feat: v0.7.3 成本收口 · 状态一致 · 死账清理 — Prompt Cache 根治 + SSRF DNS Pinning + 87 用例扩充全量回归

P1 修复面收口: Prompt Cache 根治(日期/记忆/附件三类易变内容出 system 入用户消息
  前置块 user-context.ts, system 跨 run 字节级稳定; Anthropic system 块数组化 +
  cache_control ephemeral 断言, DeepSeek 自动缓存前缀命中 — 多轮对话输入 token
  成本降数量级); 编辑重发/重新生成幽灵 Trace 双侧根治(DB truncateMessagesAfter
  同步过滤 metadata.traceSteps + 前端 trimTraceStepsByAnchor 镜像, 严格小于锚点
  时间戳, 同毫秒等值判废); sessions:deleteMessage 死通道全链路删除(渲染层零调用
  + message_count 漂移面); Ollama vision 能力门控全链路(MetonaModelInfo
  .supportsVision 贯穿 adapter/IPC/store/UI, model-capabilities.ts 三道判定纯函数,
  未知保守放行); 记忆固化节流(consolidation-policy 纯函数: 总开关 + 内容门控
  [回答>=200字符或存在成功工具调用] + 会话级 10 分钟频率窗口, 三 memory.* 配置键)

P2 安全纵深: SSRF DNS Pinning 关闭 rebinding 窗口(ssrf-guard 重构
  resolvePublicAddresses 单源; ssrf-dispatcher 以 undici Agent.connect.lookup
  钉死校验 IP, TLS SNI 保持原域名, 一次性 dispatcher 用后即毁; 代理激活显式
  退化为仅入口校验); web_fetch 重写手动逐跳重定向循环(每跳先校验后连接,
  替代 redirect:follow 内核跟跳的中间跳裸奔, 上限 5 跳); http_request 换用
  pinned fetch; web_search 可达性预检加固(私有 URL 零请求 + 不跟跳, 3xx 视为
  可达); Agent 浏览器 CORS 通配收紧为 Origin 回显 + Vary: Origin;
  ConfirmationHook.forgetSession 会话终态清理(会话删除/abort 联动/SubAgent
  终结三处接线, 根治 rememberedDecisions 泄漏)

P3 架构还债: agent.enableReflection 死配置全链路接线(main→shared→引擎→
  Orchestrator→设置开关, REFLECTING 状态真实可达); AgentLoopConfig.timeoutMs
  死字段删除; MemoryManager.cleanupExpired 挂入健康检查周期(expires_at 回收
  管道真实化); buildSafeEnv 收敛 utils/safe-env.ts 单源(run_command 与 MCP
  stdio 共用, 终结双实现漂移); Trace 生命周期治理(metadata 只保留最近 20 个
  run — keepRecentRuns 纯函数; JSONL 录制启动自动清理保留 200 个 + 设置页
  手动清理); SLO/健康快照可视化(app:healthSnapshot IPC + 设置页只读卡片 +
  审计链一键校验)

P4 能力演进: 会话标题 LLM 自动生成(TitleGenerator — 每会话幂等/并发重入复用
  同一 Promise/自定义标题不覆盖/失败静默回退, Sidebar 经 config:changed 实时
  刷新); MCP 自动重连(5s/15s/60s 退避最多 3 次, reconnecting 状态机,
  teardownConnection 内部拆除保留簿记 — 用户断开/开关关闭即时取消, 设置页
  显示第 N/3 次); 死循环检测 ABAB 乒乓模式(最近4轮 A→B→A→B 交替判定, 补齐
  docs 第五章"两状态反复切换"检测契约); i18n 第三阶段(ChatInput/LLMSettings/
  OnboardingWizard/MemoryViewer 主链路文案出层, zh-CN + en-US 双字典补齐)

测试: 737 → 824 用例(+87, 新增 8 个测试文件 + 扩展 3 个)。新覆盖: user-context
  分组/空值收缩/拼接契约、context-builder 字节级稳定性、Anthropic cache_control
  四态、consolidation-policy 九路判定矩阵、ssrf-dispatcher(pinned lookup/重定向
  解析/IP 校验)、forget-session 会话隔离、trace-lifecycle run 淘汰、
  trace-trim 严格小于边界、safe-env 净化矩阵、mcp-reconnect 退避状态机
  (fake timers)、title-generator 并发重入、SQLite 侧 truncate×TRACE 联动
  (Electron ABI)。测试驱动修复: GIT_*/ 注释终止块注释、重连计数被自身重试
  前置断开重置(拆 teardownConnection 保留簿记)、TitleGenerator 幂等占位与
  并发去重的检查顺序竞态(去重先于幂等)

版本: 0.7.3; README 同步(配置表新增 agent.enableReflection/memory.*/mcp.autoReconnect)

回归: typecheck 双端 0 错误; ESLint 0/0; 系统 Node 771 通过 53 跳过
  (better-sqlite3 ABI); Electron ABI 全量 824/824 零跳过

electron/harness/adapters/__tests__/anthropic-cache-control.test.ts

@@ -0,0 +1,133 @@
+/**
+ * Anthropic system cache_control 断言测试（v0.7.3 P1-1）
+ *
+ * Anthropic 缓存按"内容块前缀"命中：system 必须以块数组传递并在块上打
+ * cache_control 才可缓存。本文件锁定：
+ * C1 非空 system → 块数组 + {type:'ephemeral'}；
+ * C2 空 system → 保持空字符串（不发空块）；
+ * C3 thinking 模式下断言仍然存在（cache 与 thinking 不互斥）；
+ * C4 system 块文本为四分区完整拼接（roleDefinition/outputConstraints/
+ *    safetyGuidelines/dynamicReminders）。
+ */
+
+import { describe, it, expect, vi, afterEach } from 'vitest';
+
+vi.mock('electron-log', () => ({
+  default: { info: vi.fn(), warn: vi.fn(), error: vi.fn(), debug: vi.fn() },
+}));
+
+import { AnthropicAdapter } from '../anthropic.adapter';
+import type { MetonaRequest } from '../../types';
+
+function captureFetch(): { bodies: Array<Record<string, unknown>> } {
+  const bodies: Array<Record<string, unknown>> = [];
+  const genericBody = {
+    content: [{ type: 'text', text: 'ok' }],
+    usage: { input_tokens: 3, output_tokens: 2 },
+    stop_reason: 'end_turn',
+  };
+  const fetchMock = vi.fn(async (_url: string | URL, init?: RequestInit) => {
+    bodies.push(JSON.parse(String(init?.body ?? '{}')) as Record<string, unknown>);
+    return new Response(JSON.stringify(genericBody), {
+      status: 200,
+      headers: { 'Content-Type': 'application/json' },
+    });
+  });
+  vi.stubGlobal('fetch', fetchMock);
+  return { bodies };
+}
+
+afterEach(() => {
+  vi.unstubAllGlobals();
+});
+
+function makeAdapter(): AnthropicAdapter {
+  return new AnthropicAdapter({
+    provider: 'anthropic',
+    baseURL: 'http://a.test',
+    apiKey: 'k',
+    defaultModel: 'claude-sonnet-4-5',
+  });
+}
+
+function makeRequest(overrides?: Partial<MetonaRequest>): MetonaRequest {
+  return {
+    meta: {
+      sessionId: 's1',
+      iteration: 1,
+      requestId: 'r1',
+      timestamp: Date.now(),
+      agentVersion: 'test',
+    },
+    systemPrompt: {
+      roleDefinition: 'You are Metona.',
+      outputConstraints: 'Be concise.',
+      safetyGuidelines: 'Stay safe.',
+      dynamicReminders: '## Current Workspace\n`/ws`',
+    },
+    messages: [{ role: 'user', content: 'hi', timestamp: Date.now() }],
+    params: { maxTokens: 63_488, temperature: 0, stream: false },
+    ...overrides,
+  };
+}
+
+describe('AnthropicAdapter — system cache_control（P1-1）', () => {
+  it('C1: 非空 system → 单 text 块 + cache_control ephemeral', async () => {
+    const adapter = makeAdapter();
+    const { bodies } = captureFetch();
+    await adapter.send(makeRequest());
+
+    const system = bodies[0].system as Array<{
+      type: string;
+      text: string;
+      cache_control: { type: string };
+    }>;
+    expect(Array.isArray(system)).toBe(true);
+    expect(system).toHaveLength(1);
+    expect(system[0].type).toBe('text');
+    expect(system[0].cache_control).toEqual({ type: 'ephemeral' });
+  });
+
+  it('C2: 空 system → 保持空字符串（不发空块）', async () => {
+    const adapter = makeAdapter();
+    const { bodies } = captureFetch();
+    await adapter.send(
+      makeRequest({
+        systemPrompt: { roleDefinition: '', outputConstraints: '', safetyGuidelines: '' },
+      }),
+    );
+    expect(bodies[0].system).toBe('');
+  });
+
+  it('C3: thinking 模式下 cache_control 断言仍然存在', async () => {
+    const adapter = makeAdapter();
+    const { bodies } = captureFetch();
+    await adapter.send(
+      makeRequest({
+        params: {
+          maxTokens: 8192,
+          temperature: 0,
+          stream: false,
+          thinkingEnabled: true,
+          thinkingEffort: 'high',
+        },
+      }),
+    );
+    const system = bodies[0].system as Array<{ cache_control: { type: string } }>;
+    expect(system[0].cache_control).toEqual({ type: 'ephemeral' });
+    // thinking 与 cache 共存：thinking 块也在请求体中
+    expect(bodies[0].thinking).toMatchObject({ type: 'enabled' });
+  });
+
+  it('C4: system 块文本为四分区完整拼接', async () => {
+    const adapter = makeAdapter();
+    const { bodies } = captureFetch();
+    await adapter.send(makeRequest());
+
+    const system = bodies[0].system as Array<{ text: string }>;
+    expect(system[0].text).toContain('You are Metona.');
+    expect(system[0].text).toContain('Be concise.');
+    expect(system[0].text).toContain('Stay safe.');
+    expect(system[0].text).toContain('## Current Workspace');
+  });
+});

electron/harness/adapters/__tests__/provider-request-shapes.test.ts

@@ -109,29 +109,69 @@ describe('AnthropicAdapter — 请求体契约', () => {
             role: 'assistant',
             content: null,
             toolCalls: [
-              { id: 'tc_1', name: 'read_file', args: { path: 'a.txt' }, iteration: 1, timestamp: Date.now() },
+              {
+                id: 'tc_1',
+                name: 'read_file',
+                args: { path: 'a.txt' },
+                iteration: 1,
+                timestamp: Date.now(),
+              },
             ],
             timestamp: Date.now(),
           },
-          { role: 'tool', content: null, toolResult: { toolCallId: 'tc_1', toolName: 'read_file', result: 'data', success: true, durationMs: 1, timestamp: Date.now() }, timestamp: Date.now() },
+          {
+            role: 'tool',
+            content: null,
+            toolResult: {
+              toolCallId: 'tc_1',
+              toolName: 'read_file',
+              result: 'data',
+              success: true,
+              durationMs: 1,
+              timestamp: Date.now(),
+            },
+            timestamp: Date.now(),
+          },
           // 孤立 tool_result（前面没有对应 tool_use）应被过滤
-          { role: 'tool', content: null, toolResult: { toolCallId: 'tc_orphan', toolName: 'x', result: '', success: true, durationMs: 1, timestamp: Date.now() }, timestamp: Date.now() },
+          {
+            role: 'tool',
+            content: null,
+            toolResult: {
+              toolCallId: 'tc_orphan',
+              toolName: 'x',
+              result: '',
+              success: true,
+              durationMs: 1,
+              timestamp: Date.now(),
+            },
+            timestamp: Date.now(),
+          },
           { role: 'user', content: 'next?', timestamp: Date.now() },
         ],
       }),
     );
 
     const body = bodies[0];
-    expect(body.system).toContain('You are Metona.');
+    // v0.7.3 P1-1: system 转为块数组并打 cache_control 断言（稳定前缀 prompt cache）
+    const system = body.system as Array<{
+      type: string;
+      text: string;
+      cache_control: { type: string };
+    }>;
+    expect(Array.isArray(system)).toBe(true);
+    expect(system[0].text).toContain('You are Metona.');
+    expect(system[0].cache_control).toEqual({ type: 'ephemeral' });
     expect(Array.isArray(body.messages)).toBe(true);
     const msgs = body.messages as Array<{ role: string; content: Array<Record<string, unknown>> }>;
     // tool_use 的 assistant 消息存在且携带 id/name
     const assistantToolMsg = msgs.find((m) => m.role === 'assistant');
-    expect(assistantToolMsg?.content[0]).toMatchObject({ type: 'tool_use', id: 'tc_1', name: 'read_file' });
+    expect(assistantToolMsg?.content[0]).toMatchObject({
+      type: 'tool_use',
+      id: 'tc_1',
+      name: 'read_file',
+    });
     // tool 结果以 user 角色 tool_result 形态出现且配对 id 正确；孤立者被丢弃
-    const toolResultBlocks = msgs.flatMap((m) =>
-      m.content.filter((c) => c.type === 'tool_result'),
-    );
+    const toolResultBlocks = msgs.flatMap((m) => m.content.filter((c) => c.type === 'tool_result'));
     expect(toolResultBlocks).toHaveLength(1);
     expect(toolResultBlocks[0].tool_use_id).toBe('tc_1');
   });
@@ -166,7 +206,15 @@ describe('AnthropicAdapter — 请求体契约', () => {
     });
     const { bodies } = captureFetch();
     await adapter.send(
-      makeRequest({ params: { maxTokens: 1500, temperature: 0, stream: false, thinkingEnabled: true, thinkingEffort: 'low' } }),
+      makeRequest({
+        params: {
+          maxTokens: 1500,
+          temperature: 0,
+          stream: false,
+          thinkingEnabled: true,
+          thinkingEffort: 'low',
+        },
+      }),
     );
     const body = bodies[0];
     const thinking = body.thinking as { type: string; budget_tokens: number };
@@ -185,13 +233,17 @@ describe('AnthropicAdapter — 请求体契约', () => {
     });
     const { bodies } = captureFetch();
     await adapter.send(
-      makeRequest({ params: { maxTokens: 4096, temperature: 0.7, stream: false, thinkingEnabled: true } }),
+      makeRequest({
+        params: { maxTokens: 4096, temperature: 0.7, stream: false, thinkingEnabled: true },
+      }),
     );
     expect(bodies[0].temperature).toBeUndefined();
     expect(bodies[0].thinking).toBeDefined();
 
     await adapter.send(
-      makeRequest({ params: { maxTokens: 4096, temperature: 0.7, stream: false, thinkingEnabled: false } }),
+      makeRequest({
+        params: { maxTokens: 4096, temperature: 0.7, stream: false, thinkingEnabled: false },
+      }),
     );
     expect(bodies[1].temperature).toBe(0.7);
     expect(bodies[1].thinking).toBeUndefined();
@@ -236,13 +288,37 @@ describe('OllamaAdapter — 请求体契约', () => {
     const adapter = makeOllama();
     const { bodies } = captureFetch();
 
-    await adapter.send(makeRequest({ params: { maxTokens: 4096, temperature: 0, stream: false, thinkingEnabled: true, thinkingEffort: 'low' } }));
+    await adapter.send(
+      makeRequest({
+        params: {
+          maxTokens: 4096,
+          temperature: 0,
+          stream: false,
+          thinkingEnabled: true,
+          thinkingEffort: 'low',
+        },
+      }),
+    );
     expect(bodies[0].think).toBe('low');
 
-    await adapter.send(makeRequest({ params: { maxTokens: 4096, temperature: 0, stream: false, thinkingEnabled: true, thinkingEffort: 'max' } }));
+    await adapter.send(
+      makeRequest({
+        params: {
+          maxTokens: 4096,
+          temperature: 0,
+          stream: false,
+          thinkingEnabled: true,
+          thinkingEffort: 'max',
+        },
+      }),
+    );
     expect(bodies[1].think).toBe(true);
 
-    await adapter.send(makeRequest({ params: { maxTokens: 4096, temperature: 0, stream: false, thinkingEnabled: false } }));
+    await adapter.send(
+      makeRequest({
+        params: { maxTokens: 4096, temperature: 0, stream: false, thinkingEnabled: false },
+      }),
+    );
     expect(bodies[2].think).toBeUndefined();
   });
 
@@ -319,12 +395,26 @@ describe('AgnesAdapter — 思考模式对称性（v0.6.4）', () => {
     });
     const { bodies } = captureFetch();
 
-    await adapter.send(makeRequest({ params: { maxTokens: 4096, temperature: 0, stream: false, thinkingEnabled: true, thinkingEffort: 'high' } }));
+    await adapter.send(
+      makeRequest({
+        params: {
+          maxTokens: 4096,
+          temperature: 0,
+          stream: false,
+          thinkingEnabled: true,
+          thinkingEffort: 'high',
+        },
+      }),
+    );
     expect(
       ((bodies[0].chat_template_kwargs as Record<string, unknown>) ?? {}).enable_thinking,
     ).toBe(true);
 
-    await adapter.send(makeRequest({ params: { maxTokens: 4096, temperature: 0, stream: false, thinkingEnabled: false } }));
+    await adapter.send(
+      makeRequest({
+        params: { maxTokens: 4096, temperature: 0, stream: false, thinkingEnabled: false },
+      }),
+    );
     expect(
       ((bodies[1].chat_template_kwargs as Record<string, unknown>) ?? {}).enable_thinking,
     ).toBe(false);

electron/harness/adapters/anthropic.adapter.ts

@@ -248,8 +248,7 @@ export class AnthropicAdapter extends BaseAdapter {
               usage: {
                 inputTokens: messageStartInputTokens,
                 outputTokens: (usage.output_tokens as number) ?? 0,
-                totalTokens:
-                  messageStartInputTokens + ((usage.output_tokens as number) ?? 0),
+                totalTokens: messageStartInputTokens + ((usage.output_tokens as number) ?? 0),
                 // v0.6.4: 补采 Anthropic 自己的缓存字段（其他 provider 均已采集，
                 // cache_read/creation_input_tokens 与 output_tokens 同在 usage 内）
                 cacheHitTokens: (usage.cache_read_input_tokens as number) ?? undefined,
@@ -343,10 +342,7 @@ export class AnthropicAdapter extends BaseAdapter {
           try {
             args = block.argsBuffer ? JSON.parse(block.argsBuffer) : {};
           } catch (err) {
-            args = truncatedArgumentsPayload(
-              (err as Error).message,
-              block.argsBuffer.slice(-120),
-            );
+            args = truncatedArgumentsPayload((err as Error).message, block.argsBuffer.slice(-120));
           }
           yield {
             type: MetonaStreamEventType.TOOL_CALL_COMPLETE,
@@ -506,11 +502,22 @@ export class AnthropicAdapter extends BaseAdapter {
     const body: Record<string, unknown> = {
       model: this.config.defaultModel,
       max_tokens: maxTokensForRequest,
-      system,
       messages: merged,
       stream,
     };
 
+    // v0.7.3 P1-1: system 稳定前缀打 prompt cache 断言。
+    // Anthropic 缓存按"内容块前缀"命中 —— system 以字符串传递时无法附加
+    // cache_control，必须转为块数组并在最后一个块上打 {type:'ephemeral'}。
+    // 缓存前缀覆盖 tools + system（请求组装顺序 tools 在前）：system 稳定后，
+    // 多轮对话/多轮迭代复用同一前缀，输入 token 计费按缓存读价（约 1/10）。
+    // 前缀稳定性由 P1-1 保证：易变内容（日期/记忆/附件提示）已迁入用户消息。
+    if (system) {
+      body.system = [{ type: 'text', text: system, cache_control: { type: 'ephemeral' } }];
+    } else {
+      body.system = system;
+    }
+
     // 工具定义（input_schema 命名）
     if (request.tools?.length) {
       body.tools = request.tools.map((t) => ({

electron/harness/adapters/ollama.adapter.ts

@@ -59,17 +59,21 @@ export class OllamaAdapter extends BaseAdapter {
     const nativeRequest = await this.toNativeRequest(request);
 
     // #24 修复: 使用 fetchWithTimeout 替代 getFetchSignal + fetch，确保 timer 清理
-    const response = await this.fetchWithTimeout(`${this.baseURL}/api/chat`, {
-      method: 'POST',
-      headers: { 'Content-Type': 'application/json' },
-      body: JSON.stringify({ ...nativeRequest, stream: false }),
-    }, this.config.timeoutMs ?? 300_000);
+    const response = await this.fetchWithTimeout(
+      `${this.baseURL}/api/chat`,
+      {
+        method: 'POST',
+        headers: { 'Content-Type': 'application/json' },
+        body: JSON.stringify({ ...nativeRequest, stream: false }),
+      },
+      this.config.timeoutMs ?? 300_000,
+    );
 
     if (!response.ok) {
       await this.throwHttpError(response, 'Ollama API error');
     }
 
-    const data = await response.json() as Record<string, unknown>;
+    const data = (await response.json()) as Record<string, unknown>;
     return this.toMetonaResponse(data, request.meta.requestId, request.meta.iteration);
   }
 
@@ -79,11 +83,15 @@ export class OllamaAdapter extends BaseAdapter {
     const nativeRequest = await this.toNativeRequest(request);
 
     // #24 修复: 使用 fetchWithTimeout 替代 getFetchSignal + fetch，确保 timer 清理
-    const response = await this.fetchWithTimeout(`${this.baseURL}/api/chat`, {
-      method: 'POST',
-      headers: { 'Content-Type': 'application/json' },
-      body: JSON.stringify({ ...nativeRequest, stream: true }),
-    }, this.config.timeoutMs ?? 300_000);
+    const response = await this.fetchWithTimeout(
+      `${this.baseURL}/api/chat`,
+      {
+        method: 'POST',
+        headers: { 'Content-Type': 'application/json' },
+        body: JSON.stringify({ ...nativeRequest, stream: true }),
+      },
+      this.config.timeoutMs ?? 300_000,
+    );
 
     if (!response.ok || !response.body) {
       await this.throwHttpError(response, 'Ollama stream error');
@@ -209,7 +217,10 @@ export class OllamaAdapter extends BaseAdapter {
           }
         } catch (parseErr) {
           // P2-8 修复: 与 sse-stream.ts 一致，记录解析失败行便于诊断
-          log.warn(`[Ollama] Failed to parse NDJSON line: ${(parseErr as Error).message}`, trimmed.slice(0, 200));
+          log.warn(
+            `[Ollama] Failed to parse NDJSON line: ${(parseErr as Error).message}`,
+            trimmed.slice(0, 200),
+          );
         }
       }
     }
@@ -239,7 +250,13 @@ export class OllamaAdapter extends BaseAdapter {
     format?: string | object;
     images?: string[];
     options?: Record<string, unknown>;
-  }): Promise<{ response: string; thinking?: string; done: boolean; totalDuration: number; evalCount: number }> {
+  }): Promise<{
+    response: string;
+    thinking?: string;
+    done: boolean;
+    totalDuration: number;
+    evalCount: number;
+  }> {
     const response = await fetch(`${this.baseURL}/api/generate`, {
       method: 'POST',
       headers: { 'Content-Type': 'application/json' },
@@ -248,9 +265,12 @@ export class OllamaAdapter extends BaseAdapter {
     });
 
     if (!response.ok) throw new Error(`Ollama generate error: ${response.status}`);
-    const data = await response.json() as {
-      response?: string; thinking?: string; done?: boolean;
-      total_duration?: number; eval_count?: number;
+    const data = (await response.json()) as {
+      response?: string;
+      thinking?: string;
+      done?: boolean;
+      total_duration?: number;
+      eval_count?: number;
     };
 
     return {
@@ -277,7 +297,7 @@ export class OllamaAdapter extends BaseAdapter {
     });
 
     if (!response.ok) throw new Error(`Ollama embed error: ${response.status}`);
-    const data = await response.json() as { embeddings?: number[][]; total_duration?: number };
+    const data = (await response.json()) as { embeddings?: number[][]; total_duration?: number };
 
     return {
       embeddings: data.embeddings ?? [],
@@ -299,7 +319,7 @@ export class OllamaAdapter extends BaseAdapter {
         signal: AbortSignal.timeout(10_000),
       });
       if (response.ok) {
-        const data = await response.json() as {
+        const data = (await response.json()) as {
           models?: Array<{
             name: string;
             size?: number;
@@ -309,6 +329,8 @@ export class OllamaAdapter extends BaseAdapter {
         if (data.models?.length) {
           // v0.6.4 P4-1: 能力标志改为逐模型 /api/show 实测探测；单个探测失败
           // 该模型回退保守 true（不可用时行为与旧实现一致，fail-open 保可用性）
+          // v0.7.3 P1-4: supportsVision 随探测结果透出（undefined = 未知 → 前端保守放行），
+          // 供上传入口拒绝不支持图片的本地语言模型
           const enriched = await Promise.all(
             data.models.map(async (m) => {
               const caps = await this.probeCapabilities(m.name);
@@ -319,6 +341,7 @@ export class OllamaAdapter extends BaseAdapter {
                 contextWindow: OllamaAdapter.DEFAULT_CONTEXT_WINDOW,
                 supportsToolCalling: caps ? caps.supportsTools : true,
                 supportsThinking: caps ? caps.supportsThinking : true,
+                supportsVision: caps ? caps.supportsVision : undefined,
                 description: m.details
                   ? `${m.details.family ?? 'unknown'} / ${m.details.parameter_size ?? '?'} / ${m.details.quantization_level ?? '?'}`
                   : undefined,
@@ -403,7 +426,9 @@ export class OllamaAdapter extends BaseAdapter {
 
   // ===== POST /api/show =====
 
-  async showModel(model: string): Promise<{ parameters: string; template: string; capabilities: string[] } | null> {
+  async showModel(
+    model: string,
+  ): Promise<{ parameters: string; template: string; capabilities: string[] } | null> {
     try {
       const response = await fetch(`${this.baseURL}/api/show`, {
         method: 'POST',
@@ -412,7 +437,11 @@ export class OllamaAdapter extends BaseAdapter {
         signal: AbortSignal.timeout(10_000),
       });
       if (!response.ok) return null;
-      const data = await response.json() as { parameters?: string; template?: string; capabilities?: string[] };
+      const data = (await response.json()) as {
+        parameters?: string;
+        template?: string;
+        capabilities?: string[];
+      };
       return {
         parameters: data.parameters ?? '',
         template: data.template ?? '',
@@ -470,13 +499,22 @@ export class OllamaAdapter extends BaseAdapter {
 
   // ===== GET /api/ps =====
 
-  async listRunning(): Promise<Array<{ name: string; size: number; sizeVram: number; contextLength: number }>> {
+  async listRunning(): Promise<
+    Array<{ name: string; size: number; sizeVram: number; contextLength: number }>
+  > {
     try {
       const response = await fetch(`${this.baseURL}/api/ps`, {
         signal: AbortSignal.timeout(10_000),
       });
       if (!response.ok) return [];
-      const data = await response.json() as { models?: Array<{ name: string; size?: number; size_vram?: number; context_length?: number }> };
+      const data = (await response.json()) as {
+        models?: Array<{
+          name: string;
+          size?: number;
+          size_vram?: number;
+          context_length?: number;
+        }>;
+      };
       return (data.models ?? []).map((m) => ({
         name: m.name ?? '',
         size: m.size ?? 0,
@@ -496,7 +534,7 @@ export class OllamaAdapter extends BaseAdapter {
         signal: AbortSignal.timeout(5_000),
       });
       if (!response.ok) return 'unknown';
-      const data = await response.json() as { version?: string };
+      const data = (await response.json()) as { version?: string };
       return data.version ?? 'unknown';
     } catch {
       return 'unknown';
@@ -525,7 +563,9 @@ export class OllamaAdapter extends BaseAdapter {
       const buf = Buffer.from(await res.arrayBuffer());
       return buf.toString('base64');
     } catch (error) {
-      log.warn(`[Ollama] Failed to download image ${url.slice(0, 100)}: ${(error as Error).message}`);
+      log.warn(
+        `[Ollama] Failed to download image ${url.slice(0, 100)}: ${(error as Error).message}`,
+      );
       return '';
     }
   }
@@ -539,58 +579,60 @@ export class OllamaAdapter extends BaseAdapter {
           request.systemPrompt.outputConstraints,
           request.systemPrompt.safetyGuidelines,
           request.systemPrompt.dynamicReminders,
-        ].filter(Boolean).join('\n\n'),
+        ]
+          .filter(Boolean)
+          .join('\n\n'),
       },
     ];
 
     // #2 修复: 改为 for 循环以支持 async 图片下载（map 回调无法 await）
     for (const m of request.messages) {
       if (m.role === 'system') continue;
-        // C-6 修复: Ollama API 不支持 null content，assistant 仅有 tool_calls 时转为空字符串
-        const msg: Record<string, unknown> = { role: m.role, content: m.content ?? '' };
-        // Ollama 图片使用 images 字段（纯 base64 数组，不含 data: 前缀）
-        if (m.images?.length) {
-          // #2 修复: 支持公网 URL 图片，下载后转为纯 base64
-          // 之前直接将 URL 字符串传给 Ollama，导致 base64 解码错误
-          const resolvedImages: string[] = [];
-          for (const img of m.images) {
-            const url = img.url;
-            if (url.startsWith('data:')) {
-              // data:image/png;base64,iVBOR... → iVBOR...
-              const base64Part = url.split(',')[1];
-              resolvedImages.push(base64Part ?? url);
-            } else if (url.startsWith('http://') || url.startsWith('https://')) {
-              // #2 修复: 公网 URL → 下载 → 纯 base64
-              const base64 = await this.resolveImageToBase64(url);
-              if (base64) resolvedImages.push(base64);
-            } else {
-              // 已是纯 base64 字符串（无 data: 前缀）
-              resolvedImages.push(url);
-            }
+      // C-6 修复: Ollama API 不支持 null content，assistant 仅有 tool_calls 时转为空字符串
+      const msg: Record<string, unknown> = { role: m.role, content: m.content ?? '' };
+      // Ollama 图片使用 images 字段（纯 base64 数组，不含 data: 前缀）
+      if (m.images?.length) {
+        // #2 修复: 支持公网 URL 图片，下载后转为纯 base64
+        // 之前直接将 URL 字符串传给 Ollama，导致 base64 解码错误
+        const resolvedImages: string[] = [];
+        for (const img of m.images) {
+          const url = img.url;
+          if (url.startsWith('data:')) {
+            // data:image/png;base64,iVBOR... → iVBOR...
+            const base64Part = url.split(',')[1];
+            resolvedImages.push(base64Part ?? url);
+          } else if (url.startsWith('http://') || url.startsWith('https://')) {
+            // #2 修复: 公网 URL → 下载 → 纯 base64
+            const base64 = await this.resolveImageToBase64(url);
+            if (base64) resolvedImages.push(base64);
+          } else {
+            // 已是纯 base64 字符串（无 data: 前缀）
+            resolvedImages.push(url);
           }
-          msg.images = resolvedImages;
         }
-        // 工具结果
-        if (m.role === 'tool' && m.toolResult) {
-          msg.tool_call_id = m.toolResult.toolCallId;
-          // CE-2 修复: 工具失败时 result 为 null，优先用 error 字段作为 content
-          msg.content = m.toolResult.error
-            ? m.toolResult.error
-            : (typeof m.toolResult.result === 'string'
-              ? m.toolResult.result
-              : JSON.stringify(m.toolResult.result));
-        }
-        // assistant 工具调用（Ollama REST API 要求 arguments 为 JSON 字符串）
-        if (m.role === 'assistant' && m.toolCalls?.length) {
-          msg.tool_calls = m.toolCalls.map((tc) => ({
-            function: { name: tc.name, arguments: JSON.stringify(tc.args) },
-          }));
-        }
-        // 推理内容回传（保持多轮推理链完整）
-        if (m.role === 'assistant' && m.reasoningContent) {
-          (msg as Record<string, unknown>).reasoning_content = m.reasoningContent;
-        }
-        messages.push(msg);
+        msg.images = resolvedImages;
+      }
+      // 工具结果
+      if (m.role === 'tool' && m.toolResult) {
+        msg.tool_call_id = m.toolResult.toolCallId;
+        // CE-2 修复: 工具失败时 result 为 null，优先用 error 字段作为 content
+        msg.content = m.toolResult.error
+          ? m.toolResult.error
+          : typeof m.toolResult.result === 'string'
+            ? m.toolResult.result
+            : JSON.stringify(m.toolResult.result);
+      }
+      // assistant 工具调用（Ollama REST API 要求 arguments 为 JSON 字符串）
+      if (m.role === 'assistant' && m.toolCalls?.length) {
+        msg.tool_calls = m.toolCalls.map((tc) => ({
+          function: { name: tc.name, arguments: JSON.stringify(tc.args) },
+        }));
+      }
+      // 推理内容回传（保持多轮推理链完整）
+      if (m.role === 'assistant' && m.reasoningContent) {
+        (msg as Record<string, unknown>).reasoning_content = m.reasoningContent;
+      }
+      messages.push(msg);
     }
 
     const body: Record<string, unknown> = {
@@ -619,14 +661,23 @@ export class OllamaAdapter extends BaseAdapter {
 
     // Thinking 模式
     if (request.params.thinkingEnabled) {
-      const effortMap: Record<string, string | boolean> = { low: 'low', medium: 'medium', high: 'high', max: true };
+      const effortMap: Record<string, string | boolean> = {
+        low: 'low',
+        medium: 'medium',
+        high: 'high',
+        max: true,
+      };
       body.think = effortMap[request.params.thinkingEffort ?? 'high'] ?? true;
     }
 
     return body;
   }
 
-  private toMetonaResponse(data: Record<string, unknown>, requestId: string, iteration: number = 0): MetonaResponse {
+  private toMetonaResponse(
+    data: Record<string, unknown>,
+    requestId: string,
+    iteration: number = 0,
+  ): MetonaResponse {
     const message = data.message as Record<string, unknown> | undefined;
     const toolCalls = message?.tool_calls as Array<Record<string, unknown>> | undefined;
     return {
@@ -638,11 +689,14 @@ export class OllamaAdapter extends BaseAdapter {
         timestamp: Date.now(),
         perfStats: {
           loadDurationMs: data.load_duration ? (data.load_duration as number) / 1e6 : undefined,
-          promptEvalDurationMs: data.prompt_eval_duration ? (data.prompt_eval_duration as number) / 1e6 : undefined,
-          evalDurationMs: data.eval_duration ? (data.eval_duration as number) / 1e6 : undefined,
-          tokensPerSecond: data.eval_count && data.eval_duration
-            ? ((data.eval_count as number) / ((data.eval_duration as number) / 1e9))
+          promptEvalDurationMs: data.prompt_eval_duration
+            ? (data.prompt_eval_duration as number) / 1e6
             : undefined,
+          evalDurationMs: data.eval_duration ? (data.eval_duration as number) / 1e6 : undefined,
+          tokensPerSecond:
+            data.eval_count && data.eval_duration
+              ? (data.eval_count as number) / ((data.eval_duration as number) / 1e9)
+              : undefined,
         },
       },
       content: (message?.content as string) ?? '',
@@ -652,7 +706,10 @@ export class OllamaAdapter extends BaseAdapter {
         const rawArgs = fn?.arguments;
         let args: Record<string, unknown> = {};
         try {
-          args = typeof rawArgs === 'string' ? JSON.parse(rawArgs) : (rawArgs as Record<string, unknown>) ?? {};
+          args =
+            typeof rawArgs === 'string'
+              ? JSON.parse(rawArgs)
+              : ((rawArgs as Record<string, unknown>) ?? {});
         } catch (parseErr) {
           // v0.6.4: 非流式路径截断自愈对齐 —— 原 catch 静默降级 {}，与流式修复后的
           // 行为不一致。统一转为 _truncatedArguments 错误参数。
@@ -677,7 +734,10 @@ export class OllamaAdapter extends BaseAdapter {
         outputTokens: (data.eval_count as number) ?? 0,
         totalTokens: ((data.prompt_eval_count as number) ?? 0) + ((data.eval_count as number) ?? 0),
       },
-      finishReason: mapOllamaDoneReason(data.done_reason as string | undefined, !!message?.tool_calls),
+      finishReason: mapOllamaDoneReason(
+        data.done_reason as string | undefined,
+        !!message?.tool_calls,
+      ),
     };
   }
 }
@@ -693,10 +753,15 @@ function mapOllamaDoneReason(
 ): MetonaFinishReason {
   if (hasToolCalls) return MetonaFinishReason.TOOL_CALLS;
   switch (reason) {
-    case 'stop': return MetonaFinishReason.STOP;
-    case 'length': return MetonaFinishReason.LENGTH;
-    case 'load': return MetonaFinishReason.STOP;   // 冷启动加载完成，非错误
-    case 'unload': return MetonaFinishReason.STOP;
-    default: return MetonaFinishReason.STOP;
+    case 'stop':
+      return MetonaFinishReason.STOP;
+    case 'length':
+      return MetonaFinishReason.LENGTH;
+    case 'load':
+      return MetonaFinishReason.STOP; // 冷启动加载完成，非错误
+    case 'unload':
+      return MetonaFinishReason.STOP;
+    default:
+      return MetonaFinishReason.STOP;
   }
 }