<SharedString>,
68 pending_summary: Task<Option<()>>,
69 messages: Vec<Message>,
70 next_message_id: MessageId,
71 context: BTreeMap<ContextId, ContextSnapshot>,
72 context_by_message: HashMap<MessageId, Vec<ContextId>>,
73 completion_count: usize,
74 pending_completions: Vec<PendingCompletion>,
75 project: Entity<Project>,
76 tools: Arc<ToolWorkingSet>,
77 tool_use: ToolUseState,
78}
79
80impl Thread {
81 pub fn new(
82 project: Entity<Project>,
83 tools: Arc<ToolWorkingSet>,
84 _cx: &mut Context<Self>,
85 ) -> Self {
86 Self {
87 id: ThreadId::new(),
88 updated_at: Utc::now(),
89 summary: None,
90 pending_summary: Task::ready(None),
91 messages: Vec::new(),
92 next_message_id: MessageId(0),
93 context: BTreeMap::default(),
94 context_by_message: HashMap::default(),
95 completion_count: 0,
96 pending_completions: Vec::new(),
97 project,
98 tools,
99 tool_use: ToolUseState::new(),
100 }
101 }
102
103 pub fn from_saved(
104 id: ThreadId,
105 saved: SavedThread,
106 project: Entity<Project>,
107 tools: Arc<ToolWorkingSet>,
108 _cx: &mut Context<Self>,
109 ) -> Self {
110 let next_message_id = MessageId(
111 saved
112 .messages
113 .last()
114 .map(|message| message.id.0 + 1)
115 .unwrap_or(0),
116 );
117 let tool_use = ToolUseState::from_saved_messages(&saved.messages);
118
119 Self {
120 id,
121 updated_at: saved.updated_at,
122 summary: Some(saved.summary),
123 pending_summary: Task::ready(None),
124 messages: saved
125 .messages
126 .into_iter()
127 .map(|message| Message {
128 id: message.id,
129 role: message.role,
130 text: message.text,
131 })
132 .collect(),
133 next_message_id,
134 context: BTreeMap::default(),
135 context_by_message: HashMap::default(),
136 completion_count: 0,
137 pending_completions: Vec::new(),
138 project,
139 tools,
140 tool_use,
141 }
142 }
143
144 pub fn id(&self) -> &ThreadId {
145 &self.id
146 }
147
148 pub fn is_empty(&self) -> bool {
149 self.messages.is_empty()
150 }
151
152 pub fn updated_at(&self) -> DateTime<Utc> {
153 self.updated_at
154 }
155
156 pub fn touch_updated_at(&mut self) {
157 self.updated_at = Utc::now();
158 }
159
160 pub fn summary(&self) -> Option<SharedString> {
161 self.summary.clone()
162 }
163
164 pub fn summary_or_default(&self) -> SharedString {
165 const DEFAULT: SharedString = SharedString::new_static("New Thread");
166 self.summary.clone().unwrap_or(DEFAULT)
167 }
168
169 pub fn set_summary(&mut self, summary: impl Into<SharedString>, cx: &mut Context<Self>) {
170 self.summary = Some(summary.into());
171 cx.emit(ThreadEvent::SummaryChanged);
172 }
173
174 pub fn message(&self, id: MessageId) -> Option<&Message> {
175 self.messages.iter().find(|message| message.id == id)
176 }
177
178 pub fn messages(&self) -> impl Iterator<Item = &Message> {
179 self.messages.iter()
180 }
181
182 pub fn is_streaming(&self) -> bool {
183 !self.pending_completions.is_empty()
184 }
185
186 pub fn tools(&self) -> &Arc<ToolWorkingSet> {
187 &self.tools
188 }
189
190 pub fn context_for_message(&self, id: MessageId) -> Option<Vec<ContextSnapshot>> {
191 let context = self.context_by_message.get(&id)?;
192 Some(
193 context
194 .into_iter()
195 .filter_map(|context_id| self.context.get(&context_id))
196 .cloned()
197 .collect::<Vec<_>>(),
198 )
199 }
200
201 pub fn pending_tool_uses(&self) -> Vec<&PendingToolUse> {
202 self.tool_use.pending_tool_uses()
203 }
204
205 /// Returns whether all of the tool uses have finished running.
206 pub fn all_tools_finished(&self) -> bool {
207 // If the only pending tool uses left are the ones with errors, then that means that we've finished running all
208 // of the pending tools.
209 self.pending_tool_uses()
210 .into_iter()
211 .all(|tool_use| tool_use.status.is_error())
212 }
213
214 pub fn tool_uses_for_message(&self, id: MessageId) -> Vec<ToolUse> {
215 self.tool_use.tool_uses_for_message(id)
216 }
217
218 pub fn tool_results_for_message(&self, id: MessageId) -> Vec<&LanguageModelToolResult> {
219 self.tool_use.tool_results_for_message(id)
220 }
221
222 pub fn message_has_tool_results(&self, message_id: MessageId) -> bool {
223 self.tool_use.message_has_tool_results(message_id)
224 }
225
226 pub fn insert_user_message(
227 &mut self,
228 text: impl Into<String>,
229 context: Vec<ContextSnapshot>,
230 cx: &mut Context<Self>,
231 ) {
232 let message_id = self.insert_message(Role::User, text, cx);
233 let context_ids = context.iter().map(|context| context.id).collect::<Vec<_>>();
234 self.context
235 .extend(context.into_iter().map(|context| (context.id, context)));
236 self.context_by_message.insert(message_id, context_ids);
237 }
238
239 pub fn insert_message(
240 &mut self,
241 role: Role,
242 text: impl Into<String>,
243 cx: &mut Context<Self>,
244 ) -> MessageId {
245 let id = self.next_message_id.post_inc();
246 self.messages.push(Message {
247 id,
248 role,
249 text: text.into(),
250 });
251 self.touch_updated_at();
252 cx.emit(ThreadEvent::MessageAdded(id));
253 id
254 }
255
256 pub fn edit_message(
257 &mut self,
258 id: MessageId,
259 new_role: Role,
260 new_text: String,
261 cx: &mut Context<Self>,
262 ) -> bool {
263 let Some(message) = self.messages.iter_mut().find(|message| message.id == id) else {
264 return false;
265 };
266 message.role = new_role;
267 message.text = new_text;
268 self.touch_updated_at();
269 cx.emit(ThreadEvent::MessageEdited(id));
270 true
271 }
272
273 pub fn delete_message(&mut self, id: MessageId, cx: &mut Context<Self>) -> bool {
274 let Some(index) = self.messages.iter().position(|message| message.id == id) else {
275 return false;
276 };
277 self.messages.remove(index);
278 self.context_by_message.remove(&id);
279 self.touch_updated_at();
280 cx.emit(ThreadEvent::MessageDeleted(id));
281 true
282 }
283
284 /// Returns the representation of this [`Thread`] in a textual form.
285 ///
286 /// This is the representation we use when attaching a thread as context to another thread.
287 pub fn text(&self) -> String {
288 let mut text = String::new();
289
290 for message in &self.messages {
291 text.push_str(match message.role {
292 language_model::Role::User => "User:",
293 language_model::Role::Assistant => "Assistant:",
294 language_model::Role::System => "System:",
295 });
296 text.push('\n');
297
298 text.push_str(&message.text);
299 text.push('\n');
300 }
301
302 text
303 }
304
305 pub fn send_to_model(
306 &mut self,
307 model: Arc<dyn LanguageModel>,
308 request_kind: RequestKind,
309 use_tools: bool,
310 cx: &mut Context<Self>,
311 ) {
312 let mut request = self.to_completion_request(request_kind, cx);
313
314 if use_tools {
315 request.tools = self
316 .tools()
317 .tools(cx)
318 .into_iter()
319 .map(|tool| LanguageModelRequestTool {
320 name: tool.name(),
321 description: tool.description(),
322 input_schema: tool.input_schema(),
323 })
324 .collect();
325 }
326
327 self.stream_completion(request, model, cx);
328 }
329
330 pub fn to_completion_request(
331 &self,
332 request_kind: RequestKind,
333 _cx: &App,
334 ) -> LanguageModelRequest {
335 let mut request = LanguageModelRequest {
336 messages: vec![],
337 tools: Vec::new(),
338 stop: Vec::new(),
339 temperature: None,
340 };
341
342 let mut referenced_context_ids = HashSet::default();
343
344 for message in &self.messages {
345 if let Some(context_ids) = self.context_by_message.get(&message.id) {
346 referenced_context_ids.extend(context_ids);
347 }
348
349 let mut request_message = LanguageModelRequestMessage {
350 role: message.role,
351 content: Vec::new(),
352 cache: false,
353 };
354 match request_kind {
355 RequestKind::Chat => {
356 self.tool_use
357 .attach_tool_results(message.id, &mut request_message);
358 }
359 RequestKind::Summarize => {
360 // We don't care about tool use during summarization.
361 }
362 }
363
364 if !message.text.is_empty() {
365 request_message
366 .content
367 .push(MessageContent::Text(message.text.clone()));
368 }
369
370 match request_kind {
371 RequestKind::Chat => {
372 self.tool_use
373 .attach_tool_uses(message.id, &mut request_message);
374 }
375 RequestKind::Summarize => {
376 // We don't care about tool use during summarization.
377 }
378 }
379
380 request.messages.push(request_message);
381 }
382
383 if !referenced_context_ids.is_empty() {
384 let mut context_message = LanguageModelRequestMessage {
385 role: Role::User,
386 content: Vec::new(),
387 cache: false,
388 };
389
390 let referenced_context = referenced_context_ids
391 .into_iter()
392 .filter_map(|context_id| self.context.get(context_id))
393 .cloned();
394 attach_context_to_message(&mut context_message, referenced_context);
395
396 request.messages.push(context_message);
397 }
398
399 request
400 }
401
402 pub fn stream_completion(
403 &mut self,
404 request: LanguageModelRequest,
405 model: Arc<dyn LanguageModel>,
406 cx: &mut Context<Self>,
407 ) {
408 let pending_completion_id = post_inc(&mut self.completion_count);
409
410 let task = cx.spawn(|thread, mut cx| async move {
411 let stream = model.stream_completion(request, &cx);
412 let stream_completion = async {
413 let mut events = stream.await?;
414 let mut stop_reason = StopReason::EndTurn;
415
416 while let Some(event) = events.next().await {
417 let event = event?;
418
419 thread.update(&mut cx, |thread, cx| {
420 match event {
421 LanguageModelCompletionEvent::StartMessage { .. } => {
422 thread.insert_message(Role::Assistant, String::new(), cx);
423 }
424 LanguageModelCompletionEvent::Stop(reason) => {
425 stop_reason = reason;
426 }
427 LanguageModelCompletionEvent::Text(chunk) => {
428 if let Some(last_message) = thread.messages.last_mut() {
429 if last_message.role == Role::Assistant {
430 last_message.text.push_str(&chunk);
431 cx.emit(ThreadEvent::StreamedAssistantText(
432 last_message.id,
433 chunk,
434 ));
435 } else {
436 // If we won't have an Assistant message yet, assume this chunk marks the beginning
437 // of a new Assistant response.
438 //
439 // Importantly: We do *not* want to emit a `StreamedAssistantText` event here, as it
440 // will result in duplicating the text of the chunk in the rendered Markdown.
441 thread.insert_message(Role::Assistant, chunk, cx);
442 }
443 }
444 }
445 LanguageModelCompletionEvent::ToolUse(tool_use) => {
446 if let Some(last_assistant_message) = thread
447 .messages
448 .iter()
449 .rfind(|message| message.role == Role::Assistant)
450 {
451 thread
452 .tool_use
453 .request_tool_use(last_assistant_message.id, tool_use);
454 }
455 }
456 }
457
458 thread.touch_updated_at();
459 cx.emit(ThreadEvent::StreamedCompletion);
460 cx.notify();
461 })?;
462
463 smol::future::yield_now().await;
464 }
465
466 thread.update(&mut cx, |thread, cx| {
467 thread
468 .pending_completions
469 .retain(|completion| completion.id != pending_completion_id);
470
471 if thread.summary.is_none() && thread.messages.len() >= 2 {
472 thread.summarize(cx);
473 }
474 })?;
475
476 anyhow::Ok(stop_reason)
477 };
478
479 let result = stream_completion.await;
480
481 thread
482 .update(&mut cx, |thread, cx| match result.as_ref() {
483 Ok(stop_reason) => match stop_reason {
484 StopReason::ToolUse => {
485 cx.emit(ThreadEvent::UsePendingTools);
486 }
487 StopReason::EndTurn => {}
488 StopReason::MaxTokens => {}
489 },
490 Err(error) => {
491 if error.is::<PaymentRequiredError>() {
492 cx.emit(ThreadEvent::ShowError(ThreadError::PaymentRequired));
493 } else if error.is::<MaxMonthlySpendReachedError>() {
494 cx.emit(ThreadEvent::ShowError(ThreadError::MaxMonthlySpendReached));
495 } else {
496 let error_message = error
497 .chain()
498 .map(|err| err.to_string())
499 .collect::<Vec<_>>()
500 .join("\n");
501 cx.emit(ThreadEvent::ShowError(ThreadError::Message(
502 SharedString::from(error_message.clone()),
503 )));
504 }
505
506 thread.cancel_last_completion();
507 }
508 })
509 .ok();
510 });
511
512 self.pending_completions.push(PendingCompletion {
513 id: pending_completion_id,
514 _task: task,
515 });
516 }
517
518 pub fn summarize(&mut self, cx: &mut Context<Self>) {
519 let Some(provider) = LanguageModelRegistry::read_global(cx).active_provider() else {
520 return;
521 };
522 let Some(model) = LanguageModelRegistry::read_global(cx).active_model() else {
523 return;
524 };
525
526 if !provider.is_authenticated(cx) {
527 return;
528 }
529
530 let mut request = self.to_completion_request(RequestKind::Summarize, cx);
531 request.messages.push(LanguageModelRequestMessage {
532 role: Role::User,
533 content: vec![
534 "Generate a concise 3-7 word title for this conversation, omitting punctuation. Go straight to the title, without any preamble and prefix like `Here's a concise suggestion:...` or `Title:`"
535 .into(),
536 ],
537 cache: false,
538 });
539
540 self.pending_summary = cx.spawn(|this, mut cx| {
541 async move {
542 let stream = model.stream_completion_text(request, &cx);
543 let mut messages = stream.await?;
544
545 let mut new_summary = String::new();
546 while let Some(message) = messages.stream.next().await {
547 let text = message?;
548 let mut lines = text.lines();
549 new_summary.extend(lines.next());
550
551 // Stop if the LLM generated multiple lines.
552 if lines.next().is_some() {
553 break;
554 }
555 }
556
557 this.update(&mut cx, |this, cx| {
558 if !new_summary.is_empty() {
559 this.summary = Some(new_summary.into());
560 }
561
562 cx.emit(ThreadEvent::SummaryChanged);
563 })?;
564
565 anyhow::Ok(())
566 }
567 .log_err()
568 });
569 }
570
571 pub fn use_pending_tools(&mut self, cx: &mut Context<Self>) {
572 let pending_tool_uses = self
573 .pending_tool_uses()
574 .into_iter()
575 .filter(|tool_use| tool_use.status.is_idle())
576 .cloned()
577 .collect::<Vec<_>>();
578
579 for tool_use in pending_tool_uses {
580 if let Some(tool) = self.tools.tool(&tool_use.name, cx) {
581 let task = tool.run(tool_use.input, self.project.clone(), cx);
582
583 self.insert_tool_output(tool_use.id.clone(), task, cx);
584 }
585 }
586 }
587
588 pub fn insert_tool_output(
589 &mut self,
590 tool_use_id: LanguageModelToolUseId,
591 output: Task<Result<String>>,
592 cx: &mut Context<Self>,
593 ) {
594 let insert_output_task = cx.spawn(|thread, mut cx| {
595 let tool_use_id = tool_use_id.clone();
596 async move {
597 let output = output.await;
598 thread
599 .update(&mut cx, |thread, cx| {
600 thread
601 .tool_use
602 .insert_tool_output(tool_use_id.clone(), output);
603
604 cx.emit(ThreadEvent::ToolFinished { tool_use_id });
605 })
606 .ok();
607 }
608 });
609
610 self.tool_use
611 .run_pending_tool(tool_use_id, insert_output_task);
612 }
613
614 pub fn send_tool_results_to_model(
615 &mut self,
616 model: Arc<dyn LanguageModel>,
617 cx: &mut Context<Self>,
618 ) {
619 // Insert a user message to contain the tool results.
620 self.insert_user_message(
621 // TODO: Sending up a user message without any content results in the model sending back
622 // responses that also don't have any content. We currently don't handle this case well,
623 // so for now we provide some text to keep the model on track.
624 "Here are the tool results.",
625 Vec::new(),
626 cx,
627 );
628 self.send_to_model(model, RequestKind::Chat, true, cx);
629 }
630
631 /// Cancels the last pending completion, if there are any pending.
632 ///
633 /// Returns whether a completion was canceled.
634 pub fn cancel_last_completion(&mut self) -> bool {
635 if let Some(_last_completion) = self.pending_completions.pop() {
636 true
637 } else {
638 false
639 }
640 }
641}
642
643#[derive(Debug, Clone)]
644pub enum ThreadError {
645 PaymentRequired,
646 MaxMonthlySpendReached,
647 Message(SharedString),
648}
649
650#[derive(Debug, Clone)]
651pub enum ThreadEvent {
652 ShowError(ThreadError),
653 StreamedCompletion,
654 StreamedAssistantText(MessageId, String),
655 MessageAdded(MessageId),
656 MessageEdited(MessageId),
657 MessageDeleted(MessageId),
658 SummaryChanged,
659 UsePendingTools,
660 ToolFinished {
661 #[allow(unused)]
662 tool_use_id: LanguageModelToolUseId,
663 },
664}
665
666impl EventEmitter<ThreadEvent> for Thread {}
667
668struct PendingCompletion {
669 id: usize,
670 _task: Task<()>,
671}