Thread Anthropic errors into LanguageModelKnownError (#33261)

This PR is in preparation for doing automatic retries for certain
errors, e.g. Overloaded. It doesn't change behavior yet (aside from some
granularity of error messages shown to the user), but rather mostly
changes some error handling to be exhaustive enum matches instead of
`anyhow` downcasts, and leaves some comments for where the behavior
change will be in a future PR.

Release Notes:

- N/A

crates/language_models/src/provider/anthropic.rs

@@ -16,10 +16,10 @@ use gpui::{
 use http_client::HttpClient;
 use language_model::{
     AuthenticateError, LanguageModel, LanguageModelCacheConfiguration,
-    LanguageModelCompletionError, LanguageModelId, LanguageModelKnownError, LanguageModelName,
-    LanguageModelProvider, LanguageModelProviderId, LanguageModelProviderName,
-    LanguageModelProviderState, LanguageModelRequest, LanguageModelToolChoice,
-    LanguageModelToolResultContent, MessageContent, RateLimiter, Role,
+    LanguageModelCompletionError, LanguageModelId, LanguageModelName, LanguageModelProvider,
+    LanguageModelProviderId, LanguageModelProviderName, LanguageModelProviderState,
+    LanguageModelRequest, LanguageModelToolChoice, LanguageModelToolResultContent, MessageContent,
+    RateLimiter, Role,
 };
 use language_model::{LanguageModelCompletionEvent, LanguageModelToolUse, StopReason};
 use schemars::JsonSchema;
@@ -407,14 +407,7 @@ impl AnthropicModel {
             let api_key = api_key.context("Missing Anthropic API Key")?;
             let request =
                 anthropic::stream_completion(http_client.as_ref(), &api_url, &api_key, request);
-            request.await.map_err(|err| match err {
-                AnthropicError::RateLimit(duration) => {
-                    LanguageModelCompletionError::RateLimit(duration)
-                }
-                err @ (AnthropicError::ApiError(..) | AnthropicError::Other(..)) => {
-                    LanguageModelCompletionError::Other(anthropic_err_to_anyhow(err))
-                }
-            })
+            request.await.map_err(Into::into)
         }
         .boxed()
     }
@@ -714,7 +707,7 @@ impl AnthropicEventMapper {
         events.flat_map(move |event| {
             futures::stream::iter(match event {
                 Ok(event) => self.map_event(event),
-                Err(error) => vec![Err(LanguageModelCompletionError::Other(anyhow!(error)))],
+                Err(error) => vec![Err(error.into())],
             })
         })
     }
@@ -859,9 +852,7 @@ impl AnthropicEventMapper {
                 vec![Ok(LanguageModelCompletionEvent::Stop(self.stop_reason))]
             }
             Event::Error { error } => {
-                vec![Err(LanguageModelCompletionError::Other(anyhow!(
-                    AnthropicError::ApiError(error)
-                )))]
+                vec![Err(error.into())]
             }
             _ => Vec::new(),
         }
@@ -874,16 +865,6 @@ struct RawToolUse {
     input_json: String,
 }
 
-pub fn anthropic_err_to_anyhow(err: AnthropicError) -> anyhow::Error {
-    if let AnthropicError::ApiError(api_err) = &err {
-        if let Some(tokens) = api_err.match_window_exceeded() {
-            return anyhow!(LanguageModelKnownError::ContextWindowLimitExceeded { tokens });
-        }
-    }
-
-    anyhow!(err)
-}
-
 /// Updates usage data by preferring counts from `new`.
 fn update_usage(usage: &mut Usage, new: &Usage) {
     if let Some(input_tokens) = new.input_tokens {