feat: implement model selection for --auto and add corresponding tests

Author: lanternian

src/providers/copilot/client.ts

@@ -36,6 +36,7 @@ const shouldRetryResponse = (response: Response): boolean =>
 
 const buildHeaders = (
   provider: CopilotProviderContext,
+  path: string,
   init: RequestInit,
   options: FetchCopilotOptions,
 ): Headers => {
@@ -47,10 +48,7 @@ const buildHeaders = (
   headers.set("editor-plugin-version", EDITOR_PLUGIN_VERSION)
   headers.set("user-agent", USER_AGENT)
   headers.set("openai-intent", "conversation-panel")
-  headers.set(
-    "x-github-api-version",
-    runtimeState.autoSessionToken ? AUTO_MODE_API_VERSION : API_VERSION,
-  )
+  headers.set("x-github-api-version", API_VERSION)
   headers.set("x-request-id", randomUUID())
   headers.set("x-vscode-user-agent-library-version", "electron-fetch")
 
@@ -62,8 +60,9 @@ const buildHeaders = (
     headers.set("x-initiator", options.initiator)
   }
 
-  if (runtimeState.autoSessionToken) {
-    headers.set("copilot-session-token", runtimeState.autoSessionToken)
+  if (runtimeState.autoSessionToken && (path.startsWith("/chat/completions") || path.startsWith("/responses"))) {
+    headers.set("copilot-session-token", runtimeState.autoSessionToken),
+    headers.set("x-github-api-version", AUTO_MODE_API_VERSION)
   }
 
   if (!headers.has("content-type") && init.body !== undefined) {
@@ -132,7 +131,7 @@ export const fetchCopilot = async (
       await traceCopilotRequest(path, init, options, attempt)
       const response = await fetch(`${provider.baseUrl}${path}`, {
         ...init,
-        headers: buildHeaders(provider, init, options),
+        headers: buildHeaders(provider, path, init, options),
       })
 
       if (!shouldRetryResponse(response) || attempt === MAX_FETCH_ATTEMPTS) {

src/start.ts

@@ -174,22 +174,6 @@ export const start = defineCommand({
       showToken: args["show-token"],
     })
 
-    if (args.auto) {
-      try {
-        const session = await enableAutoMode(config)
-        consola.success(
-          `Auto mode enabled${
-            session.available_models?.length ?
-              ` (available models: ${session.available_models.join(", ")})`
-            : ""
-          }`,
-        )
-      } catch (error) {
-        consola.error("Failed to enable auto mode:", error)
-        process.exit(1)
-      }
-    }
-
     const server = startServer(config)
 
     consola.success(
@@ -300,8 +284,15 @@ export const start = defineCommand({
         )
       }
     }
-
-    const models = await fetchAvailableModels(config)
+    
+    let models
+    if (args.auto) {
+      const session = await enableAutoMode(config)
+      models = session.available_models ?? []
+    }
+    else {
+      models = await fetchAvailableModels(config)    
+    }
     const supportedIds = new Set(
       MODEL_CAPABILITIES.flatMap((m) => [m.id, ...(m.aliases ?? [])]),
     )
@@ -316,16 +307,7 @@ export const start = defineCommand({
             .map((id) => getPublicModelId(id)),
         )
       : fallbackModelIds
-    const autoAllowed =
-      args.auto && runtimeState.autoAvailableModels?.length ?
-        new Set(runtimeState.autoAvailableModels.map(getPublicModelId))
-      : undefined
-    const autoFilteredPickable =
-      autoAllowed ? pickable.filter((id) => autoAllowed.has(id)) : pickable
-    const finalPickable =
-      autoFilteredPickable.length > 0 ? autoFilteredPickable
-      : pickable.length > 0 ? pickable
-      : fallbackModelIds
+    const finalPickable = pickable.length > 0 ? pickable : fallbackModelIds
     if (models.length > 0) {
       consola.info(
         `Available models:\n${models