Merge czy-all into dev

Resolve PR #39 conflicts by preserving dev-side quota and routing fixes while bringing in upstream Responses/Messages updates.

Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>

Author: cuipengfei

desktop/package.json

@@ -1,6 +1,6 @@
 {
   "name": "copilot-api-desktop",
-  "version": "1.10.12",
+  "version": "1.10.22",
   "description": "Copilot API Desktop App",
   "main": "out/main/index.js",
   "scripts": {

package.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://json.schemastore.org/package.json",
   "name": "@jeffreycao/copilot-api",
-  "version": "1.10.12",
+  "version": "1.10.22",
   "description": "OpenAI and Anthropic-compatible gateway for GitHub Copilot or Codex or third-party providers.",
   "keywords": [
     "ai-gateway",

src/lib/api-config.ts

@@ -150,7 +150,7 @@ const OPENCODE_VERSION = "opencode/1.14.29"
 const OPENCODE_LLM_USER_AGENT =
   "opencode/1.14.29 ai-sdk/provider-utils/4.0.23 runtime/bun/1.3.13, opencode/1.14.29"
 
-export const COPILOT_VERSION = "0.47.1"
+export const COPILOT_VERSION = "0.50.1"
 const EDITOR_PLUGIN_VERSION = `copilot-chat/${COPILOT_VERSION}`
 const USER_AGENT = `GitHubCopilotChat/${COPILOT_VERSION}`
 const CLAUDE_AGENT_USER_AGENT =

src/lib/models.ts

@@ -2,14 +2,19 @@ import type { Model } from "~/services/copilot/get-models"
 
 import { state } from "~/lib/state"
 
+export interface NormalizedSdkModelId {
+  family: string
+  version: string
+}
+
 export const findEndpointModel = (sdkModelId: string): Model | undefined => {
   const models = state.models?.data ?? []
   const exactMatch = models.find((m) => m.id === sdkModelId)
   if (exactMatch) {
     return exactMatch
   }
 
-  const normalized = _normalizeSdkModelId(sdkModelId)
+  const normalized = normalizeSdkModelId(sdkModelId)
   if (!normalized) {
     return undefined
   }
@@ -33,30 +38,30 @@ export const findEndpointModel = (sdkModelId: string): Model | undefined => {
  * - "claude-haiku-3-5-20250514" -> { family: "haiku", version: "3.5" }
  * - "claude-haiku-4.5" -> { family: "haiku", version: "4.5" }
  */
-const _normalizeSdkModelId = (
+export const normalizeSdkModelId = (
   sdkModelId: string,
-): { family: string; version: string } | undefined => {
+): NormalizedSdkModelId | undefined => {
   const lower = sdkModelId.toLowerCase()
 
   // Strip date suffix (8 digits at the end)
   const withoutDate = lower.replace(/-\d{8}$/, "")
 
-  // Pattern 1: claude-{family}-{major}-{minor} (e.g., claude-opus-4-5, claude-haiku-3-5)
-  const pattern1 = withoutDate.match(/^claude-(\w+)-(\d+)-(\d+)$/)
+  // Pattern 1: claude-{family}-{major}.{minor} (e.g., claude-haiku-4.5)
+  const pattern1 = withoutDate.match(/^claude-(\w+)-(\d+)\.(\d+)$/)
   if (pattern1) {
     return { family: pattern1[1], version: `${pattern1[2]}.${pattern1[3]}` }
   }
 
-  // Pattern 2: claude-{major}-{minor}-{family} (e.g., claude-3-5-sonnet)
-  const pattern2 = withoutDate.match(/^claude-(\d+)-(\d+)-(\w+)$/)
+  // Pattern 2: claude-{family}-{major}-{minor} (e.g., claude-opus-4-5, claude-haiku-3-5)
+  const pattern2 = withoutDate.match(/^claude-(\w+)-(\d+)-(\d+)$/)
   if (pattern2) {
-    return { family: pattern2[3], version: `${pattern2[1]}.${pattern2[2]}` }
+    return { family: pattern2[1], version: `${pattern2[2]}.${pattern2[3]}` }
   }
 
-  // Pattern 3: claude-{family}-{major}.{minor} (e.g., claude-haiku-4.5)
-  const pattern3 = withoutDate.match(/^claude-(\w+)-(\d+)\.(\d+)$/)
+  // Pattern 3: claude-{major}-{minor}-{family} (e.g., claude-3-5-sonnet)
+  const pattern3 = withoutDate.match(/^claude-(\d+)-(\d+)-(\w+)$/)
   if (pattern3) {
-    return { family: pattern3[1], version: `${pattern3[2]}.${pattern3[3]}` }
+    return { family: pattern3[3], version: `${pattern3[1]}.${pattern3[2]}` }
   }
 
   // Pattern 4: claude-{family}-{major} (e.g., claude-sonnet-4)

src/lib/utils.ts

@@ -20,13 +20,60 @@ export const sleep = (ms: number) =>
 export const isNullish = (value: unknown): value is null | undefined =>
   value === null || value === undefined
 
-export async function cacheModels(): Promise<void> {
-  const models = await getCopilotModels()
+// Periodically refresh models so long-running daemons pick up new SKUs.
+const MODELS_REFRESH_BASE_MS = 30 * 60 * 1000
+let modelsRefreshTimer: ReturnType<typeof setTimeout> | null = null
+
+export const stopModelsRefreshLoop = () => {
+  if (modelsRefreshTimer) {
+    clearTimeout(modelsRefreshTimer)
+    modelsRefreshTimer = null
+  }
+}
 
+type ModelsFetcher = typeof getCopilotModels
+
+const refreshModels = async (fetcher: ModelsFetcher) => {
+  const prevIds = new Set(state.models?.data.map((m) => m.id) ?? [])
+  const models = await fetcher()
   state.models = {
     ...models,
     data: models.data.filter((model) => model.model_picker_enabled),
   }
+  const nextIds = state.models.data.map((m) => m.id)
+  const added = nextIds.filter((id) => !prevIds.has(id))
+  if (added.length > 0) {
+    consola.info(`Models refresh: ${added.length} new -- ${added.join(", ")}`)
+  } else {
+    consola.debug(`Models refresh: no changes (${nextIds.length} total)`)
+  }
+}
+
+const scheduleModelsRefresh = (fetcher: ModelsFetcher, intervalMs: number) => {
+  const jitter = Math.floor(Math.random() * (intervalMs / 6))
+  const delay = intervalMs + jitter
+  consola.debug(
+    `Scheduling next models refresh in ${Math.round(delay / 1000)} seconds`,
+  )
+
+  stopModelsRefreshLoop()
+  modelsRefreshTimer = setTimeout(async () => {
+    try {
+      await refreshModels(fetcher)
+    } catch (error) {
+      consola.warn("Failed to refresh models, keeping previous cache.", error)
+    } finally {
+      scheduleModelsRefresh(fetcher, intervalMs)
+    }
+  }, delay)
+}
+
+export async function cacheModels(
+  fetcher: ModelsFetcher = getCopilotModels,
+  intervalMs: number = MODELS_REFRESH_BASE_MS,
+): Promise<void> {
+  await refreshModels(fetcher)
+  scheduleModelsRefresh(fetcher, intervalMs)
 }
 
 export const cacheVSCodeVersion = async () => {

src/routes/messages/anthropic-types.ts

@@ -2,7 +2,7 @@
 
 export interface AnthropicMessagesPayload {
   model: string
-  messages: Array<AnthropicMessage>
+  messages: Array<AnthropicInputMessage>
   cache_control?: AnthropicCacheControl | null
   system?: string | Array<AnthropicTextBlock>
   stop_sequences?: Array<string>
@@ -125,7 +125,13 @@ export interface AnthropicAssistantMessage {
   content: string | Array<AnthropicAssistantContentBlock>
 }
 
+export interface AnthropicSystemMessage {
+  role: "system"
+  content: string | Array<AnthropicTextBlock>
+}
+
 export type AnthropicMessage = AnthropicUserMessage | AnthropicAssistantMessage
+export type AnthropicInputMessage = AnthropicMessage | AnthropicSystemMessage
 
 export interface AnthropicTool {
   name: string

src/routes/messages/count-tokens-handler.ts

@@ -18,6 +18,7 @@ import { type Model } from "~/services/copilot/get-models"
 import { findEndpointModel } from "../../lib/models"
 import { type AnthropicMessagesPayload } from "./anthropic-types"
 import { translateToOpenAI } from "./non-stream-translation"
+import { normalizeSystemMessages } from "./preprocess"
 
 export const resolveCountTokensModel = (
   modelId: string,
@@ -92,6 +93,7 @@ async function countTokensViaAnthropic(
 export async function handleCountTokens(c: Context) {
   const anthropicPayload = await c.req.json<AnthropicMessagesPayload>()
   anthropicPayload.model = resolveMappedModel(anthropicPayload.model)
+  normalizeSystemMessages(anthropicPayload)
 
   const providerModelAlias = parseProviderModelAlias(anthropicPayload.model)
   if (providerModelAlias) {

src/routes/messages/handler.ts

@@ -37,6 +37,7 @@ import {
   getCompactType,
   getLastMessageContentCacheControl,
   mergeToolResultForClaude,
+  normalizeSystemMessages,
   sanitizeIdeTools,
   stripToolReferenceTurnBoundary,
 } from "./preprocess"
@@ -70,10 +71,13 @@ export async function handleCompletion(c: Context) {
     })
   }
 
+  debugJson(logger, "Anthropic request payload:", anthropicPayload)
+
+  normalizeSystemMessages(anthropicPayload)
+
   await checkRateLimit(state)
 
   const originalModel = anthropicPayload.model
-  debugJson(logger, "Anthropic request payload:", anthropicPayload)
 
   const compactType = getCompactType(anthropicPayload)
   if (compactType !== 0) {

src/routes/messages/non-stream-translation.ts

@@ -16,6 +16,7 @@ import {
   type AnthropicAssistantContentBlock,
   type AnthropicAssistantMessage,
   type AnthropicDocumentBlock,
+  type AnthropicMessage,
   type AnthropicMessagesPayload,
   type AnthropicResponse,
   type AnthropicTextBlock,
@@ -124,10 +125,11 @@ function translateAnthropicMessagesToOpenAI(
   capabilities: TranslationCapabilities,
 ): Array<Message> {
   const systemMessages = handleSystemPrompt(payload.system)
-  const otherMessages = payload.messages.flatMap((message) =>
-    message.role === "user" ?
-      handleUserMessage(message, capabilities)
-    : handleAssistantMessage(message, modelId, capabilities),
+  const otherMessages = (payload.messages as Array<AnthropicMessage>).flatMap(
+    (message) =>
+      message.role === "user" ?
+        handleUserMessage(message, capabilities)
+      : handleAssistantMessage(message, modelId, capabilities),
   )
   return [...systemMessages, ...otherMessages]
 }