feat: keep managed SSH prompts running on Hosts

Managed SSH prompts previously depended on Desktop-owned provider and transcript state, so accepted work could stop or become uncertain when the Desktop, network, or power disappeared. The Agent now owns accepted OpenCode prompt execution, model and tool rounds, cancellation, and the durable semantic transcript, while Desktop remains authoritative for permanent projects, conversations, tasks, and messages.

Recovered work attaches to the original conversation without replay. The project switcher also groups managed SSH projects by saved Host and shows live Agent connectivity beside each group without opening a connection merely to read status. Remote project subtitles now show only the path, and the Agent lock advances to 0.11.14 with Runtime ACP v4.

Release note: 托管 SSH 任务在桌面端、网络或电源中断后可继续在主机上运行，并能回到原会话精确恢复；项目菜单现按主机分组并显示实时连接状态。

src/shared/model-bridge-contracts.ts

@@ -59,6 +59,120 @@ const modelNameSchema = utf8StringSchema(
     'Model name cannot contain control characters'
   )
 
+const modelProfileTextSchema = (label: string) =>
+  utf8StringSchema(512, {
+    minimumBytes: 1,
+    label
+  })
+    .refine(
+      (value) => value.trim() === value,
+      `${label} must not have outer whitespace`
+    )
+    .refine(
+      (value) => !/[\0\r\n]/u.test(value),
+      `${label} cannot contain control characters`
+    )
+
+const modelProviderBaseUrlSchema = utf8StringSchema(4_096, {
+  minimumBytes: 1,
+  label: 'Model provider base URL'
+}).superRefine((value, context) => {
+  let url: URL
+  try {
+    url = new URL(value)
+  } catch {
+    context.addIssue({
+      code: 'custom',
+      message: 'Model provider base URL is invalid'
+    })
+    return
+  }
+  if (
+    !['http:', 'https:'].includes(url.protocol) ||
+    url.username !== '' ||
+    url.password !== '' ||
+    url.hash !== ''
+  ) {
+    context.addIssue({
+      code: 'custom',
+      message: 'Model provider base URL must be credential-free HTTP(S)'
+    })
+  }
+})
+
+/**
+ * A bounded prompt-scoped provider snapshot. The API key is deliberately kept
+ * out of ModelBridgePolicy because policy is safe to persist and log, while
+ * this profile must only be retained by the Agent for the active prompt.
+ */
+export const agentPromptModelProfileSchema = z
+  .object({
+    profileId: modelProfileTextSchema('Model profile identifier'),
+    modelProfileDigest: sha256DigestSchema,
+    provider: z.enum(['anthropic', 'openai']),
+    baseUrl: modelProviderBaseUrlSchema,
+    model: modelNameSchema,
+    protocol: modelBridgeModelProtocolSchema,
+    authentication: z.enum(['api-key', 'none']),
+    apiKey: utf8StringSchema(16 * 1_024, {
+      minimumBytes: 1,
+      label: 'Model provider API key'
+    }).optional(),
+    capabilities: z
+      .object({
+        imageInput: z.boolean()
+      })
+      .strict(),
+    limits: z
+      .object({
+        maximumOutputTokens: z.number().int().min(1).max(1_000_000),
+        maximumModelCalls: z.number().int().min(1).max(10_000),
+        maximumTotalOutputTokens: z.number().int().min(1).max(10_000_000),
+        requestTimeoutMilliseconds: z
+          .number()
+          .int()
+          .min(1)
+          .max(300_000)
+      })
+      .strict()
+  })
+  .strict()
+  .superRefine((profile, context) => {
+    if (
+      (profile.authentication === 'api-key') !==
+      (profile.apiKey !== undefined)
+    ) {
+      context.addIssue({
+        code: 'custom',
+        path: ['apiKey'],
+        message: 'API-key authentication requires exactly one prompt credential'
+      })
+    }
+    if (
+      profile.provider === 'anthropic' !==
+      (profile.protocol === 'anthropic-messages')
+    ) {
+      context.addIssue({
+        code: 'custom',
+        path: ['provider'],
+        message: 'Model provider does not match its wire protocol'
+      })
+    }
+    if (
+      profile.limits.maximumOutputTokens >
+      profile.limits.maximumTotalOutputTokens
+    ) {
+      context.addIssue({
+        code: 'custom',
+        path: ['limits', 'maximumOutputTokens'],
+        message: 'Per-call output limit exceeds the prompt total'
+      })
+    }
+  })
+export type AgentPromptModelProfile = z.infer<
+  typeof agentPromptModelProfileSchema
+>
+
 export const modelBridgePolicySchema = z
   .object({
     protocol: modelBridgeModelProtocolSchema,