Add OpenAI/LiteLLM provider support with settings UI

- Add openai/ provider module with translator, client, convert, request/response translators
- Add shared provider/ types (ConversationMessage, MessageRole, ProviderConfig enum)
- Wire OpenAI-compatible provider dispatch alongside Bedrock in response_stream.rs
- Add ai.openai.* settings (enabled, base_url, api_key, model, models)
- Add OpenAI/LiteLLM settings page with model fetch, picker, and config UI
- Extend model menu items and llms.rs to surface LiteLLM models
- Update WARP.md with OpenAI provider architecture docs

crates/input_classifier/src/heuristic_classifier/mod.rs

@@ -55,7 +55,8 @@ impl InputClassifier for HeuristicClassifier {
             return InputType::Shell;
         }
 
-        let result = self.classify_input(input, context)
+        let result = self
+            .classify_input(input, context)
             .await
             .map(|result| result.to_input_type())
             .unwrap_or(context.current_input_type);

crates/local_inference/src/engine/mod.rs

@@ -2,8 +2,8 @@ mod generation;
 mod model_loader;
 
 use std::path::PathBuf;
-use std::sync::atomic::{AtomicBool, Ordering};
 use std::sync::Arc;
+use std::sync::atomic::{AtomicBool, Ordering};
 
 use anyhow::{Context as _, Result};
 use candle_core::{DType, Tensor};

crates/local_inference/src/tasks/input_classification.rs

@@ -91,7 +91,9 @@ impl InputClassificationTask {
             temperature: 0.0,
             top_p: 1.0,
         };
-        let output = engine.generate_cancellable(&prompt, &config, cancel).await?;
+        let output = engine
+            .generate_cancellable(&prompt, &config, cancel)
+            .await?;
         Ok(Self::parse_output(&output))
     }
 }

crates/local_inference/tests/e2e_scenarios.rs

@@ -179,7 +179,11 @@ async fn scenario_tab_naming_development_contexts() {
         },
         TabCase {
             label: "Python data science",
-            commands: vec!["jupyter notebook", "pip install pandas", "python analysis.py"],
+            commands: vec![
+                "jupyter notebook",
+                "pip install pandas",
+                "python analysis.py",
+            ],
             cwd: "/home/user/research/data-pipeline",
         },
         TabCase {

crates/local_inference/tests/integration.rs

@@ -270,7 +270,10 @@ async fn test_classify_multiple_sequential_calls() {
         ("git status", InputCategory::Shell),
         ("what does this error mean?", InputCategory::AgentPrompt),
         ("npm install express", InputCategory::Shell),
-        ("refactor this to use async/await", InputCategory::AgentPrompt),
+        (
+            "refactor this to use async/await",
+            InputCategory::AgentPrompt,
+        ),
     ];
 
     for (input, expected_category) in inputs {
@@ -286,7 +289,10 @@ async fn test_classify_multiple_sequential_calls() {
             .await
             .expect("classification failed");
 
-        println!("\"{input}\" -> {:?} (confidence: {:.2})", result.category, result.confidence);
+        println!(
+            "\"{input}\" -> {:?} (confidence: {:.2})",
+            result.category, result.confidence
+        );
         assert_eq!(
             result.category, expected_category,
             "expected {expected_category:?} for \"{input}\", got {:?}",