feat(server): add /v1/models endpoint for OpenAI-compatible model discovery

Open WebUI and other OpenAI-compatible clients call GET /v1/models to
discover available models. Lists all models from enabled providers via
the model registry, respects disabled models, and handles Ollama models
from TOML config.

src/server/mod.rs

@@ -3,7 +3,7 @@ use axum::{
     extract::State,
     response::IntoResponse,
     response::sse::{Event, Sse},
-    routing::post,
+    routing::{get, post},
 };
 use futures::stream::StreamExt;
 use std::sync::Arc;
@@ -24,6 +24,7 @@ use crate::{
 pub fn router(state: AppState) -> Router {
     Router::new()
         .route("/v1/chat/completions", post(chat_completions))
+        .route("/v1/models", get(list_models))
         .layer(axum::middleware::from_fn_with_state(
             state.clone(),
             rate_limiting::middleware::rate_limit_middleware,
@@ -31,6 +32,60 @@ pub fn router(state: AppState) -> Router {
         .with_state(state)
 }
 
+/// GET /v1/models — OpenAI-compatible model listing.
+/// Returns all models from enabled providers so clients like Open WebUI can
+/// discover which models are available through the proxy.
+async fn list_models(
+    State(state): State<AppState>,
+    _auth: AuthenticatedClient,
+) -> Result<Json<serde_json::Value>, AppError> {
+    let registry = &state.model_registry;
+    let providers = state.provider_manager.get_all_providers().await;
+
+    let mut models = Vec::new();
+
+    for provider in &providers {
+        let provider_name = provider.name();
+
+        // Find this provider's models in the registry
+        if let Some(provider_info) = registry.providers.get(provider_name) {
+            for (model_id, meta) in &provider_info.models {
+                // Skip disabled models via the config cache
+                if let Some(cfg) = state.model_config_cache.get(model_id).await {
+                    if !cfg.enabled {
+                        continue;
+                    }
+                }
+
+                models.push(serde_json::json!({
+                    "id": model_id,
+                    "object": "model",
+                    "created": 0,
+                    "owned_by": provider_name,
+                    "name": meta.name,
+                }));
+            }
+        }
+
+        // For Ollama, models are configured in the TOML, not the registry
+        if provider_name == "ollama" {
+            for model_id in &state.config.providers.ollama.models {
+                models.push(serde_json::json!({
+                    "id": model_id,
+                    "object": "model",
+                    "created": 0,
+                    "owned_by": "ollama",
+                }));
+            }
+        }
+    }
+
+    Ok(Json(serde_json::json!({
+        "object": "list",
+        "data": models
+    })))
+}
+
 async fn get_model_cost(
     model: &str,
     prompt_tokens: u32,