Rebasing, going about this another way
This commit is contained in:
Ryan Ward
+131
-5
@@ -16,11 +16,17 @@ use crate::{
|
||||
network::{NetworkStatus, NetworkStatusEvent, NetworkStatusKind},
|
||||
report_error,
|
||||
server::server_api::ServerApiProvider,
|
||||
settings::ai::{AISettings, AISettingsChangedEvent, BedrockModelConfig},
|
||||
workspaces::user_workspaces::{UserWorkspaces, UserWorkspacesEvent},
|
||||
};
|
||||
|
||||
use settings::Setting;
|
||||
|
||||
use super::execution_profiles::profiles::AIExecutionProfilesModel;
|
||||
|
||||
#[cfg(not(target_family = "wasm"))]
|
||||
use super::bedrock::models::get_effective_models;
|
||||
|
||||
pub use ai::LLMId;
|
||||
|
||||
/// Checks if a user's' API key is being used for the given provider.
|
||||
@@ -36,6 +42,7 @@ pub fn is_using_api_key_for_provider(provider: &LLMProvider, app: &AppContext) -
|
||||
LLMProvider::OpenAI => api_keys.is_some_and(|keys| keys.openai.is_some()),
|
||||
LLMProvider::Anthropic => api_keys.is_some_and(|keys| keys.anthropic.is_some()),
|
||||
LLMProvider::Google => api_keys.is_some_and(|keys| keys.google.is_some()),
|
||||
LLMProvider::Bedrock => true,
|
||||
_ => false,
|
||||
}
|
||||
}
|
||||
@@ -89,6 +96,7 @@ pub enum LLMProvider {
|
||||
Anthropic,
|
||||
Google,
|
||||
Xai,
|
||||
Bedrock,
|
||||
Unknown,
|
||||
}
|
||||
|
||||
@@ -99,6 +107,7 @@ impl LLMProvider {
|
||||
LLMProvider::OpenAI => Some(Icon::OpenAILogo),
|
||||
LLMProvider::Anthropic => Some(Icon::ClaudeLogo),
|
||||
LLMProvider::Google => Some(Icon::GeminiLogo),
|
||||
LLMProvider::Bedrock => Some(Icon::BedrockLogo),
|
||||
LLMProvider::Xai => None,
|
||||
LLMProvider::Unknown => None,
|
||||
}
|
||||
@@ -498,11 +507,9 @@ struct AvailableLLMsUpdate {
|
||||
pub struct LLMPreferences {
|
||||
models_by_feature: ModelsByFeature,
|
||||
last_update: Option<AvailableLLMsUpdate>,
|
||||
// Stores temporary model overrides for a given terminal view.
|
||||
// NOTE: We only store an override if the model selected by the user is different
|
||||
// from the base LLM for the active profile. This means that if the user selects the
|
||||
// profile's default model and changes their profile, the model will update to that profile's default.
|
||||
base_llm_for_terminal_view: HashMap<EntityId, LLMId>,
|
||||
#[cfg(not(target_family = "wasm"))]
|
||||
bedrock_models_fetched: bool,
|
||||
}
|
||||
|
||||
impl LLMPreferences {
|
||||
@@ -534,12 +541,34 @@ impl LLMPreferences {
|
||||
}
|
||||
});
|
||||
|
||||
#[cfg(not(target_family = "wasm"))]
|
||||
ctx.subscribe_to_model(&AISettings::handle(ctx), |me, event, ctx| {
|
||||
if matches!(
|
||||
event,
|
||||
AISettingsChangedEvent::BedrockEnabled { .. }
|
||||
| AISettingsChangedEvent::BedrockModels { .. }
|
||||
| AISettingsChangedEvent::BedrockCrossRegionInference { .. }
|
||||
| AISettingsChangedEvent::BedrockRegion { .. }
|
||||
) {
|
||||
if matches!(event, AISettingsChangedEvent::BedrockEnabled { .. }) {
|
||||
let enabled = *AISettings::as_ref(ctx).bedrock_enabled.value();
|
||||
if enabled && !me.bedrock_models_fetched {
|
||||
me.trigger_bedrock_discovery(ctx);
|
||||
}
|
||||
}
|
||||
me.inject_bedrock_models(ctx);
|
||||
ctx.emit(LLMPreferencesEvent::UpdatedAvailableLLMs);
|
||||
}
|
||||
});
|
||||
|
||||
let base_llm_for_terminal_view = HashMap::new();
|
||||
|
||||
let me = Self {
|
||||
let mut me = Self {
|
||||
models_by_feature,
|
||||
last_update: None,
|
||||
base_llm_for_terminal_view,
|
||||
#[cfg(not(target_family = "wasm"))]
|
||||
bedrock_models_fetched: false,
|
||||
};
|
||||
|
||||
// In agent mode eval builds, eagerly kick off a fetch of the model list from the server
|
||||
@@ -549,9 +578,103 @@ impl LLMPreferences {
|
||||
#[cfg(feature = "agent_mode_evals")]
|
||||
me.refresh_available_models(ctx);
|
||||
|
||||
#[cfg(not(target_family = "wasm"))]
|
||||
{
|
||||
me.inject_bedrock_models(ctx);
|
||||
if *AISettings::as_ref(ctx).bedrock_enabled.value() {
|
||||
me.trigger_bedrock_discovery(ctx);
|
||||
}
|
||||
}
|
||||
|
||||
me
|
||||
}
|
||||
|
||||
#[cfg(not(target_family = "wasm"))]
|
||||
fn trigger_bedrock_discovery(&mut self, ctx: &mut ModelContext<Self>) {
|
||||
use crate::ai::bedrock::client::BedrockClientConfig;
|
||||
use crate::ai::bedrock::discovery::discover_inference_profiles;
|
||||
|
||||
self.bedrock_models_fetched = true;
|
||||
|
||||
let settings = AISettings::as_ref(ctx);
|
||||
let config = BedrockClientConfig {
|
||||
auth_method: settings.bedrock_auth_method.value().clone(),
|
||||
profile: settings.bedrock_profile.value().clone(),
|
||||
region: settings.bedrock_region.value().clone(),
|
||||
access_key_id: settings.bedrock_access_key_id.value().clone(),
|
||||
secret_access_key: settings.bedrock_secret_access_key.value().clone(),
|
||||
cross_region_inference: *settings.bedrock_cross_region_inference.value(),
|
||||
fallback_to_warp: *settings.bedrock_fallback_to_warp.value(),
|
||||
};
|
||||
|
||||
ctx.spawn(
|
||||
async move { discover_inference_profiles(&config).await },
|
||||
|me, result, ctx| match result {
|
||||
Ok(models) => {
|
||||
AISettings::handle(ctx).update(ctx, |settings, ctx| {
|
||||
let _ = settings.bedrock_models.set_value(models, ctx);
|
||||
});
|
||||
me.inject_bedrock_models(ctx);
|
||||
ctx.emit(LLMPreferencesEvent::UpdatedAvailableLLMs);
|
||||
}
|
||||
Err(e) => {
|
||||
log::error!("Failed to discover Bedrock inference profiles: {e}");
|
||||
}
|
||||
},
|
||||
);
|
||||
}
|
||||
|
||||
#[cfg(not(target_family = "wasm"))]
|
||||
fn inject_bedrock_models(&mut self, ctx: &AppContext) {
|
||||
self.models_by_feature
|
||||
.agent_mode
|
||||
.choices
|
||||
.retain(|m| m.provider != LLMProvider::Bedrock);
|
||||
|
||||
let settings = AISettings::as_ref(ctx);
|
||||
if !*settings.bedrock_enabled.value() {
|
||||
return;
|
||||
}
|
||||
|
||||
let user_models: Vec<BedrockModelConfig> = settings.bedrock_models.value().clone();
|
||||
let region = settings.bedrock_region.value().clone();
|
||||
let cross_region = *settings.bedrock_cross_region_inference.value();
|
||||
|
||||
let effective = get_effective_models(&user_models);
|
||||
for model in effective {
|
||||
let model_id = if cross_region && !region.is_empty() {
|
||||
super::bedrock::models::apply_cross_region_prefix(&model.model_id, ®ion)
|
||||
} else {
|
||||
model.model_id.clone()
|
||||
};
|
||||
|
||||
let llm_info = LLMInfo {
|
||||
id: LLMId::from(model_id.as_str()),
|
||||
display_name: model.display_name.clone(),
|
||||
base_model_name: model.display_name.clone(),
|
||||
reasoning_level: None,
|
||||
usage_metadata: LLMUsageMetadata {
|
||||
request_multiplier: 1,
|
||||
credit_multiplier: None,
|
||||
},
|
||||
description: Some("AWS Bedrock".to_string()),
|
||||
disable_reason: None,
|
||||
vision_supported: model.vision_supported,
|
||||
spec: None,
|
||||
provider: LLMProvider::Bedrock,
|
||||
host_configs: HashMap::from([(
|
||||
LLMModelHost::AwsBedrock,
|
||||
RoutingHostConfig {
|
||||
enabled: true,
|
||||
model_routing_host: LLMModelHost::AwsBedrock,
|
||||
},
|
||||
)]),
|
||||
discount_percentage: None,
|
||||
};
|
||||
self.models_by_feature.agent_mode.choices.push(llm_info);
|
||||
}
|
||||
}
|
||||
|
||||
/// Returns the `LLMInfo` for the base LLM to be used for an Agent Mode request.
|
||||
pub fn get_active_base_model<'a>(
|
||||
&'a self,
|
||||
@@ -924,6 +1047,9 @@ impl LLMPreferences {
|
||||
}
|
||||
}
|
||||
|
||||
#[cfg(not(target_family = "wasm"))]
|
||||
self.inject_bedrock_models(ctx);
|
||||
|
||||
// Clear any model selections where the model is no longer supported.
|
||||
let profiles_model = AIExecutionProfilesModel::handle(ctx);
|
||||
profiles_model.update(ctx, |profiles, ctx| {
|
||||
|
||||
Reference in New Issue
Block a user