diff --git a/README.md b/README.md
index 143a93d..70869f3 100644
--- a/README.md
+++ b/README.md
@@ -2,7 +2,7 @@
 
 ## Introduction
 
-Provides the same agent OpenAI API interface for different LLM models and supports deployment to any Edge Runtime environment.
+Provides the same proxy OpenAI API interface for different LLM models, and supports deployment to any Edge Runtime environment.
 
 Supported models
 
@@ -18,25 +18,25 @@ Supported models
 
 Environment variables
 
-- `API_KEY`: proxy API Key, which must be set when calling the proxy API
+- `API_KEY`: Proxy API Key, required when calling the proxy API
 
-- OpenAI: Support OpenAI models, such as `gpt-4o-mini`
+- OpenAI: Supports OpenAI models, e.g. `gpt-4o-mini`
   - `OPENAI_API_KEY`: OpenAI API Key
-- VertexAI Anthropic: Support Anthropic models on Google Vertex AI, such as `claude-3-5-sonnet@20240620`
+- VertexAI Anthropic: Supports Anthropic models on Google Vertex AI, e.g. `claude-3-5-sonnet@20240620`
   - `VERTEX_ANTROPIC_GOOGLE_SA_CLIENT_EMAIL`: Google Cloud Service Account Email
   - `VERTEX_ANTROPIC_GOOGLE_SA_PRIVATE_KEY`: Google Cloud Service Account Private Key
   - `VERTEX_ANTROPIC_REGION`: Google Vertex AI Anthropic Region
   - `VERTEX_ANTROPIC_PROJECTID`: Google Vertex AI Anthropic Project ID
-- Anthropic: Support Anthropic models, such as `claude-3-5-sonnet-20240620`
+- Anthropic: Supports Anthropic models, e.g. `claude-3-5-sonnet-20240620`
   - `ANTROPIC_API_KEY`: Anthropic API Key
-- Google Gemini: Support Google Gemini models, such as `gemini-1.5-flash`
+- Google Gemini: Supports Google Gemini models, e.g. `gemini-1.5-flash`
   - `GOOGLE_GEN_AI_API_KEY`: Google Gemini API Key
 
 ## Usage
 
-Once deployed successfully, different models can be called through OpenAI’s API interface.
+Once deployed successfully, you can call different models through OpenAI's API interface.
 
-For example, calling OpenAI’s API interface:
+For example, calling OpenAI's API interface:
 
 ```bash
 curl http://localhost:8787/v1/chat/completions \
@@ -53,7 +53,7 @@ curl http://localhost:8787/v1/chat/completions \
    }'
 ```
 
-Or call Anthropic’s API interface:
+Or calling Anthropic's API interface:
 
 ```bash
 curl http://localhost:8787/v1/chat/completions \
@@ -70,7 +70,7 @@ curl http://localhost:8787/v1/chat/completions \
    }'
 ```
 
-And can be used in OpenAI’s official SDK, for example:
+And it can be used in OpenAI's official SDK, for example:
 
 ```ts
 const openai = new OpenAI({
diff --git a/package.json b/package.json
index b13e6d6..a032015 100644
--- a/package.json
+++ b/package.json
@@ -3,7 +3,8 @@
   "type": "module",
   "scripts": {
     "dev": "wrangler dev src/index.ts",
-    "deploy": "wrangler deploy --minify src/index.ts"
+    "deploy": "wrangler deploy --minify src/index.ts",
+    "test": "vitest run"
   },
   "dependencies": {
     "@anthropic-ai/sdk": "^0.27.0",
diff --git a/src/__tests__/index.test.ts b/src/__tests__/index.test.ts
index 1316e21..755e4b3 100644
--- a/src/__tests__/index.test.ts
+++ b/src/__tests__/index.test.ts
@@ -1,11 +1,12 @@
 import { beforeAll, beforeEach, describe, expect, it } from 'vitest'
 import app from '..'
 import OpenAI from 'openai'
-import { omit } from 'lodash-es'
+import { omit, pick } from 'lodash-es'
 import { AnthropicVertex } from '@anthropic-ai/vertex-sdk'
 import { GoogleAuth } from 'google-auth-library'
 import { TextBlock } from '@anthropic-ai/sdk/resources/messages.mjs'
 import { anthropic } from '../llm/anthropic'
+import { testClient } from 'hono/testing'
 
 const MOCK_ENV = {
   API_KEY: import.meta.env.VITE_API_KEY,
@@ -196,7 +197,13 @@ describe('stream', () => {
 describe('list models', async () => {
   it('no api key', async () => {
     const r = (await (
-      await app.request('/v1/models', {}, {})
+      await app.request(
+        '/v1/models',
+        {
+          headers: { Authorization: 'Bearer ' + import.meta.env.VITE_API_KEY },
+        },
+        pick(MOCK_ENV, 'API_KEY'),
+      )
     ).json()) as OpenAI.Models.ModelsPage
     expect(r.data).empty
   })
@@ -204,18 +211,15 @@ describe('list models', async () => {
     const r = (await (
       await app.request(
         '/v1/models',
-        {},
-        { ANTROPIC_API_KEY: import.meta.env.VITE_ANTROPIC_API_KEY },
+        {
+          headers: {
+            Authorization: 'Bearer ' + import.meta.env.VITE_API_KEY,
+          },
+        },
+        pick(MOCK_ENV, 'API_KEY', 'ANTROPIC_API_KEY'),
       )
     ).json()) as OpenAI.Models.ModelsPage
-    expect(r.data).deep.eq(
-      anthropic({} as any).supportModels.map((it) => ({
-        id: it,
-        object: 'model',
-        owned_by: 'system',
-        created: 0,
-      })),
-    )
+    expect(r.data.map((it) => it.id)).deep.eq(anthropic(MOCK_ENV).supportModels)
   })
   it('openai api key', async () => {
     const client = new OpenAI({
@@ -225,6 +229,6 @@ describe('list models', async () => {
       },
     })
     const r = await client.models.list()
-    console.log(r)
+    expect(r.data).not.empty
   })
 })
diff --git a/src/claude/__tests__/authenticate.test.ts b/src/claude/__tests__/authenticate.test.ts
index b242d39..649c39c 100644
--- a/src/claude/__tests__/authenticate.test.ts
+++ b/src/claude/__tests__/authenticate.test.ts
@@ -4,16 +4,16 @@ import { it, expect } from 'vitest'
 
 it('should authenticate', async () => {
   const token = await authenticate({
-    clientEmail: process.env.VITE_GOOGLE_SA_CLIENT_EMAIL!,
-    privateKey: process.env.VITE_GOOGLE_SA_PRIVATE_KEY!,
+    clientEmail: import.meta.env.VITE_VERTEX_ANTROPIC_GOOGLE_SA_CLIENT_EMAIL!,
+    privateKey: import.meta.env.VITE_VERTEX_ANTROPIC_GOOGLE_SA_PRIVATE_KEY!,
   })
   expect(token).toBeDefined()
 })
 
 it('should call anthropic vertex', async () => {
   const token = await authenticate({
-    clientEmail: process.env.VITE_GOOGLE_SA_CLIENT_EMAIL!,
-    privateKey: process.env.VITE_GOOGLE_SA_PRIVATE_KEY!,
+    clientEmail: import.meta.env.VITE_VERTEX_ANTROPIC_GOOGLE_SA_CLIENT_EMAIL!,
+    privateKey: import.meta.env.VITE_VERTEX_ANTROPIC_GOOGLE_SA_PRIVATE_KEY!,
   })
   const client = new AnthropicVertexWeb({
     accessToken: token.access_token,
@@ -42,8 +42,8 @@ function buildUrl({
 
 it('should authenticate on edge runtime', async () => {
   const token = await authenticate({
-    clientEmail: process.env.VITE_GOOGLE_SA_CLIENT_EMAIL!,
-    privateKey: process.env.VITE_GOOGLE_SA_PRIVATE_KEY!,
+    clientEmail: import.meta.env.VITE_VERTEX_ANTROPIC_GOOGLE_SA_CLIENT_EMAIL!,
+    privateKey: import.meta.env.VITE_VERTEX_ANTROPIC_GOOGLE_SA_PRIVATE_KEY!,
   })
   expect(token).toBeDefined()
   const url = buildUrl({
diff --git a/src/claude/__tests__/AnthropicVertex.test.ts b/src/claude/__tests__/web.test.ts
similarity index 96%
rename from src/claude/__tests__/AnthropicVertex.test.ts
rename to src/claude/__tests__/web.test.ts
index 184fa2e..0679307 100644
--- a/src/claude/__tests__/AnthropicVertex.test.ts
+++ b/src/claude/__tests__/web.test.ts
@@ -1,6 +1,9 @@
 import { expect, it } from 'vitest'
 import { AnthropicVertexWeb } from '../web'
-import { AnthropicVertex, AnthropicVertex as OriginAnthropicVertex } from '@anthropic-ai/vertex-sdk'
+import {
+  AnthropicVertex,
+  AnthropicVertex as OriginAnthropicVertex,
+} from '@anthropic-ai/vertex-sdk'
 import { authenticate } from '../authenticate'
 import { omit } from 'lodash-es'
 import { ChatAnthropic } from 'langchain-anthropic-edge'
@@ -53,7 +56,7 @@ it('should call Custom Client and Origin Client', async () => {
   expect(r1).not.undefined
   expect(r2).not.undefined
   expect(omit(r1, 'id')).deep.equal(omit(r2, 'id'))
-}, 10_000)
+}, 20_000)
 
 it('Call Anthropic Vertex on nodejs', async () => {
   const client = new AnthropicVertex({
diff --git a/src/claude/authenticate.ts b/src/claude/authenticate.ts
index c20d903..e0e12f2 100644
--- a/src/claude/authenticate.ts
+++ b/src/claude/authenticate.ts
@@ -62,6 +62,9 @@ export async function authenticate(options: {
   clientEmail: string
   privateKey: string
 }): Promise<Token> {
+  if (!options.clientEmail || !options.privateKey) {
+    throw new Error('clientEmail and privateKey are required')
+  }
   if (token === null) {
     token = await createToken(options)
   } else if (token.expires_at < Math.floor(Date.now() / 1000)) {
diff --git a/src/llm/anthropic.ts b/src/llm/anthropic.ts
index 246a8a7..b2014b9 100644
--- a/src/llm/anthropic.ts
+++ b/src/llm/anthropic.ts
@@ -149,7 +149,6 @@ export function anthropicBase(
       signal.onabort = () => stream.controller.abort()
       const chunks: AnthropicTypes.Messages.RawMessageStreamEvent[] = []
       let start: AnthropicTypes.Messages.Message | undefined
-      let delta: AnthropicTypes.Messages.MessageDeltaEvent | undefined
       for await (const it of stream) {
         chunks.push(it)
         const fileds = () => ({
@@ -176,28 +175,29 @@ export function anthropicBase(
             ],
           } as OpenAI.ChatCompletionChunk
         } else if (it.type === 'message_delta') {
-          delta = it
-          yield {
-            ...fileds(),
-            choices: [
-              {
-                index: 0,
-                delta: {},
-                finish_reason: 'stop',
+          if (req.stream_options?.include_usage) {
+            yield {
+              ...fileds(),
+              choices: [],
+              usage: {
+                prompt_tokens: start!.usage.input_tokens,
+                completion_tokens: it!.usage.output_tokens,
+                total_tokens:
+                  start!.usage.input_tokens + it!.usage.output_tokens,
               },
-            ],
-          } as OpenAI.ChatCompletionChunk
-        } else if (it.type === 'message_stop') {
-          yield {
-            ...fileds(),
-            choices: [],
-            usage: {
-              prompt_tokens: start!.usage.input_tokens,
-              completion_tokens: delta!.usage.output_tokens,
-              total_tokens:
-                start!.usage.input_tokens + delta!.usage.output_tokens,
-            },
-          } as OpenAI.ChatCompletionChunk
+            } as OpenAI.ChatCompletionChunk
+          } else {
+            yield {
+              ...fileds(),
+              choices: [
+                {
+                  index: 0,
+                  delta: {},
+                  finish_reason: 'stop',
+                },
+              ],
+            } as OpenAI.ChatCompletionChunk
+          }
         }
       }
     },
diff --git a/src/llm/google.ts b/src/llm/google.ts
index b2ce356..b613e5f 100644
--- a/src/llm/google.ts
+++ b/src/llm/google.ts
@@ -1,5 +1,5 @@
 import { IChat } from './base'
-import { GoogleGenerativeAI, Tool } from '@google/generative-ai'
+import { GoogleGenerativeAI } from '@google/generative-ai'
 import GoogleAI from '@google/generative-ai'
 import OpenAI from 'openai'
 
@@ -128,7 +128,11 @@ export function google(env: Record<string, string>): IChat {
           ],
         } as OpenAI.ChatCompletionChunk
       }
-      if (req.stream_options?.include_usage && last) {
+      if (!last) {
+        throw new Error('No response from google')
+      }
+
+      if (req.stream_options?.include_usage) {
         yield {
           ...fields(),
           choices: [],
@@ -138,6 +142,17 @@ export function google(env: Record<string, string>): IChat {
             total_tokens: last.usageMetadata!.totalTokenCount,
           },
         } as OpenAI.ChatCompletionChunk
+      } else {
+        yield {
+          ...fields(),
+          choices: [
+            {
+              index: 0,
+              delta: {},
+              finish_reason: 'stop',
+            },
+          ],
+        } as OpenAI.ChatCompletionChunk
       }
     },
   }