feat: initial release v0.3.0

tests/unit/ai-runtime/errors.test.ts

@@ -0,0 +1,34 @@
+import { describe, expect, it } from 'vitest'
+import { toAiRuntimeError } from '@/lib/ai-runtime/errors'
+
+describe('toAiRuntimeError empty response mapping', () => {
+  it('maps nested Gemini empty response signal to EMPTY_RESPONSE even when status is 429', () => {
+    const upstreamError = new Error('Too Many Requests') as Error & {
+      status?: number
+      cause?: unknown
+    }
+    upstreamError.status = 429
+    upstreamError.cause = {
+      error: {
+        message: 'received empty response from Gemini: no meaningful content in candidates (request id: x)',
+        type: 'channel_error',
+        code: 'channel:empty_response',
+      },
+      code: 429,
+      status: 'Too Many Requests',
+    }
+
+    const runtimeError = toAiRuntimeError(upstreamError)
+    expect(runtimeError.code).toBe('EMPTY_RESPONSE')
+    expect(runtimeError.retryable).toBe(true)
+  })
+
+  it('keeps RATE_LIMIT when there is no empty response signal', () => {
+    const runtimeError = toAiRuntimeError({
+      status: 429,
+      message: 'Too Many Requests',
+    })
+    expect(runtimeError.code).toBe('RATE_LIMIT')
+    expect(runtimeError.retryable).toBe(true)
+  })
+})