From 882e298ddeb9b0da739af373965f55e724402a3e Mon Sep 17 00:00:00 2001
From: rejectliu <rejectliu@qq.com>
Date: Wed, 19 Nov 2025 16:00:36 +0800
Subject: [PATCH 01/13] update compose config

---
 .env.example       | 1 +
 .gitignore         | 4 +++-
 docker-compose.yml | 9 +++++++++
 3 files changed, 13 insertions(+), 1 deletion(-)
 create mode 100644 .env.example
 create mode 100644 docker-compose.yml

diff --git a/.env.example b/.env.example
new file mode 100644
index 000000000..6e237e5f5
--- /dev/null
+++ b/.env.example
@@ -0,0 +1 @@
+GH_TOKEN=
diff --git a/.gitignore b/.gitignore
index 577a4f199..73ebed921 100644
--- a/.gitignore
+++ b/.gitignore
@@ -11,4 +11,6 @@ node_modules/
 .eslintcache
 
 # build output
-dist/
\ No newline at end of file
+dist/
+
+.env
diff --git a/docker-compose.yml b/docker-compose.yml
new file mode 100644
index 000000000..ffed9d5f0
--- /dev/null
+++ b/docker-compose.yml
@@ -0,0 +1,9 @@
+version: "3.8"
+services:
+  copilot-api:
+    build: .
+    ports:
+      - "4141:4141"
+    environment:
+      - GH_TOKEN=${GH_TOKEN}
+    restart: unless-stopped

From 6cb662de9e1cf3ab23a79dfaa1165fd9bf5aeda6 Mon Sep 17 00:00:00 2001
From: rejectliu <rejectliu@qq.com>
Date: Wed, 19 Nov 2025 16:04:55 +0800
Subject: [PATCH 02/13] update compose config

---
 docker-compose.yml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/docker-compose.yml b/docker-compose.yml
index ffed9d5f0..c1ce596de 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -1,6 +1,7 @@
 version: "3.8"
 services:
   copilot-api:
+    container_name: copilot-api
     build: .
     ports:
       - "4141:4141"

From 3b0bdf1a37377bff006fbeb4596775a398a09404 Mon Sep 17 00:00:00 2001
From: rejectliu <rejectliu@qq.com>
Date: Sun, 14 Dec 2025 22:15:09 +0800
Subject: [PATCH 03/13] =?UTF-8?q?docker=20=E6=94=AF=E6=8C=81=E6=8C=81?=
 =?UTF-8?q?=E4=B9=85=E5=8C=96token=E5=88=B0=E5=AE=B9=E5=99=A8=E5=86=85?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 docker-compose.yml | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/docker-compose.yml b/docker-compose.yml
index c1ce596de..f3bd18691 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -5,6 +5,10 @@ services:
     build: .
     ports:
       - "4141:4141"
-    environment:
-      - GH_TOKEN=${GH_TOKEN}
+    env_file: .env
+    volumes:
+      - copilot-data:/root/.local/share/copilot-api
     restart: unless-stopped
+
+volumes:
+  copilot-data:

From 67bc65ccf494828ab72dbc5c51217f93fbe67620 Mon Sep 17 00:00:00 2001
From: rejectliu <rejectliu@qq.com>
Date: Mon, 15 Dec 2025 08:06:11 +0800
Subject: [PATCH 04/13] remove .env

---
 docker-compose.yml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/docker-compose.yml b/docker-compose.yml
index f3bd18691..1b1f502fd 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -5,7 +5,6 @@ services:
     build: .
     ports:
       - "4141:4141"
-    env_file: .env
     volumes:
       - copilot-data:/root/.local/share/copilot-api
     restart: unless-stopped

From 64b67e3824b0b039dda2cf98034b5b5a985f96c8 Mon Sep 17 00:00:00 2001
From: rejectliu <rejectliu@qq.com>
Date: Mon, 15 Dec 2025 08:06:35 +0800
Subject: [PATCH 05/13] remove version

---
 docker-compose.yml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/docker-compose.yml b/docker-compose.yml
index 1b1f502fd..61d744e2d 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -1,4 +1,3 @@
-version: "3.8"
 services:
   copilot-api:
     container_name: copilot-api

From b89199d5e0172731f36847ff0bf28d98f47a94f8 Mon Sep 17 00:00:00 2001
From: rejectliu <rejectliu@qq.com>
Date: Sat, 3 Jan 2026 12:24:39 +0800
Subject: [PATCH 06/13] =?UTF-8?q?fix:=20=E5=86=85=E9=83=A8=E5=BC=BA?=
 =?UTF-8?q?=E5=88=B6=E4=BD=BF=E7=94=A8=E6=B5=81=E5=BC=8F=E6=A8=A1=E5=BC=8F?=
 =?UTF-8?q?=E9=81=BF=E5=85=8D=E9=95=BF=E6=97=B6=E9=97=B4=E8=AF=B7=E6=B1=82?=
 =?UTF-8?q?=E8=B6=85=E6=97=B6?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- 新增 createChatCompletionsStream 函数，强制使用流式模式
- 非流式请求改为收集流式数据块后合并返回
- 解决 ECONNRESET 超时问题

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
---
 bun.lock                                      |   1 +
 src/routes/chat-completions/handler.ts        | 185 ++++++++++++++++--
 .../copilot/create-chat-completions.ts        |  40 +++-
 3 files changed, 211 insertions(+), 15 deletions(-)

diff --git a/bun.lock b/bun.lock
index 20e895e7f..9ece87578 100644
--- a/bun.lock
+++ b/bun.lock
@@ -1,5 +1,6 @@
 {
   "lockfileVersion": 1,
+  "configVersion": 0,
   "workspaces": {
     "": {
       "name": "copilot-api",
diff --git a/src/routes/chat-completions/handler.ts b/src/routes/chat-completions/handler.ts
index 04a5ae9ed..265bd692b 100644
--- a/src/routes/chat-completions/handler.ts
+++ b/src/routes/chat-completions/handler.ts
@@ -1,3 +1,4 @@
+import type { ServerSentEventMessage } from "fetch-event-stream"
 import type { Context } from "hono"
 
 import consola from "consola"
@@ -9,7 +10,8 @@ import { state } from "~/lib/state"
 import { getTokenCount } from "~/lib/tokenizer"
 import { isNullish } from "~/lib/utils"
 import {
-  createChatCompletions,
+  createChatCompletionsStream,
+  type ChatCompletionChunk,
   type ChatCompletionResponse,
   type ChatCompletionsPayload,
 } from "~/services/copilot/create-chat-completions"
@@ -47,22 +49,177 @@ export async function handleCompletion(c: Context) {
     consola.debug("Set max_tokens to:", JSON.stringify(payload.max_tokens))
   }
 
-  const response = await createChatCompletions(payload)
+  // 记录客户端是否请求流式响应
+  const clientWantsStream = payload.stream === true
 
-  if (isNonStreaming(response)) {
-    consola.debug("Non-streaming response:", JSON.stringify(response))
-    return c.json(response)
+  // 内部始终使用流式模式，避免长时间请求超时导致 ECONNRESET
+  const response = await createChatCompletionsStream(payload)
+
+  // 如果客户端请求流式响应，直接透传
+  if (clientWantsStream) {
+    consola.debug("Streaming response")
+    return streamSSE(c, async (stream) => {
+      for await (const chunk of response) {
+        consola.debug("Streaming chunk:", JSON.stringify(chunk))
+        await stream.writeSSE(chunk as SSEMessage)
+      }
+    })
   }
 
-  consola.debug("Streaming response")
-  return streamSSE(c, async (stream) => {
-    for await (const chunk of response) {
-      consola.debug("Streaming chunk:", JSON.stringify(chunk))
-      await stream.writeSSE(chunk as SSEMessage)
+  // 客户端请求非流式响应，收集流式数据块并合并
+  consola.debug("Collecting stream chunks for non-streaming response")
+  const nonStreamResponse = await collectStreamToResponse(response)
+  consola.debug("Non-streaming response:", JSON.stringify(nonStreamResponse))
+  return c.json(nonStreamResponse)
+}
+
+type FinishReason = "stop" | "length" | "tool_calls" | "content_filter"
+
+type ToolCallAccumulator = {
+  id: string
+  type: "function"
+  function: { name: string; arguments: string }
+}
+
+type StreamAccumulator = {
+  id: string
+  model: string
+  created: number
+  systemFingerprint?: string
+  finishReason: FinishReason
+  content: string
+  toolCalls: Map<number, ToolCallAccumulator>
+  usage?: ChatCompletionResponse["usage"]
+}
+
+/**
+ * 将流式响应的数据块合并为非流式响应格式
+ */
+async function collectStreamToResponse(
+  stream: AsyncGenerator<ServerSentEventMessage, void, unknown>,
+): Promise<ChatCompletionResponse> {
+  const accumulator = createAccumulator()
+
+  for await (const chunk of stream) {
+    if (!chunk.data) continue
+    if (chunk.data === "[DONE]") break
+
+    const parsed = parseChunkDataOrLog(chunk.data)
+    if (!parsed) continue
+    applyChunkToAccumulator(parsed, accumulator)
+  }
+
+  return buildResponse(accumulator)
+}
+
+function createAccumulator(): StreamAccumulator {
+  return {
+    id: "",
+    model: "",
+    created: 0,
+    finishReason: "stop",
+    content: "",
+    toolCalls: new Map(),
+  }
+}
+
+function parseChunkDataOrLog(data: unknown): ChatCompletionChunk | null {
+  if (typeof data !== "string") return null
+
+  try {
+    return JSON.parse(data) as ChatCompletionChunk
+  } catch (error) {
+    consola.debug("Failed to parse SSE chunk data", {
+      dataPreview: data.slice(0, 500),
+      error,
+    })
+    return null
+  }
+}
+
+function applyChunkToAccumulator(
+  parsed: ChatCompletionChunk,
+  accumulator: StreamAccumulator,
+) {
+  if (!accumulator.id && parsed.id) accumulator.id = parsed.id
+  if (!accumulator.model && parsed.model) accumulator.model = parsed.model
+  if (!accumulator.created && parsed.created)
+    accumulator.created = parsed.created
+  if (!accumulator.systemFingerprint && parsed.system_fingerprint) {
+    accumulator.systemFingerprint = parsed.system_fingerprint
+  }
+  if (parsed.usage) accumulator.usage = parsed.usage
+
+  const choice = parsed.choices.at(0)
+  if (!choice) return
+
+  if (choice.finish_reason) accumulator.finishReason = choice.finish_reason
+
+  if (typeof choice.delta.content === "string") {
+    accumulator.content += choice.delta.content
+  }
+
+  if (choice.delta.tool_calls) {
+    mergeToolCalls(accumulator.toolCalls, choice.delta.tool_calls)
+  }
+}
+
+function mergeToolCalls(
+  toolCalls: Map<number, ToolCallAccumulator>,
+  deltas: NonNullable<
+    ChatCompletionChunk["choices"][number]["delta"]["tool_calls"]
+  >,
+) {
+  for (const delta of deltas) {
+    const existing = toolCalls.get(delta.index)
+    if (!existing) {
+      toolCalls.set(delta.index, {
+        id: delta.id ?? "",
+        type: "function",
+        function: {
+          name: delta.function?.name ?? "",
+          arguments: delta.function?.arguments ?? "",
+        },
+      })
+      continue
+    }
+
+    if (!existing.id && delta.id) existing.id = delta.id
+    if (!existing.function.name && delta.function?.name) {
+      existing.function.name = delta.function.name
     }
-  })
+    if (delta.function?.arguments) {
+      existing.function.arguments += delta.function.arguments
+    }
+  }
 }
 
-const isNonStreaming = (
-  response: Awaited<ReturnType<typeof createChatCompletions>>,
-): response is ChatCompletionResponse => Object.hasOwn(response, "choices")
+function buildResponse(accumulator: StreamAccumulator): ChatCompletionResponse {
+  const response: ChatCompletionResponse = {
+    id: accumulator.id,
+    object: "chat.completion",
+    created: accumulator.created,
+    model: accumulator.model,
+    choices: [
+      {
+        index: 0,
+        message: {
+          role: "assistant",
+          content: accumulator.content || null,
+          ...(accumulator.toolCalls.size > 0 && {
+            tool_calls: Array.from(accumulator.toolCalls.values()),
+          }),
+        },
+        logprobs: null,
+        finish_reason: accumulator.finishReason,
+      },
+    ],
+  }
+
+  if (accumulator.systemFingerprint) {
+    response.system_fingerprint = accumulator.systemFingerprint
+  }
+  if (accumulator.usage) response.usage = accumulator.usage
+
+  return response
+}
diff --git a/src/services/copilot/create-chat-completions.ts b/src/services/copilot/create-chat-completions.ts
index 8534151da..ca2270ce3 100644
--- a/src/services/copilot/create-chat-completions.ts
+++ b/src/services/copilot/create-chat-completions.ts
@@ -1,5 +1,5 @@
 import consola from "consola"
-import { events } from "fetch-event-stream"
+import { events, type ServerSentEventMessage } from "fetch-event-stream"
 
 import { copilotHeaders, copilotBaseUrl } from "~/lib/api-config"
 import { HTTPError } from "~/lib/error"
@@ -46,6 +46,44 @@ export const createChatCompletions = async (
   return (await response.json()) as ChatCompletionResponse
 }
 
+/**
+ * 强制使用流式模式的版本，返回类型始终是 AsyncGenerator
+ * 用于避免非流式请求的超时问题
+ */
+export const createChatCompletionsStream = async (
+  payload: Omit<ChatCompletionsPayload, "stream">,
+): Promise<AsyncGenerator<ServerSentEventMessage, void, unknown>> => {
+  if (!state.copilotToken) throw new Error("Copilot token not found")
+
+  const enableVision = payload.messages.some(
+    (x) =>
+      typeof x.content !== "string"
+      && x.content?.some((x) => x.type === "image_url"),
+  )
+
+  const isAgentCall = payload.messages.some((msg) =>
+    ["assistant", "tool"].includes(msg.role),
+  )
+
+  const headers: Record<string, string> = {
+    ...copilotHeaders(state, enableVision),
+    "X-Initiator": isAgentCall ? "agent" : "user",
+  }
+
+  const response = await fetch(`${copilotBaseUrl(state)}/chat/completions`, {
+    method: "POST",
+    headers,
+    body: JSON.stringify({ ...payload, stream: true }),
+  })
+
+  if (!response.ok) {
+    consola.error("Failed to create chat completions", response)
+    throw new HTTPError("Failed to create chat completions", response)
+  }
+
+  return events(response)
+}
+
 // Streaming types
 
 export interface ChatCompletionChunk {

From 3733486aa2e32fbed4fd93a3cd0d8ab77d974efe Mon Sep 17 00:00:00 2001
From: rejectliu <rejectliu@qq.com>
Date: Sat, 3 Jan 2026 12:39:27 +0800
Subject: [PATCH 07/13] log to inform non-stream

---
 src/routes/chat-completions/handler.ts | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/src/routes/chat-completions/handler.ts b/src/routes/chat-completions/handler.ts
index 265bd692b..ee7e7b536 100644
--- a/src/routes/chat-completions/handler.ts
+++ b/src/routes/chat-completions/handler.ts
@@ -66,6 +66,10 @@ export async function handleCompletion(c: Context) {
     })
   }
 
+  consola.info(
+    "Client requested non-streaming response; using upstream streaming and collecting chunks",
+  )
+
   // 客户端请求非流式响应，收集流式数据块并合并
   consola.debug("Collecting stream chunks for non-streaming response")
   const nonStreamResponse = await collectStreamToResponse(response)

From daba1725a5308519a3000734507140b4ba334257 Mon Sep 17 00:00:00 2001
From: rejectliu <rejectliu@qq.com>
Date: Sat, 3 Jan 2026 12:43:09 +0800
Subject: [PATCH 08/13] Revert "log to inform non-stream"

This reverts commit 3733486aa2e32fbed4fd93a3cd0d8ab77d974efe.
---
 src/routes/chat-completions/handler.ts | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/src/routes/chat-completions/handler.ts b/src/routes/chat-completions/handler.ts
index ee7e7b536..265bd692b 100644
--- a/src/routes/chat-completions/handler.ts
+++ b/src/routes/chat-completions/handler.ts
@@ -66,10 +66,6 @@ export async function handleCompletion(c: Context) {
     })
   }
 
-  consola.info(
-    "Client requested non-streaming response; using upstream streaming and collecting chunks",
-  )
-
   // 客户端请求非流式响应，收集流式数据块并合并
   consola.debug("Collecting stream chunks for non-streaming response")
   const nonStreamResponse = await collectStreamToResponse(response)

From ae8f9d03d4e529ae0dcaf84eb0a596e6e5650f02 Mon Sep 17 00:00:00 2001
From: rejectliu <rejectliu@qq.com>
Date: Wed, 28 Jan 2026 11:07:50 +0800
Subject: [PATCH 09/13] feat: add manual token refresh command

Add `refresh-token` CLI command and `/token/refresh` API endpoint to manually
refresh Copilot token without restarting the container.

Usage: `docker compose exec copilot-api refresh-token`

Co-Authored-By: Claude (claude-opus-4.5) <noreply@anthropic.com>
---
 Dockerfile                |  4 +++-
 entrypoint.sh             | 18 +++++++++++-------
 refresh-token             |  5 +++++
 src/main.ts               |  9 ++++++++-
 src/refresh-token.ts      | 37 +++++++++++++++++++++++++++++++++++++
 src/routes/token/route.ts | 16 ++++++++++++++++
 6 files changed, 80 insertions(+), 9 deletions(-)
 create mode 100644 refresh-token
 create mode 100644 src/refresh-token.ts

diff --git a/Dockerfile b/Dockerfile
index 1265220ef..39be38106 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -21,5 +21,7 @@ HEALTHCHECK --interval=30s --timeout=5s --start-period=10s --retries=3 \
   CMD wget --spider -q http://localhost:4141/ || exit 1
 
 COPY entrypoint.sh /entrypoint.sh
-RUN chmod +x /entrypoint.sh
+COPY refresh-token /usr/local/bin/refresh-token
+RUN chmod +x /entrypoint.sh /usr/local/bin/refresh-token
+
 ENTRYPOINT ["/entrypoint.sh"]
diff --git a/entrypoint.sh b/entrypoint.sh
index dfe63c902..cc1c741ce 100644
--- a/entrypoint.sh
+++ b/entrypoint.sh
@@ -1,9 +1,13 @@
 #!/bin/sh
-if [ "$1" = "--auth" ]; then
-  # Run auth command
-  exec bun run dist/main.js auth
-else
-  # Default command
-  exec bun run dist/main.js start -g "$GH_TOKEN" "$@"
-fi
+case "$1" in
+  --auth)
+    exec bun run dist/main.js auth
+    ;;
+  refresh-token)
+    exec bun run dist/main.js refresh-token
+    ;;
+  *)
+    exec bun run dist/main.js start -g "$GH_TOKEN" "$@"
+    ;;
+esac
 
diff --git a/refresh-token b/refresh-token
new file mode 100644
index 000000000..20d14123f
--- /dev/null
+++ b/refresh-token
@@ -0,0 +1,5 @@
+#!/bin/sh
+set -e
+
+echo "Refreshing Copilot token..."
+bun run /app/dist/main.js refresh-token "$@"
diff --git a/src/main.ts b/src/main.ts
index 4f6ca784b..afe174d54 100644
--- a/src/main.ts
+++ b/src/main.ts
@@ -5,6 +5,7 @@ import { defineCommand, runMain } from "citty"
 import { auth } from "./auth"
 import { checkUsage } from "./check-usage"
 import { debug } from "./debug"
+import { refreshToken } from "./refresh-token"
 import { start } from "./start"
 
 const main = defineCommand({
@@ -13,7 +14,13 @@ const main = defineCommand({
     description:
       "A wrapper around GitHub Copilot API to make it OpenAI compatible, making it usable for other tools.",
   },
-  subCommands: { auth, start, "check-usage": checkUsage, debug },
+  subCommands: {
+    auth,
+    start,
+    "check-usage": checkUsage,
+    debug,
+    "refresh-token": refreshToken,
+  },
 })
 
 await runMain(main)
diff --git a/src/refresh-token.ts b/src/refresh-token.ts
new file mode 100644
index 000000000..1a8a956ab
--- /dev/null
+++ b/src/refresh-token.ts
@@ -0,0 +1,37 @@
+import { defineCommand } from "citty"
+import consola from "consola"
+
+export const refreshToken = defineCommand({
+  meta: {
+    name: "refresh-token",
+    description: "Manually refresh the Copilot token via API",
+  },
+  args: {
+    port: {
+      type: "string",
+      alias: "p",
+      default: "4141",
+      description: "The port the server is running on",
+    },
+  },
+  async run({ args }) {
+    const port = args.port
+    const url = `http://localhost:${port}/token/refresh`
+
+    try {
+      const response = await fetch(url, { method: "POST" })
+      const data = await response.json()
+
+      if (response.ok && data.success) {
+        consola.success("Token refreshed successfully")
+      } else {
+        consola.error("Failed to refresh token:", data.error || "Unknown error")
+        process.exit(1)
+      }
+    } catch (error) {
+      consola.error("Failed to connect to server:", error)
+      consola.info("Make sure the server is running on port", port)
+      process.exit(1)
+    }
+  },
+})
diff --git a/src/routes/token/route.ts b/src/routes/token/route.ts
index dd0456d9a..1b1ae41f1 100644
--- a/src/routes/token/route.ts
+++ b/src/routes/token/route.ts
@@ -1,6 +1,7 @@
 import { Hono } from "hono"
 
 import { state } from "~/lib/state"
+import { getCopilotToken } from "~/services/github/get-copilot-token"
 
 export const tokenRoute = new Hono()
 
@@ -14,3 +15,18 @@ tokenRoute.get("/", (c) => {
     return c.json({ error: "Failed to fetch token", token: null }, 500)
   }
 })
+
+tokenRoute.post("/refresh", async (c) => {
+  try {
+    const { token } = await getCopilotToken()
+    state.copilotToken = token
+    console.log("Copilot token manually refreshed")
+    return c.json({
+      success: true,
+      message: "Token refreshed successfully",
+    })
+  } catch (error) {
+    console.error("Error refreshing token:", error)
+    return c.json({ error: "Failed to refresh token", success: false }, 500)
+  }
+})

From 39c7c03cb71a1f9051d75dda841ce30cc82da416 Mon Sep 17 00:00:00 2001
From: rejectliu <rejectliu@qq.com>
Date: Thu, 29 Jan 2026 15:27:52 +0800
Subject: [PATCH 10/13] feat: auto refresh token on 401 with 1 hour cooldown

When a request fails with 401 (token expired), automatically attempt to
refresh the Copilot token and retry the request. Includes a 1 hour
cooldown to prevent repeated refresh attempts if the error is caused by
other issues.

Co-Authored-By: Claude (claude-opus-4.5) <noreply@anthropic.com>
---
 src/lib/token.ts                              | 29 +++++++++
 .../copilot/create-chat-completions.ts        | 59 ++++++++++---------
 2 files changed, 61 insertions(+), 27 deletions(-)

diff --git a/src/lib/token.ts b/src/lib/token.ts
index fc8d2785f..4b6b49549 100644
--- a/src/lib/token.ts
+++ b/src/lib/token.ts
@@ -12,6 +12,35 @@ import { state } from "./state"
 
 const readGithubToken = () => fs.readFile(PATHS.GITHUB_TOKEN_PATH, "utf8")
 
+const REFRESH_COOLDOWN_MS = 60 * 60 * 1000 // 1 hour
+let lastRefreshAttempt = 0
+
+export async function refreshCopilotTokenOnError(): Promise<boolean> {
+  const now = Date.now()
+  if (now - lastRefreshAttempt < REFRESH_COOLDOWN_MS) {
+    consola.warn(
+      "Token refresh on error skipped: cooldown not elapsed (1 hour limit)",
+    )
+    return false
+  }
+
+  lastRefreshAttempt = now
+  consola.info("Attempting to refresh Copilot token due to request error")
+
+  try {
+    const { token } = await getCopilotToken()
+    state.copilotToken = token
+    consola.info("Copilot token refreshed successfully after error")
+    if (state.showToken) {
+      consola.info("Refreshed Copilot token:", token)
+    }
+    return true
+  } catch (error) {
+    consola.error("Failed to refresh Copilot token on error:", error)
+    return false
+  }
+}
+
 const writeGithubToken = (token: string) =>
   fs.writeFile(PATHS.GITHUB_TOKEN_PATH, token)
 
diff --git a/src/services/copilot/create-chat-completions.ts b/src/services/copilot/create-chat-completions.ts
index ca2270ce3..8c915209c 100644
--- a/src/services/copilot/create-chat-completions.ts
+++ b/src/services/copilot/create-chat-completions.ts
@@ -4,10 +4,12 @@ import { events, type ServerSentEventMessage } from "fetch-event-stream"
 import { copilotHeaders, copilotBaseUrl } from "~/lib/api-config"
 import { HTTPError } from "~/lib/error"
 import { state } from "~/lib/state"
+import { refreshCopilotTokenOnError } from "~/lib/token"
 
-export const createChatCompletions = async (
+async function doFetch(
   payload: ChatCompletionsPayload,
-) => {
+  streamOverride?: boolean,
+) {
   if (!state.copilotToken) throw new Error("Copilot token not found")
 
   const enableVision = payload.messages.some(
@@ -16,23 +18,39 @@ export const createChatCompletions = async (
       && x.content?.some((x) => x.type === "image_url"),
   )
 
-  // Agent/user check for X-Initiator header
-  // Determine if any message is from an agent ("assistant" or "tool")
   const isAgentCall = payload.messages.some((msg) =>
     ["assistant", "tool"].includes(msg.role),
   )
 
-  // Build headers and add X-Initiator
   const headers: Record<string, string> = {
     ...copilotHeaders(state, enableVision),
     "X-Initiator": isAgentCall ? "agent" : "user",
   }
 
-  const response = await fetch(`${copilotBaseUrl(state)}/chat/completions`, {
+  const body =
+    streamOverride !== undefined ?
+      { ...payload, stream: streamOverride }
+    : payload
+
+  return fetch(`${copilotBaseUrl(state)}/chat/completions`, {
     method: "POST",
     headers,
-    body: JSON.stringify(payload),
+    body: JSON.stringify(body),
   })
+}
+
+export const createChatCompletions = async (
+  payload: ChatCompletionsPayload,
+) => {
+  let response = await doFetch(payload)
+
+  if (response.status === 401) {
+    consola.warn("Got 401, attempting token refresh")
+    const refreshed = await refreshCopilotTokenOnError()
+    if (refreshed) {
+      response = await doFetch(payload)
+    }
+  }
 
   if (!response.ok) {
     consola.error("Failed to create chat completions", response)
@@ -53,29 +71,16 @@ export const createChatCompletions = async (
 export const createChatCompletionsStream = async (
   payload: Omit<ChatCompletionsPayload, "stream">,
 ): Promise<AsyncGenerator<ServerSentEventMessage, void, unknown>> => {
-  if (!state.copilotToken) throw new Error("Copilot token not found")
-
-  const enableVision = payload.messages.some(
-    (x) =>
-      typeof x.content !== "string"
-      && x.content?.some((x) => x.type === "image_url"),
-  )
-
-  const isAgentCall = payload.messages.some((msg) =>
-    ["assistant", "tool"].includes(msg.role),
-  )
+  let response = await doFetch(payload as ChatCompletionsPayload, true)
 
-  const headers: Record<string, string> = {
-    ...copilotHeaders(state, enableVision),
-    "X-Initiator": isAgentCall ? "agent" : "user",
+  if (response.status === 401) {
+    consola.warn("Got 401, attempting token refresh")
+    const refreshed = await refreshCopilotTokenOnError()
+    if (refreshed) {
+      response = await doFetch(payload as ChatCompletionsPayload, true)
+    }
   }
 
-  const response = await fetch(`${copilotBaseUrl(state)}/chat/completions`, {
-    method: "POST",
-    headers,
-    body: JSON.stringify({ ...payload, stream: true }),
-  })
-
   if (!response.ok) {
     consola.error("Failed to create chat completions", response)
     throw new HTTPError("Failed to create chat completions", response)

From 9a296d30436ea62653b51c74cf7197b31cbb6659 Mon Sep 17 00:00:00 2001
From: rejectliu <rejectliu@qq.com>
Date: Sat, 28 Feb 2026 20:36:16 +0800
Subject: [PATCH 11/13] feat: centralize model name normalization and add model
 aliases

Extract model alias mapping into a dedicated module (model-normalization.ts)
to unify model name handling across the codebase. This enables Claude Code
to work with versioned model names (e.g. claude-opus-4-6[1M]) by mapping
them to their canonical Copilot model IDs. Also fixes a bug in
count-tokens-handler where the pre-normalization model name was used for
model lookup.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 src/lib/model-normalization.ts                | 39 +++++++++++++++++++
 src/routes/messages/count-tokens-handler.ts   |  2 +-
 src/routes/messages/non-stream-translation.ts | 13 +------
 src/routes/models/route.ts                    | 33 +++++++++++-----
 .../copilot/create-chat-completions.ts        | 14 +++++--
 src/start.ts                                  | 10 +++--
 6 files changed, 83 insertions(+), 28 deletions(-)
 create mode 100644 src/lib/model-normalization.ts

diff --git a/src/lib/model-normalization.ts b/src/lib/model-normalization.ts
new file mode 100644
index 000000000..f5f8602a4
--- /dev/null
+++ b/src/lib/model-normalization.ts
@@ -0,0 +1,39 @@
+const modelAliases: Record<string, string> = {
+  "claude-opus-4-6[1M]": "claude-opus-4.6-1m",
+  "claude-opus-4-6": "claude-opus-4.6-1m",
+  "claude-sonnet-4-6": "claude-sonnet-4.6",
+}
+
+const reverseModelAliases = new Map<string, Array<string>>()
+for (const [alias, canonical] of Object.entries(modelAliases)) {
+  const aliases = reverseModelAliases.get(canonical) ?? []
+  aliases.push(alias)
+  reverseModelAliases.set(canonical, aliases)
+}
+
+export function normalizeModelName(modelId: string): string {
+  return modelAliases[modelId] ?? modelId
+}
+
+export function getModelAliases(modelId: string): Array<string> {
+  return reverseModelAliases.get(modelId) ?? []
+}
+
+export function expandModelIdsWithAliases(
+  modelIds: Array<string>,
+): Array<string> {
+  const expandedModelIds: Array<string> = []
+  const seenModelIds = new Set<string>()
+
+  for (const modelId of modelIds) {
+    for (const variant of [modelId, ...getModelAliases(modelId)]) {
+      if (seenModelIds.has(variant)) {
+        continue
+      }
+      seenModelIds.add(variant)
+      expandedModelIds.push(variant)
+    }
+  }
+
+  return expandedModelIds
+}
diff --git a/src/routes/messages/count-tokens-handler.ts b/src/routes/messages/count-tokens-handler.ts
index 2ec849cb8..939cee317 100644
--- a/src/routes/messages/count-tokens-handler.ts
+++ b/src/routes/messages/count-tokens-handler.ts
@@ -20,7 +20,7 @@ export async function handleCountTokens(c: Context) {
     const openAIPayload = translateToOpenAI(anthropicPayload)
 
     const selectedModel = state.models?.data.find(
-      (model) => model.id === anthropicPayload.model,
+      (model) => model.id === openAIPayload.model,
     )
 
     if (!selectedModel) {
diff --git a/src/routes/messages/non-stream-translation.ts b/src/routes/messages/non-stream-translation.ts
index dc41e6382..dd22038f6 100644
--- a/src/routes/messages/non-stream-translation.ts
+++ b/src/routes/messages/non-stream-translation.ts
@@ -1,3 +1,4 @@
+import { normalizeModelName } from "~/lib/model-normalization"
 import {
   type ChatCompletionResponse,
   type ChatCompletionsPayload,
@@ -30,7 +31,7 @@ export function translateToOpenAI(
   payload: AnthropicMessagesPayload,
 ): ChatCompletionsPayload {
   return {
-    model: translateModelName(payload.model),
+    model: normalizeModelName(payload.model),
     messages: translateAnthropicMessagesToOpenAI(
       payload.messages,
       payload.system,
@@ -46,16 +47,6 @@ export function translateToOpenAI(
   }
 }
 
-function translateModelName(model: string): string {
-  // Subagent requests use a specific model number which Copilot doesn't support
-  if (model.startsWith("claude-sonnet-4-")) {
-    return model.replace(/^claude-sonnet-4-.*/, "claude-sonnet-4")
-  } else if (model.startsWith("claude-opus-")) {
-    return model.replace(/^claude-opus-4-.*/, "claude-opus-4")
-  }
-  return model
-}
-
 function translateAnthropicMessagesToOpenAI(
   anthropicMessages: Array<AnthropicMessage>,
   system: string | Array<AnthropicTextBlock> | undefined,
diff --git a/src/routes/models/route.ts b/src/routes/models/route.ts
index 5254e2af7..1a3b6e42e 100644
--- a/src/routes/models/route.ts
+++ b/src/routes/models/route.ts
@@ -1,6 +1,10 @@
 import { Hono } from "hono"
 
 import { forwardError } from "~/lib/error"
+import {
+  expandModelIdsWithAliases,
+  normalizeModelName,
+} from "~/lib/model-normalization"
 import { state } from "~/lib/state"
 import { cacheModels } from "~/lib/utils"
 
@@ -13,15 +17,26 @@ modelRoutes.get("/", async (c) => {
       await cacheModels()
     }
 
-    const models = state.models?.data.map((model) => ({
-      id: model.id,
-      object: "model",
-      type: "model",
-      created: 0, // No date available from source
-      created_at: new Date(0).toISOString(), // No date available from source
-      owned_by: model.vendor,
-      display_name: model.name,
-    }))
+    const modelById = new Map(
+      state.models?.data.map((model) => [model.id, model]),
+    )
+    const modelIds = expandModelIdsWithAliases(
+      state.models?.data.map((model) => model.id) ?? [],
+    )
+    const models = modelIds.flatMap((modelId) => {
+      const sourceModel = modelById.get(normalizeModelName(modelId))
+      if (!sourceModel) return []
+
+      return {
+        id: modelId,
+        object: "model",
+        type: "model",
+        created: 0, // No date available from source
+        created_at: new Date(0).toISOString(), // No date available from source
+        owned_by: sourceModel.vendor,
+        display_name: sourceModel.name,
+      }
+    })
 
     return c.json({
       object: "list",
diff --git a/src/services/copilot/create-chat-completions.ts b/src/services/copilot/create-chat-completions.ts
index 8c915209c..e7cecf3a3 100644
--- a/src/services/copilot/create-chat-completions.ts
+++ b/src/services/copilot/create-chat-completions.ts
@@ -3,6 +3,7 @@ import { events, type ServerSentEventMessage } from "fetch-event-stream"
 
 import { copilotHeaders, copilotBaseUrl } from "~/lib/api-config"
 import { HTTPError } from "~/lib/error"
+import { normalizeModelName } from "~/lib/model-normalization"
 import { state } from "~/lib/state"
 import { refreshCopilotTokenOnError } from "~/lib/token"
 
@@ -12,13 +13,18 @@ async function doFetch(
 ) {
   if (!state.copilotToken) throw new Error("Copilot token not found")
 
-  const enableVision = payload.messages.some(
+  const normalizedPayload: ChatCompletionsPayload = {
+    ...payload,
+    model: normalizeModelName(payload.model),
+  }
+
+  const enableVision = normalizedPayload.messages.some(
     (x) =>
       typeof x.content !== "string"
       && x.content?.some((x) => x.type === "image_url"),
   )
 
-  const isAgentCall = payload.messages.some((msg) =>
+  const isAgentCall = normalizedPayload.messages.some((msg) =>
     ["assistant", "tool"].includes(msg.role),
   )
 
@@ -29,8 +35,8 @@ async function doFetch(
 
   const body =
     streamOverride !== undefined ?
-      { ...payload, stream: streamOverride }
-    : payload
+      { ...normalizedPayload, stream: streamOverride }
+    : normalizedPayload
 
   return fetch(`${copilotBaseUrl(state)}/chat/completions`, {
     method: "POST",
diff --git a/src/start.ts b/src/start.ts
index 14abbbdff..9fe1a8caf 100644
--- a/src/start.ts
+++ b/src/start.ts
@@ -6,6 +6,7 @@ import consola from "consola"
 import { serve, type ServerHandler } from "srvx"
 import invariant from "tiny-invariant"
 
+import { expandModelIdsWithAliases } from "./lib/model-normalization"
 import { ensurePaths } from "./lib/paths"
 import { initProxyFromEnv } from "./lib/proxy"
 import { generateEnvScript } from "./lib/shell"
@@ -59,9 +60,12 @@ export async function runServer(options: RunServerOptions): Promise<void> {
 
   await setupCopilotToken()
   await cacheModels()
+  const availableModelIds = expandModelIdsWithAliases(
+    state.models?.data.map((model) => model.id) ?? [],
+  )
 
   consola.info(
-    `Available models: \n${state.models?.data.map((model) => `- ${model.id}`).join("\n")}`,
+    `Available models: \n${availableModelIds.map((modelId) => `- ${modelId}`).join("\n")}`,
   )
 
   const serverUrl = `http://localhost:${options.port}`
@@ -73,7 +77,7 @@ export async function runServer(options: RunServerOptions): Promise<void> {
       "Select a model to use with Claude Code",
       {
         type: "select",
-        options: state.models.data.map((model) => model.id),
+        options: availableModelIds,
       },
     )
 
@@ -81,7 +85,7 @@ export async function runServer(options: RunServerOptions): Promise<void> {
       "Select a small model to use with Claude Code",
       {
         type: "select",
-        options: state.models.data.map((model) => model.id),
+        options: availableModelIds,
       },
     )
 

From 7b4ed3169d38d5db29b5603c368f1baa4615f320 Mon Sep 17 00:00:00 2001
From: rejectliu <rejectliu@qq.com>
Date: Sat, 28 Feb 2026 20:36:55 +0800
Subject: [PATCH 12/13] feat: add claude-haiku-4-5 as alias for
 claude-haiku-4.5

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 src/lib/model-normalization.ts | 1 +
 1 file changed, 1 insertion(+)

diff --git a/src/lib/model-normalization.ts b/src/lib/model-normalization.ts
index f5f8602a4..8a807401e 100644
--- a/src/lib/model-normalization.ts
+++ b/src/lib/model-normalization.ts
@@ -2,6 +2,7 @@ const modelAliases: Record<string, string> = {
   "claude-opus-4-6[1M]": "claude-opus-4.6-1m",
   "claude-opus-4-6": "claude-opus-4.6-1m",
   "claude-sonnet-4-6": "claude-sonnet-4.6",
+  "claude-haiku-4-5": "claude-haiku-4.5",
 }
 
 const reverseModelAliases = new Map<string, Array<string>>()

From f4b06ff9729b51951304e7a8c53935967abc48b4 Mon Sep 17 00:00:00 2001
From: rejectliu <rejectliu@qq.com>
Date: Sat, 28 Feb 2026 20:38:54 +0800
Subject: [PATCH 13/13] docs: explain why claude-opus-4-6 maps to 1m variant

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 src/lib/model-normalization.ts | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/src/lib/model-normalization.ts b/src/lib/model-normalization.ts
index 8a807401e..e3bf67d3f 100644
--- a/src/lib/model-normalization.ts
+++ b/src/lib/model-normalization.ts
@@ -1,3 +1,7 @@
+// When configuring claude-opus-4-6[1M] in Claude Code, the [1M] suffix only
+// activates the client-side 1M context window. The actual model name sent in
+// requests is still claude-opus-4-6. So to enable 1M context support, we must
+// map claude-opus-4-6 to claude-opus-4.6-1m (not claude-opus-4.6).
 const modelAliases: Record<string, string> = {
   "claude-opus-4-6[1M]": "claude-opus-4.6-1m",
   "claude-opus-4-6": "claude-opus-4.6-1m",