1use anthropic::AnthropicModelMode;
2use anyhow::{Context as _, Result, anyhow};
3use client::{Client, ModelRequestUsage, UserStore, zed_urls};
4use futures::{
5 AsyncBufReadExt, FutureExt, Stream, StreamExt, future::BoxFuture, stream::BoxStream,
6};
7use google_ai::GoogleModelMode;
8use gpui::{
9 AnyElement, AnyView, App, AsyncApp, Context, Entity, SemanticVersion, Subscription, Task,
10};
11use http_client::http::{HeaderMap, HeaderValue};
12use http_client::{AsyncBody, HttpClient, Method, Response, StatusCode};
13use language_model::{
14 AuthenticateError, LanguageModel, LanguageModelCacheConfiguration,
15 LanguageModelCompletionError, LanguageModelCompletionEvent, LanguageModelId, LanguageModelName,
16 LanguageModelProvider, LanguageModelProviderId, LanguageModelProviderName,
17 LanguageModelProviderState, LanguageModelProviderTosView, LanguageModelRequest,
18 LanguageModelToolChoice, LanguageModelToolSchemaFormat, LlmApiToken,
19 ModelRequestLimitReachedError, PaymentRequiredError, RateLimiter, RefreshLlmTokenListener,
20};
21use proto::Plan;
22use release_channel::AppVersion;
23use schemars::JsonSchema;
24use serde::{Deserialize, Serialize, de::DeserializeOwned};
25use settings::SettingsStore;
26use smol::io::{AsyncReadExt, BufReader};
27use std::pin::Pin;
28use std::str::FromStr as _;
29use std::sync::Arc;
30use std::time::Duration;
31use thiserror::Error;
32use ui::{TintColor, prelude::*};
33use util::{ResultExt as _, maybe};
34use zed_llm_client::{
35 CLIENT_SUPPORTS_STATUS_MESSAGES_HEADER_NAME, CURRENT_PLAN_HEADER_NAME, CompletionBody,
36 CompletionRequestStatus, CountTokensBody, CountTokensResponse, EXPIRED_LLM_TOKEN_HEADER_NAME,
37 ListModelsResponse, MODEL_REQUESTS_RESOURCE_HEADER_VALUE,
38 SERVER_SUPPORTS_STATUS_MESSAGES_HEADER_NAME, SUBSCRIPTION_LIMIT_RESOURCE_HEADER_NAME,
39 TOOL_USE_LIMIT_REACHED_HEADER_NAME, ZED_VERSION_HEADER_NAME,
40};
41
42use crate::provider::anthropic::{AnthropicEventMapper, count_anthropic_tokens, into_anthropic};
43use crate::provider::google::{GoogleEventMapper, into_google};
44use crate::provider::open_ai::{OpenAiEventMapper, count_open_ai_tokens, into_open_ai};
45
46const PROVIDER_ID: LanguageModelProviderId = language_model::ZED_CLOUD_PROVIDER_ID;
47const PROVIDER_NAME: LanguageModelProviderName = language_model::ZED_CLOUD_PROVIDER_NAME;
48
49#[derive(Default, Clone, Debug, PartialEq)]
50pub struct ZedDotDevSettings {
51 pub available_models: Vec<AvailableModel>,
52}
53
54#[derive(Clone, Debug, PartialEq, Serialize, Deserialize, JsonSchema)]
55#[serde(rename_all = "lowercase")]
56pub enum AvailableProvider {
57 Anthropic,
58 OpenAi,
59 Google,
60}
61
62#[derive(Clone, Debug, PartialEq, Serialize, Deserialize, JsonSchema)]
63pub struct AvailableModel {
64 /// The provider of the language model.
65 pub provider: AvailableProvider,
66 /// The model's name in the provider's API. e.g. claude-3-5-sonnet-20240620
67 pub name: String,
68 /// The name displayed in the UI, such as in the assistant panel model dropdown menu.
69 pub display_name: Option<String>,
70 /// The size of the context window, indicating the maximum number of tokens the model can process.
71 pub max_tokens: usize,
72 /// The maximum number of output tokens allowed by the model.
73 pub max_output_tokens: Option<u64>,
74 /// The maximum number of completion tokens allowed by the model (o1-* only)
75 pub max_completion_tokens: Option<u64>,
76 /// Override this model with a different Anthropic model for tool calls.
77 pub tool_override: Option<String>,
78 /// Indicates whether this custom model supports caching.
79 pub cache_configuration: Option<LanguageModelCacheConfiguration>,
80 /// The default temperature to use for this model.
81 pub default_temperature: Option<f32>,
82 /// Any extra beta headers to provide when using the model.
83 #[serde(default)]
84 pub extra_beta_headers: Vec<String>,
85 /// The model's mode (e.g. thinking)
86 pub mode: Option<ModelMode>,
87}
88
89#[derive(Default, Clone, Debug, PartialEq, Serialize, Deserialize, JsonSchema)]
90#[serde(tag = "type", rename_all = "lowercase")]
91pub enum ModelMode {
92 #[default]
93 Default,
94 Thinking {
95 /// The maximum number of tokens to use for reasoning. Must be lower than the model's `max_output_tokens`.
96 budget_tokens: Option<u32>,
97 },
98}
99
100impl From<ModelMode> for AnthropicModelMode {
101 fn from(value: ModelMode) -> Self {
102 match value {
103 ModelMode::Default => AnthropicModelMode::Default,
104 ModelMode::Thinking { budget_tokens } => AnthropicModelMode::Thinking { budget_tokens },
105 }
106 }
107}
108
109pub struct CloudLanguageModelProvider {
110 client: Arc<Client>,
111 state: gpui::Entity<State>,
112 _maintain_client_status: Task<()>,
113}
114
115pub struct State {
116 client: Arc<Client>,
117 llm_api_token: LlmApiToken,
118 user_store: Entity<UserStore>,
119 status: client::Status,
120 accept_terms: Option<Task<Result<()>>>,
121 models: Vec<Arc<zed_llm_client::LanguageModel>>,
122 default_model: Option<Arc<zed_llm_client::LanguageModel>>,
123 default_fast_model: Option<Arc<zed_llm_client::LanguageModel>>,
124 recommended_models: Vec<Arc<zed_llm_client::LanguageModel>>,
125 _fetch_models_task: Task<()>,
126 _settings_subscription: Subscription,
127 _llm_token_subscription: Subscription,
128}
129
130impl State {
131 fn new(
132 client: Arc<Client>,
133 user_store: Entity<UserStore>,
134 status: client::Status,
135 cx: &mut Context<Self>,
136 ) -> Self {
137 let refresh_llm_token_listener = RefreshLlmTokenListener::global(cx);
138
139 Self {
140 client: client.clone(),
141 llm_api_token: LlmApiToken::default(),
142 user_store,
143 status,
144 accept_terms: None,
145 models: Vec::new(),
146 default_model: None,
147 default_fast_model: None,
148 recommended_models: Vec::new(),
149 _fetch_models_task: cx.spawn(async move |this, cx| {
150 maybe!(async move {
151 let (client, llm_api_token) = this
152 .read_with(cx, |this, _cx| (client.clone(), this.llm_api_token.clone()))?;
153
154 loop {
155 let status = this.read_with(cx, |this, _cx| this.status)?;
156 if matches!(status, client::Status::Connected { .. }) {
157 break;
158 }
159
160 cx.background_executor()
161 .timer(Duration::from_millis(100))
162 .await;
163 }
164
165 let response = Self::fetch_models(client, llm_api_token).await?;
166 cx.update(|cx| {
167 this.update(cx, |this, cx| {
168 let mut models = Vec::new();
169
170 for model in response.models {
171 models.push(Arc::new(model.clone()));
172
173 // Right now we represent thinking variants of models as separate models on the client,
174 // so we need to insert variants for any model that supports thinking.
175 if model.supports_thinking {
176 models.push(Arc::new(zed_llm_client::LanguageModel {
177 id: zed_llm_client::LanguageModelId(
178 format!("{}-thinking", model.id).into(),
179 ),
180 display_name: format!("{} Thinking", model.display_name),
181 ..model
182 }));
183 }
184 }
185
186 this.default_model = models
187 .iter()
188 .find(|model| model.id == response.default_model)
189 .cloned();
190 this.default_fast_model = models
191 .iter()
192 .find(|model| model.id == response.default_fast_model)
193 .cloned();
194 this.recommended_models = response
195 .recommended_models
196 .iter()
197 .filter_map(|id| models.iter().find(|model| &model.id == id))
198 .cloned()
199 .collect();
200 this.models = models;
201 cx.notify();
202 })
203 })??;
204
205 anyhow::Ok(())
206 })
207 .await
208 .context("failed to fetch Zed models")
209 .log_err();
210 }),
211 _settings_subscription: cx.observe_global::<SettingsStore>(|_, cx| {
212 cx.notify();
213 }),
214 _llm_token_subscription: cx.subscribe(
215 &refresh_llm_token_listener,
216 |this, _listener, _event, cx| {
217 let client = this.client.clone();
218 let llm_api_token = this.llm_api_token.clone();
219 cx.spawn(async move |_this, _cx| {
220 llm_api_token.refresh(&client).await?;
221 anyhow::Ok(())
222 })
223 .detach_and_log_err(cx);
224 },
225 ),
226 }
227 }
228
229 fn is_signed_out(&self) -> bool {
230 self.status.is_signed_out()
231 }
232
233 fn authenticate(&self, cx: &mut Context<Self>) -> Task<Result<()>> {
234 let client = self.client.clone();
235 cx.spawn(async move |state, cx| {
236 client
237 .authenticate_and_connect(true, &cx)
238 .await
239 .into_response()?;
240 state.update(cx, |_, cx| cx.notify())
241 })
242 }
243
244 fn has_accepted_terms_of_service(&self, cx: &App) -> bool {
245 self.user_store
246 .read(cx)
247 .current_user_has_accepted_terms()
248 .unwrap_or(false)
249 }
250
251 fn accept_terms_of_service(&mut self, cx: &mut Context<Self>) {
252 let user_store = self.user_store.clone();
253 self.accept_terms = Some(cx.spawn(async move |this, cx| {
254 let _ = user_store
255 .update(cx, |store, cx| store.accept_terms_of_service(cx))?
256 .await;
257 this.update(cx, |this, cx| {
258 this.accept_terms = None;
259 cx.notify()
260 })
261 }));
262 }
263
264 async fn fetch_models(
265 client: Arc<Client>,
266 llm_api_token: LlmApiToken,
267 ) -> Result<ListModelsResponse> {
268 let http_client = &client.http_client();
269 let token = llm_api_token.acquire(&client).await?;
270
271 let request = http_client::Request::builder()
272 .method(Method::GET)
273 .uri(http_client.build_zed_llm_url("/models", &[])?.as_ref())
274 .header("Authorization", format!("Bearer {token}"))
275 .body(AsyncBody::empty())?;
276 let mut response = http_client
277 .send(request)
278 .await
279 .context("failed to send list models request")?;
280
281 if response.status().is_success() {
282 let mut body = String::new();
283 response.body_mut().read_to_string(&mut body).await?;
284 return Ok(serde_json::from_str(&body)?);
285 } else {
286 let mut body = String::new();
287 response.body_mut().read_to_string(&mut body).await?;
288 anyhow::bail!(
289 "error listing models.\nStatus: {:?}\nBody: {body}",
290 response.status(),
291 );
292 }
293 }
294}
295
296impl CloudLanguageModelProvider {
297 pub fn new(user_store: Entity<UserStore>, client: Arc<Client>, cx: &mut App) -> Self {
298 let mut status_rx = client.status();
299 let status = *status_rx.borrow();
300
301 let state = cx.new(|cx| State::new(client.clone(), user_store.clone(), status, cx));
302
303 let state_ref = state.downgrade();
304 let maintain_client_status = cx.spawn(async move |cx| {
305 while let Some(status) = status_rx.next().await {
306 if let Some(this) = state_ref.upgrade() {
307 _ = this.update(cx, |this, cx| {
308 if this.status != status {
309 this.status = status;
310 cx.notify();
311 }
312 });
313 } else {
314 break;
315 }
316 }
317 });
318
319 Self {
320 client,
321 state: state.clone(),
322 _maintain_client_status: maintain_client_status,
323 }
324 }
325
326 fn create_language_model(
327 &self,
328 model: Arc<zed_llm_client::LanguageModel>,
329 llm_api_token: LlmApiToken,
330 ) -> Arc<dyn LanguageModel> {
331 Arc::new(CloudLanguageModel {
332 id: LanguageModelId(SharedString::from(model.id.0.clone())),
333 model,
334 llm_api_token: llm_api_token.clone(),
335 client: self.client.clone(),
336 request_limiter: RateLimiter::new(4),
337 })
338 }
339}
340
341impl LanguageModelProviderState for CloudLanguageModelProvider {
342 type ObservableEntity = State;
343
344 fn observable_entity(&self) -> Option<gpui::Entity<Self::ObservableEntity>> {
345 Some(self.state.clone())
346 }
347}
348
349impl LanguageModelProvider for CloudLanguageModelProvider {
350 fn id(&self) -> LanguageModelProviderId {
351 PROVIDER_ID
352 }
353
354 fn name(&self) -> LanguageModelProviderName {
355 PROVIDER_NAME
356 }
357
358 fn icon(&self) -> IconName {
359 IconName::AiZed
360 }
361
362 fn default_model(&self, cx: &App) -> Option<Arc<dyn LanguageModel>> {
363 let default_model = self.state.read(cx).default_model.clone()?;
364 let llm_api_token = self.state.read(cx).llm_api_token.clone();
365 Some(self.create_language_model(default_model, llm_api_token))
366 }
367
368 fn default_fast_model(&self, cx: &App) -> Option<Arc<dyn LanguageModel>> {
369 let default_fast_model = self.state.read(cx).default_fast_model.clone()?;
370 let llm_api_token = self.state.read(cx).llm_api_token.clone();
371 Some(self.create_language_model(default_fast_model, llm_api_token))
372 }
373
374 fn recommended_models(&self, cx: &App) -> Vec<Arc<dyn LanguageModel>> {
375 let llm_api_token = self.state.read(cx).llm_api_token.clone();
376 self.state
377 .read(cx)
378 .recommended_models
379 .iter()
380 .cloned()
381 .map(|model| self.create_language_model(model, llm_api_token.clone()))
382 .collect()
383 }
384
385 fn provided_models(&self, cx: &App) -> Vec<Arc<dyn LanguageModel>> {
386 let llm_api_token = self.state.read(cx).llm_api_token.clone();
387 self.state
388 .read(cx)
389 .models
390 .iter()
391 .cloned()
392 .map(|model| self.create_language_model(model, llm_api_token.clone()))
393 .collect()
394 }
395
396 fn is_authenticated(&self, cx: &App) -> bool {
397 let state = self.state.read(cx);
398 !state.is_signed_out() && state.has_accepted_terms_of_service(cx)
399 }
400
401 fn authenticate(&self, _cx: &mut App) -> Task<Result<(), AuthenticateError>> {
402 Task::ready(Ok(()))
403 }
404
405 fn configuration_view(&self, _: &mut Window, cx: &mut App) -> AnyView {
406 cx.new(|_| ConfigurationView {
407 state: self.state.clone(),
408 })
409 .into()
410 }
411
412 fn must_accept_terms(&self, cx: &App) -> bool {
413 !self.state.read(cx).has_accepted_terms_of_service(cx)
414 }
415
416 fn render_accept_terms(
417 &self,
418 view: LanguageModelProviderTosView,
419 cx: &mut App,
420 ) -> Option<AnyElement> {
421 render_accept_terms(self.state.clone(), view, cx)
422 }
423
424 fn reset_credentials(&self, _cx: &mut App) -> Task<Result<()>> {
425 Task::ready(Ok(()))
426 }
427}
428
429fn render_accept_terms(
430 state: Entity<State>,
431 view_kind: LanguageModelProviderTosView,
432 cx: &mut App,
433) -> Option<AnyElement> {
434 if state.read(cx).has_accepted_terms_of_service(cx) {
435 return None;
436 }
437
438 let accept_terms_disabled = state.read(cx).accept_terms.is_some();
439
440 let thread_fresh_start = matches!(view_kind, LanguageModelProviderTosView::ThreadFreshStart);
441 let thread_empty_state = matches!(view_kind, LanguageModelProviderTosView::ThreadtEmptyState);
442
443 let terms_button = Button::new("terms_of_service", "Terms of Service")
444 .style(ButtonStyle::Subtle)
445 .icon(IconName::ArrowUpRight)
446 .icon_color(Color::Muted)
447 .icon_size(IconSize::XSmall)
448 .when(thread_empty_state, |this| this.label_size(LabelSize::Small))
449 .on_click(move |_, _window, cx| cx.open_url("https://zed.dev/terms-of-service"));
450
451 let button_container = h_flex().child(
452 Button::new("accept_terms", "I accept the Terms of Service")
453 .when(!thread_empty_state, |this| {
454 this.full_width()
455 .style(ButtonStyle::Tinted(TintColor::Accent))
456 .icon(IconName::Check)
457 .icon_position(IconPosition::Start)
458 .icon_size(IconSize::Small)
459 })
460 .when(thread_empty_state, |this| {
461 this.style(ButtonStyle::Tinted(TintColor::Warning))
462 .label_size(LabelSize::Small)
463 })
464 .disabled(accept_terms_disabled)
465 .on_click({
466 let state = state.downgrade();
467 move |_, _window, cx| {
468 state
469 .update(cx, |state, cx| state.accept_terms_of_service(cx))
470 .ok();
471 }
472 }),
473 );
474
475 let form = if thread_empty_state {
476 h_flex()
477 .w_full()
478 .flex_wrap()
479 .justify_between()
480 .child(
481 h_flex()
482 .child(
483 Label::new("To start using Zed AI, please read and accept the")
484 .size(LabelSize::Small),
485 )
486 .child(terms_button),
487 )
488 .child(button_container)
489 } else {
490 v_flex()
491 .w_full()
492 .gap_2()
493 .child(
494 h_flex()
495 .flex_wrap()
496 .when(thread_fresh_start, |this| this.justify_center())
497 .child(Label::new(
498 "To start using Zed AI, please read and accept the",
499 ))
500 .child(terms_button),
501 )
502 .child({
503 match view_kind {
504 LanguageModelProviderTosView::PromptEditorPopup => {
505 button_container.w_full().justify_end()
506 }
507 LanguageModelProviderTosView::Configuration => {
508 button_container.w_full().justify_start()
509 }
510 LanguageModelProviderTosView::ThreadFreshStart => {
511 button_container.w_full().justify_center()
512 }
513 LanguageModelProviderTosView::ThreadtEmptyState => div().w_0(),
514 }
515 })
516 };
517
518 Some(form.into_any())
519}
520
521pub struct CloudLanguageModel {
522 id: LanguageModelId,
523 model: Arc<zed_llm_client::LanguageModel>,
524 llm_api_token: LlmApiToken,
525 client: Arc<Client>,
526 request_limiter: RateLimiter,
527}
528
529struct PerformLlmCompletionResponse {
530 response: Response<AsyncBody>,
531 usage: Option<ModelRequestUsage>,
532 tool_use_limit_reached: bool,
533 includes_status_messages: bool,
534}
535
536impl CloudLanguageModel {
537 async fn perform_llm_completion(
538 client: Arc<Client>,
539 llm_api_token: LlmApiToken,
540 app_version: Option<SemanticVersion>,
541 body: CompletionBody,
542 ) -> Result<PerformLlmCompletionResponse> {
543 let http_client = &client.http_client();
544
545 let mut token = llm_api_token.acquire(&client).await?;
546 let mut refreshed_token = false;
547
548 loop {
549 let request_builder = http_client::Request::builder()
550 .method(Method::POST)
551 .uri(http_client.build_zed_llm_url("/completions", &[])?.as_ref());
552 let request_builder = if let Some(app_version) = app_version {
553 request_builder.header(ZED_VERSION_HEADER_NAME, app_version.to_string())
554 } else {
555 request_builder
556 };
557
558 let request = request_builder
559 .header("Content-Type", "application/json")
560 .header("Authorization", format!("Bearer {token}"))
561 .header(CLIENT_SUPPORTS_STATUS_MESSAGES_HEADER_NAME, "true")
562 .body(serde_json::to_string(&body)?.into())?;
563 let mut response = http_client.send(request).await?;
564 let status = response.status();
565 if status.is_success() {
566 let includes_status_messages = response
567 .headers()
568 .get(SERVER_SUPPORTS_STATUS_MESSAGES_HEADER_NAME)
569 .is_some();
570
571 let tool_use_limit_reached = response
572 .headers()
573 .get(TOOL_USE_LIMIT_REACHED_HEADER_NAME)
574 .is_some();
575
576 let usage = if includes_status_messages {
577 None
578 } else {
579 ModelRequestUsage::from_headers(response.headers()).ok()
580 };
581
582 return Ok(PerformLlmCompletionResponse {
583 response,
584 usage,
585 includes_status_messages,
586 tool_use_limit_reached,
587 });
588 }
589
590 if !refreshed_token
591 && response
592 .headers()
593 .get(EXPIRED_LLM_TOKEN_HEADER_NAME)
594 .is_some()
595 {
596 token = llm_api_token.refresh(&client).await?;
597 refreshed_token = true;
598 continue;
599 }
600
601 if status == StatusCode::FORBIDDEN
602 && response
603 .headers()
604 .get(SUBSCRIPTION_LIMIT_RESOURCE_HEADER_NAME)
605 .is_some()
606 {
607 if let Some(MODEL_REQUESTS_RESOURCE_HEADER_VALUE) = response
608 .headers()
609 .get(SUBSCRIPTION_LIMIT_RESOURCE_HEADER_NAME)
610 .and_then(|resource| resource.to_str().ok())
611 {
612 if let Some(plan) = response
613 .headers()
614 .get(CURRENT_PLAN_HEADER_NAME)
615 .and_then(|plan| plan.to_str().ok())
616 .and_then(|plan| zed_llm_client::Plan::from_str(plan).ok())
617 {
618 let plan = match plan {
619 zed_llm_client::Plan::ZedFree => Plan::Free,
620 zed_llm_client::Plan::ZedPro => Plan::ZedPro,
621 zed_llm_client::Plan::ZedProTrial => Plan::ZedProTrial,
622 };
623 return Err(anyhow!(ModelRequestLimitReachedError { plan }));
624 }
625 }
626 } else if status == StatusCode::PAYMENT_REQUIRED {
627 return Err(anyhow!(PaymentRequiredError));
628 }
629
630 let mut body = String::new();
631 let headers = response.headers().clone();
632 response.body_mut().read_to_string(&mut body).await?;
633 return Err(anyhow!(ApiError {
634 status,
635 body,
636 headers
637 }));
638 }
639 }
640}
641
642#[derive(Debug, Error)]
643#[error("cloud language model request failed with status {status}: {body}")]
644struct ApiError {
645 status: StatusCode,
646 body: String,
647 headers: HeaderMap<HeaderValue>,
648}
649
650impl From<ApiError> for LanguageModelCompletionError {
651 fn from(error: ApiError) -> Self {
652 let retry_after = None;
653 LanguageModelCompletionError::from_http_status(
654 PROVIDER_NAME,
655 error.status,
656 error.body,
657 retry_after,
658 )
659 }
660}
661
662impl LanguageModel for CloudLanguageModel {
663 fn id(&self) -> LanguageModelId {
664 self.id.clone()
665 }
666
667 fn name(&self) -> LanguageModelName {
668 LanguageModelName::from(self.model.display_name.clone())
669 }
670
671 fn provider_id(&self) -> LanguageModelProviderId {
672 PROVIDER_ID
673 }
674
675 fn provider_name(&self) -> LanguageModelProviderName {
676 PROVIDER_NAME
677 }
678
679 fn upstream_provider_id(&self) -> LanguageModelProviderId {
680 use zed_llm_client::LanguageModelProvider::*;
681 match self.model.provider {
682 Anthropic => language_model::ANTHROPIC_PROVIDER_ID,
683 OpenAi => language_model::OPEN_AI_PROVIDER_ID,
684 Google => language_model::GOOGLE_PROVIDER_ID,
685 }
686 }
687
688 fn upstream_provider_name(&self) -> LanguageModelProviderName {
689 use zed_llm_client::LanguageModelProvider::*;
690 match self.model.provider {
691 Anthropic => language_model::ANTHROPIC_PROVIDER_NAME,
692 OpenAi => language_model::OPEN_AI_PROVIDER_NAME,
693 Google => language_model::GOOGLE_PROVIDER_NAME,
694 }
695 }
696
697 fn supports_tools(&self) -> bool {
698 self.model.supports_tools
699 }
700
701 fn supports_images(&self) -> bool {
702 self.model.supports_images
703 }
704
705 fn supports_tool_choice(&self, choice: LanguageModelToolChoice) -> bool {
706 match choice {
707 LanguageModelToolChoice::Auto
708 | LanguageModelToolChoice::Any
709 | LanguageModelToolChoice::None => true,
710 }
711 }
712
713 fn supports_burn_mode(&self) -> bool {
714 self.model.supports_max_mode
715 }
716
717 fn telemetry_id(&self) -> String {
718 format!("zed.dev/{}", self.model.id)
719 }
720
721 fn tool_input_format(&self) -> LanguageModelToolSchemaFormat {
722 match self.model.provider {
723 zed_llm_client::LanguageModelProvider::Anthropic
724 | zed_llm_client::LanguageModelProvider::OpenAi => {
725 LanguageModelToolSchemaFormat::JsonSchema
726 }
727 zed_llm_client::LanguageModelProvider::Google => {
728 LanguageModelToolSchemaFormat::JsonSchemaSubset
729 }
730 }
731 }
732
733 fn max_token_count(&self) -> u64 {
734 self.model.max_token_count as u64
735 }
736
737 fn cache_configuration(&self) -> Option<LanguageModelCacheConfiguration> {
738 match &self.model.provider {
739 zed_llm_client::LanguageModelProvider::Anthropic => {
740 Some(LanguageModelCacheConfiguration {
741 min_total_token: 2_048,
742 should_speculate: true,
743 max_cache_anchors: 4,
744 })
745 }
746 zed_llm_client::LanguageModelProvider::OpenAi
747 | zed_llm_client::LanguageModelProvider::Google => None,
748 }
749 }
750
751 fn count_tokens(
752 &self,
753 request: LanguageModelRequest,
754 cx: &App,
755 ) -> BoxFuture<'static, Result<u64>> {
756 match self.model.provider {
757 zed_llm_client::LanguageModelProvider::Anthropic => count_anthropic_tokens(request, cx),
758 zed_llm_client::LanguageModelProvider::OpenAi => {
759 let model = match open_ai::Model::from_id(&self.model.id.0) {
760 Ok(model) => model,
761 Err(err) => return async move { Err(anyhow!(err)) }.boxed(),
762 };
763 count_open_ai_tokens(request, model, cx)
764 }
765 zed_llm_client::LanguageModelProvider::Google => {
766 let client = self.client.clone();
767 let llm_api_token = self.llm_api_token.clone();
768 let model_id = self.model.id.to_string();
769 let generate_content_request =
770 into_google(request, model_id.clone(), GoogleModelMode::Default);
771 async move {
772 let http_client = &client.http_client();
773 let token = llm_api_token.acquire(&client).await?;
774
775 let request_body = CountTokensBody {
776 provider: zed_llm_client::LanguageModelProvider::Google,
777 model: model_id,
778 provider_request: serde_json::to_value(&google_ai::CountTokensRequest {
779 generate_content_request,
780 })?,
781 };
782 let request = http_client::Request::builder()
783 .method(Method::POST)
784 .uri(
785 http_client
786 .build_zed_llm_url("/count_tokens", &[])?
787 .as_ref(),
788 )
789 .header("Content-Type", "application/json")
790 .header("Authorization", format!("Bearer {token}"))
791 .body(serde_json::to_string(&request_body)?.into())?;
792 let mut response = http_client.send(request).await?;
793 let status = response.status();
794 let headers = response.headers().clone();
795 let mut response_body = String::new();
796 response
797 .body_mut()
798 .read_to_string(&mut response_body)
799 .await?;
800
801 if status.is_success() {
802 let response_body: CountTokensResponse =
803 serde_json::from_str(&response_body)?;
804
805 Ok(response_body.tokens as u64)
806 } else {
807 Err(anyhow!(ApiError {
808 status,
809 body: response_body,
810 headers
811 }))
812 }
813 }
814 .boxed()
815 }
816 }
817 }
818
819 fn stream_completion(
820 &self,
821 request: LanguageModelRequest,
822 cx: &AsyncApp,
823 ) -> BoxFuture<
824 'static,
825 Result<
826 BoxStream<'static, Result<LanguageModelCompletionEvent, LanguageModelCompletionError>>,
827 LanguageModelCompletionError,
828 >,
829 > {
830 let thread_id = request.thread_id.clone();
831 let prompt_id = request.prompt_id.clone();
832 let intent = request.intent;
833 let mode = request.mode;
834 let app_version = cx.update(|cx| AppVersion::global(cx)).ok();
835 match self.model.provider {
836 zed_llm_client::LanguageModelProvider::Anthropic => {
837 let request = into_anthropic(
838 request,
839 self.model.id.to_string(),
840 1.0,
841 self.model.max_output_tokens as u64,
842 if self.model.id.0.ends_with("-thinking") {
843 AnthropicModelMode::Thinking {
844 budget_tokens: Some(4_096),
845 }
846 } else {
847 AnthropicModelMode::Default
848 },
849 );
850 let client = self.client.clone();
851 let llm_api_token = self.llm_api_token.clone();
852 let future = self.request_limiter.stream(async move {
853 let PerformLlmCompletionResponse {
854 response,
855 usage,
856 includes_status_messages,
857 tool_use_limit_reached,
858 } = Self::perform_llm_completion(
859 client.clone(),
860 llm_api_token,
861 app_version,
862 CompletionBody {
863 thread_id,
864 prompt_id,
865 intent,
866 mode,
867 provider: zed_llm_client::LanguageModelProvider::Anthropic,
868 model: request.model.clone(),
869 provider_request: serde_json::to_value(&request)
870 .map_err(|e| anyhow!(e))?,
871 },
872 )
873 .await
874 .map_err(|err| match err.downcast::<ApiError>() {
875 Ok(api_err) => anyhow!(LanguageModelCompletionError::from(api_err)),
876 Err(err) => anyhow!(err),
877 })?;
878
879 let mut mapper = AnthropicEventMapper::new();
880 Ok(map_cloud_completion_events(
881 Box::pin(
882 response_lines(response, includes_status_messages)
883 .chain(usage_updated_event(usage))
884 .chain(tool_use_limit_reached_event(tool_use_limit_reached)),
885 ),
886 move |event| mapper.map_event(event),
887 ))
888 });
889 async move { Ok(future.await?.boxed()) }.boxed()
890 }
891 zed_llm_client::LanguageModelProvider::OpenAi => {
892 let client = self.client.clone();
893 let model = match open_ai::Model::from_id(&self.model.id.0) {
894 Ok(model) => model,
895 Err(err) => return async move { Err(anyhow!(err).into()) }.boxed(),
896 };
897 let request = into_open_ai(
898 request,
899 model.id(),
900 model.supports_parallel_tool_calls(),
901 None,
902 );
903 let llm_api_token = self.llm_api_token.clone();
904 let future = self.request_limiter.stream(async move {
905 let PerformLlmCompletionResponse {
906 response,
907 usage,
908 includes_status_messages,
909 tool_use_limit_reached,
910 } = Self::perform_llm_completion(
911 client.clone(),
912 llm_api_token,
913 app_version,
914 CompletionBody {
915 thread_id,
916 prompt_id,
917 intent,
918 mode,
919 provider: zed_llm_client::LanguageModelProvider::OpenAi,
920 model: request.model.clone(),
921 provider_request: serde_json::to_value(&request)
922 .map_err(|e| anyhow!(e))?,
923 },
924 )
925 .await?;
926
927 let mut mapper = OpenAiEventMapper::new();
928 Ok(map_cloud_completion_events(
929 Box::pin(
930 response_lines(response, includes_status_messages)
931 .chain(usage_updated_event(usage))
932 .chain(tool_use_limit_reached_event(tool_use_limit_reached)),
933 ),
934 move |event| mapper.map_event(event),
935 ))
936 });
937 async move { Ok(future.await?.boxed()) }.boxed()
938 }
939 zed_llm_client::LanguageModelProvider::Google => {
940 let client = self.client.clone();
941 let request =
942 into_google(request, self.model.id.to_string(), GoogleModelMode::Default);
943 let llm_api_token = self.llm_api_token.clone();
944 let future = self.request_limiter.stream(async move {
945 let PerformLlmCompletionResponse {
946 response,
947 usage,
948 includes_status_messages,
949 tool_use_limit_reached,
950 } = Self::perform_llm_completion(
951 client.clone(),
952 llm_api_token,
953 app_version,
954 CompletionBody {
955 thread_id,
956 prompt_id,
957 intent,
958 mode,
959 provider: zed_llm_client::LanguageModelProvider::Google,
960 model: request.model.model_id.clone(),
961 provider_request: serde_json::to_value(&request)
962 .map_err(|e| anyhow!(e))?,
963 },
964 )
965 .await?;
966
967 let mut mapper = GoogleEventMapper::new();
968 Ok(map_cloud_completion_events(
969 Box::pin(
970 response_lines(response, includes_status_messages)
971 .chain(usage_updated_event(usage))
972 .chain(tool_use_limit_reached_event(tool_use_limit_reached)),
973 ),
974 move |event| mapper.map_event(event),
975 ))
976 });
977 async move { Ok(future.await?.boxed()) }.boxed()
978 }
979 }
980 }
981}
982
983#[derive(Serialize, Deserialize)]
984#[serde(rename_all = "snake_case")]
985pub enum CloudCompletionEvent<T> {
986 Status(CompletionRequestStatus),
987 Event(T),
988}
989
990fn map_cloud_completion_events<T, F>(
991 stream: Pin<Box<dyn Stream<Item = Result<CloudCompletionEvent<T>>> + Send>>,
992 mut map_callback: F,
993) -> BoxStream<'static, Result<LanguageModelCompletionEvent, LanguageModelCompletionError>>
994where
995 T: DeserializeOwned + 'static,
996 F: FnMut(T) -> Vec<Result<LanguageModelCompletionEvent, LanguageModelCompletionError>>
997 + Send
998 + 'static,
999{
1000 stream
1001 .flat_map(move |event| {
1002 futures::stream::iter(match event {
1003 Err(error) => {
1004 vec![Err(LanguageModelCompletionError::from(error))]
1005 }
1006 Ok(CloudCompletionEvent::Status(event)) => {
1007 vec![Ok(LanguageModelCompletionEvent::StatusUpdate(event))]
1008 }
1009 Ok(CloudCompletionEvent::Event(event)) => map_callback(event),
1010 })
1011 })
1012 .boxed()
1013}
1014
1015fn usage_updated_event<T>(
1016 usage: Option<ModelRequestUsage>,
1017) -> impl Stream<Item = Result<CloudCompletionEvent<T>>> {
1018 futures::stream::iter(usage.map(|usage| {
1019 Ok(CloudCompletionEvent::Status(
1020 CompletionRequestStatus::UsageUpdated {
1021 amount: usage.amount as usize,
1022 limit: usage.limit,
1023 },
1024 ))
1025 }))
1026}
1027
1028fn tool_use_limit_reached_event<T>(
1029 tool_use_limit_reached: bool,
1030) -> impl Stream<Item = Result<CloudCompletionEvent<T>>> {
1031 futures::stream::iter(tool_use_limit_reached.then(|| {
1032 Ok(CloudCompletionEvent::Status(
1033 CompletionRequestStatus::ToolUseLimitReached,
1034 ))
1035 }))
1036}
1037
1038fn response_lines<T: DeserializeOwned>(
1039 response: Response<AsyncBody>,
1040 includes_status_messages: bool,
1041) -> impl Stream<Item = Result<CloudCompletionEvent<T>>> {
1042 futures::stream::try_unfold(
1043 (String::new(), BufReader::new(response.into_body())),
1044 move |(mut line, mut body)| async move {
1045 match body.read_line(&mut line).await {
1046 Ok(0) => Ok(None),
1047 Ok(_) => {
1048 let event = if includes_status_messages {
1049 serde_json::from_str::<CloudCompletionEvent<T>>(&line)?
1050 } else {
1051 CloudCompletionEvent::Event(serde_json::from_str::<T>(&line)?)
1052 };
1053
1054 line.clear();
1055 Ok(Some((event, (line, body))))
1056 }
1057 Err(e) => Err(e.into()),
1058 }
1059 },
1060 )
1061}
1062
1063struct ConfigurationView {
1064 state: gpui::Entity<State>,
1065}
1066
1067impl ConfigurationView {
1068 fn authenticate(&mut self, cx: &mut Context<Self>) {
1069 self.state.update(cx, |state, cx| {
1070 state.authenticate(cx).detach_and_log_err(cx);
1071 });
1072 cx.notify();
1073 }
1074}
1075
1076impl Render for ConfigurationView {
1077 fn render(&mut self, _: &mut Window, cx: &mut Context<Self>) -> impl IntoElement {
1078 const ZED_PRICING_URL: &str = "https://zed.dev/pricing";
1079
1080 let is_connected = !self.state.read(cx).is_signed_out();
1081 let user_store = self.state.read(cx).user_store.read(cx);
1082 let plan = user_store.current_plan();
1083 let subscription_period = user_store.subscription_period();
1084 let eligible_for_trial = user_store.trial_started_at().is_none();
1085 let has_accepted_terms = self.state.read(cx).has_accepted_terms_of_service(cx);
1086
1087 let is_pro = plan == Some(proto::Plan::ZedPro);
1088 let subscription_text = match (plan, subscription_period) {
1089 (Some(proto::Plan::ZedPro), Some(_)) => {
1090 "You have access to Zed's hosted LLMs through your Zed Pro subscription."
1091 }
1092 (Some(proto::Plan::ZedProTrial), Some(_)) => {
1093 "You have access to Zed's hosted LLMs through your Zed Pro trial."
1094 }
1095 (Some(proto::Plan::Free), Some(_)) => {
1096 "You have basic access to Zed's hosted LLMs through your Zed Free subscription."
1097 }
1098 _ => {
1099 if eligible_for_trial {
1100 "Subscribe for access to Zed's hosted LLMs. Start with a 14 day free trial."
1101 } else {
1102 "Subscribe for access to Zed's hosted LLMs."
1103 }
1104 }
1105 };
1106 let manage_subscription_buttons = if is_pro {
1107 h_flex().child(
1108 Button::new("manage_settings", "Manage Subscription")
1109 .style(ButtonStyle::Tinted(TintColor::Accent))
1110 .on_click(cx.listener(|_, _, _, cx| cx.open_url(&zed_urls::account_url(cx)))),
1111 )
1112 } else {
1113 h_flex()
1114 .gap_2()
1115 .child(
1116 Button::new("learn_more", "Learn more")
1117 .style(ButtonStyle::Subtle)
1118 .on_click(cx.listener(|_, _, _, cx| cx.open_url(ZED_PRICING_URL))),
1119 )
1120 .child(
1121 Button::new("upgrade", "Upgrade")
1122 .style(ButtonStyle::Subtle)
1123 .color(Color::Accent)
1124 .on_click(
1125 cx.listener(|_, _, _, cx| cx.open_url(&zed_urls::account_url(cx))),
1126 ),
1127 )
1128 };
1129
1130 if is_connected {
1131 v_flex()
1132 .gap_3()
1133 .w_full()
1134 .children(render_accept_terms(
1135 self.state.clone(),
1136 LanguageModelProviderTosView::Configuration,
1137 cx,
1138 ))
1139 .when(has_accepted_terms, |this| {
1140 this.child(subscription_text)
1141 .child(manage_subscription_buttons)
1142 })
1143 } else {
1144 v_flex()
1145 .gap_2()
1146 .child(Label::new("Use Zed AI to access hosted language models."))
1147 .child(
1148 Button::new("sign_in", "Sign In")
1149 .icon_color(Color::Muted)
1150 .icon(IconName::Github)
1151 .icon_position(IconPosition::Start)
1152 .on_click(cx.listener(move |this, _, _, cx| this.authenticate(cx))),
1153 )
1154 }
1155 }
1156}