mirror of
https://github.com/Syngnat/GoNavi.git
synced 2026-07-24 15:20:03 +08:00
b815e7b296
- 新增 MCP 参数逐项说明，覆盖未知参数、位置参数和常见运行时参数 - 对敏感参数值做脱敏展示，避免提示区泄露 token 或 api key - 将逐项说明拆分到独立 util，并接入 inspect_mcp_draft 诊断输出
502 lines
18 KiB
TypeScript
502 lines
18 KiB
TypeScript
import { describe, expect, it, vi } from 'vitest';
|
||
|
||
import type { AIToolCall, SavedConnection } from '../../types';
|
||
import { executeLocalAIToolCall } from './aiLocalToolExecutor';
|
||
|
||
const buildConnection = (): SavedConnection => ({
|
||
id: 'conn-1',
|
||
name: '主库',
|
||
config: {
|
||
type: 'mysql',
|
||
host: '127.0.0.1',
|
||
port: 3306,
|
||
user: 'root',
|
||
},
|
||
});
|
||
|
||
const buildToolCall = (name: string, args: Record<string, unknown>): AIToolCall => ({
|
||
id: `call-${name}`,
|
||
type: 'function',
|
||
function: {
|
||
name,
|
||
arguments: JSON.stringify(args),
|
||
},
|
||
});
|
||
|
||
describe('aiLocalToolExecutor AI config inspection tools', () => {
|
||
it('returns the current ai runtime snapshot so the model can inspect provider, safety, skills, and tools', async () => {
|
||
const result = await executeLocalAIToolCall({
|
||
toolCall: buildToolCall('inspect_ai_runtime', {}),
|
||
connections: [buildConnection()],
|
||
mcpTools: [{
|
||
alias: 'browser_open',
|
||
originalName: 'browser_open',
|
||
serverId: 'server-1',
|
||
serverName: 'browser',
|
||
title: '打开浏览器',
|
||
description: '打开页面',
|
||
}],
|
||
skills: [{
|
||
id: 'skill-1',
|
||
name: '结构审查',
|
||
systemPrompt: '先核对字段',
|
||
enabled: true,
|
||
scopes: ['database'],
|
||
requiredTools: ['get_columns'],
|
||
}],
|
||
dynamicModels: ['gpt-5.4', 'gpt-4.1'],
|
||
toolContextMap: new Map(),
|
||
runtime: {
|
||
getDatabases: vi.fn(),
|
||
getTables: vi.fn(),
|
||
getAIRuntimeState: vi.fn().mockResolvedValue({
|
||
activeProviderId: 'provider-1',
|
||
providers: [{
|
||
id: 'provider-1',
|
||
type: 'openai',
|
||
name: 'OpenAI 主账号',
|
||
apiKey: '',
|
||
hasSecret: true,
|
||
baseUrl: 'https://api.openai.com/v1',
|
||
model: 'gpt-5.4',
|
||
models: ['gpt-5.4', 'gpt-4.1'],
|
||
maxTokens: 32000,
|
||
temperature: 0.2,
|
||
}],
|
||
safetyLevel: 'readonly',
|
||
contextLevel: 'with_samples',
|
||
}),
|
||
},
|
||
});
|
||
|
||
expect(result.success).toBe(true);
|
||
expect(result.content).toContain('"hasActiveProvider":true');
|
||
expect(result.content).toContain('"name":"OpenAI 主账号"');
|
||
expect(result.content).toContain('"safetyLevel":"readonly"');
|
||
expect(result.content).toContain('"contextLevel":"with_samples"');
|
||
expect(result.content).toContain('"enabledSkillCount":1');
|
||
expect(result.content).toContain('"alias":"browser_open"');
|
||
expect(result.content).toContain('"builtinToolCount":');
|
||
});
|
||
|
||
it('returns the current ai safety snapshot so the model can inspect write boundaries and readonly guards', async () => {
|
||
const result = await executeLocalAIToolCall({
|
||
toolCall: buildToolCall('inspect_ai_safety', {}),
|
||
connections: [{
|
||
id: 'jvm-1',
|
||
name: 'JVM 诊断环境',
|
||
config: {
|
||
type: 'jvm',
|
||
host: '10.0.0.8',
|
||
port: 0,
|
||
user: '',
|
||
jvm: {
|
||
environment: 'uat',
|
||
readOnly: true,
|
||
diagnostic: {
|
||
transport: 'agent-bridge',
|
||
allowObserveCommands: true,
|
||
allowTraceCommands: true,
|
||
allowMutatingCommands: false,
|
||
},
|
||
},
|
||
},
|
||
}],
|
||
tabs: [{
|
||
id: 'diag-tab-1',
|
||
title: 'JVM 诊断',
|
||
type: 'jvm-diagnostic',
|
||
connectionId: 'jvm-1',
|
||
readOnly: true,
|
||
}],
|
||
activeTabId: 'diag-tab-1',
|
||
mcpTools: [],
|
||
toolContextMap: new Map(),
|
||
runtime: {
|
||
getDatabases: vi.fn(),
|
||
getTables: vi.fn(),
|
||
getAIRuntimeState: vi.fn().mockResolvedValue({
|
||
safetyLevel: 'readwrite',
|
||
}),
|
||
},
|
||
});
|
||
|
||
expect(result.success).toBe(true);
|
||
expect(result.content).toContain('"safetyLevel":"readwrite"');
|
||
expect(result.content).toContain('"allowDML":true');
|
||
expect(result.content).toContain('"allowDDL":false');
|
||
expect(result.content).toContain('"readOnly":true');
|
||
expect(result.content).toContain('"allowMutatingCommands":false');
|
||
expect(result.content).toContain('allowMutating=true');
|
||
});
|
||
|
||
it('returns the current ai provider snapshot so the model can inspect provider readiness and model selection', async () => {
|
||
const result = await executeLocalAIToolCall({
|
||
toolCall: buildToolCall('inspect_ai_providers', {}),
|
||
connections: [buildConnection()],
|
||
mcpTools: [],
|
||
dynamicModels: ['gpt-5.4', 'gpt-4.1-mini'],
|
||
toolContextMap: new Map(),
|
||
runtime: {
|
||
getDatabases: vi.fn(),
|
||
getTables: vi.fn(),
|
||
getAIRuntimeState: vi.fn().mockResolvedValue({
|
||
activeProviderId: 'provider-1',
|
||
providers: [
|
||
{
|
||
id: 'provider-1',
|
||
type: 'openai',
|
||
name: 'OpenAI 主账号',
|
||
apiKey: '',
|
||
hasSecret: true,
|
||
baseUrl: 'https://api.openai.com/v1',
|
||
model: 'gpt-5.4',
|
||
models: ['gpt-5.4', 'gpt-4.1'],
|
||
maxTokens: 32000,
|
||
temperature: 0.2,
|
||
},
|
||
{
|
||
id: 'provider-2',
|
||
type: 'custom',
|
||
name: '自建代理',
|
||
apiKey: '',
|
||
hasSecret: false,
|
||
baseUrl: '',
|
||
model: '',
|
||
models: [],
|
||
headers: {
|
||
Authorization: 'Bearer secret-token',
|
||
},
|
||
maxTokens: 16000,
|
||
temperature: 0.7,
|
||
},
|
||
],
|
||
}),
|
||
},
|
||
});
|
||
|
||
expect(result.success).toBe(true);
|
||
expect(result.content).toContain('"providerCount":2');
|
||
expect(result.content).toContain('"missingSecretCount":1');
|
||
expect(result.content).toContain('"name":"OpenAI 主账号"');
|
||
expect(result.content).toContain('"name":"自建代理"');
|
||
expect(result.content).toContain('"issues":["missing_secret","missing_base_url","missing_selected_model","missing_declared_models"]');
|
||
expect(result.content).not.toContain('secret-token');
|
||
});
|
||
|
||
it('returns the current chat readiness snapshot so the model can inspect why ai input cannot send yet', async () => {
|
||
const result = await executeLocalAIToolCall({
|
||
toolCall: buildToolCall('inspect_ai_chat_readiness', {}),
|
||
connections: [buildConnection()],
|
||
mcpTools: [],
|
||
dynamicModels: ['gpt-5.5', 'gpt-4.1-mini'],
|
||
activeContext: {
|
||
connectionId: 'conn-1',
|
||
dbName: 'demo',
|
||
},
|
||
aiContexts: {
|
||
'conn-1:demo': [{
|
||
dbName: 'demo',
|
||
tableName: 'orders',
|
||
ddl: 'CREATE TABLE orders (...)',
|
||
}],
|
||
},
|
||
toolContextMap: new Map(),
|
||
runtime: {
|
||
getDatabases: vi.fn(),
|
||
getTables: vi.fn(),
|
||
getAIRuntimeState: vi.fn().mockResolvedValue({
|
||
activeProviderId: 'provider-1',
|
||
providers: [{
|
||
id: 'provider-1',
|
||
type: 'openai',
|
||
name: 'OpenAI 主账号',
|
||
apiKey: '',
|
||
hasSecret: true,
|
||
baseUrl: 'https://api.openai.com/v1',
|
||
model: '',
|
||
models: ['gpt-5.5', 'gpt-4.1-mini'],
|
||
maxTokens: 32000,
|
||
temperature: 0.2,
|
||
}],
|
||
}),
|
||
},
|
||
});
|
||
|
||
expect(result.success).toBe(true);
|
||
expect(result.content).toContain('"status":"missing_model"');
|
||
expect(result.content).toContain('"contextAttachedCount":1');
|
||
expect(result.content).toContain('"selectableModelCount":2');
|
||
expect(result.content).toContain('OpenAI 主账号');
|
||
});
|
||
|
||
it('returns the ai tool catalog so the model can choose probes and build arguments', async () => {
|
||
const result = await executeLocalAIToolCall({
|
||
toolCall: buildToolCall('inspect_ai_tool_catalog', {
|
||
keyword: 'mcp',
|
||
limit: 8,
|
||
}),
|
||
connections: [buildConnection()],
|
||
mcpTools: [{
|
||
alias: 'github_create_issue',
|
||
originalName: 'create_issue',
|
||
serverId: 'github-server',
|
||
serverName: 'GitHub',
|
||
title: '创建 Issue',
|
||
description: 'Create a GitHub issue',
|
||
inputSchema: {
|
||
type: 'object',
|
||
required: ['owner', 'repo', 'title'],
|
||
properties: {
|
||
owner: { type: 'string', description: '仓库 owner' },
|
||
repo: { type: 'string', description: '仓库名' },
|
||
title: { type: 'string', description: 'Issue 标题' },
|
||
},
|
||
},
|
||
}],
|
||
toolContextMap: new Map(),
|
||
runtime: {
|
||
getDatabases: vi.fn(),
|
||
getTables: vi.fn(),
|
||
},
|
||
});
|
||
|
||
expect(result.success).toBe(true);
|
||
expect(result.content).toContain('"keyword":"mcp"');
|
||
expect(result.content).toContain('新增 MCP 填写指引');
|
||
expect(result.content).toContain('"name":"inspect_mcp_draft"');
|
||
expect(result.content).toContain('"name":"fullCommand"');
|
||
expect(result.content).toContain('"alias":"github_create_issue"');
|
||
expect(result.content).toContain('"requiredParameters":["owner","repo","title"]');
|
||
expect(result.content).toContain('调用带参数工具前');
|
||
});
|
||
|
||
it('returns the current mcp setup snapshot so the model can inspect configured servers and client install state', async () => {
|
||
const result = await executeLocalAIToolCall({
|
||
toolCall: buildToolCall('inspect_mcp_setup', {}),
|
||
connections: [buildConnection()],
|
||
mcpTools: [{
|
||
alias: 'browser_open',
|
||
originalName: 'browser_open',
|
||
serverId: 'server-1',
|
||
serverName: 'Browser',
|
||
title: '打开页面',
|
||
description: '打开页面',
|
||
}],
|
||
toolContextMap: new Map(),
|
||
runtime: {
|
||
getDatabases: vi.fn(),
|
||
getTables: vi.fn(),
|
||
getMCPServers: vi.fn().mockResolvedValue([
|
||
{
|
||
id: 'server-1',
|
||
name: 'Browser',
|
||
transport: 'stdio',
|
||
command: 'uvx',
|
||
args: ['mcp-server-browser'],
|
||
env: {
|
||
OPENAI_API_KEY: '***',
|
||
},
|
||
enabled: true,
|
||
timeoutSeconds: 20,
|
||
},
|
||
{
|
||
id: 'server-2',
|
||
name: 'Broken',
|
||
transport: 'stdio',
|
||
command: '',
|
||
args: [],
|
||
env: {},
|
||
enabled: true,
|
||
timeoutSeconds: 1,
|
||
},
|
||
]),
|
||
getMCPClientInstallStatuses: vi.fn().mockResolvedValue([
|
||
{
|
||
client: 'codex',
|
||
displayName: 'Codex',
|
||
installed: true,
|
||
matchesCurrent: false,
|
||
clientDetected: true,
|
||
clientCommand: 'codex',
|
||
clientPath: 'C:/Tools/codex.exe',
|
||
configPath: 'C:/Users/demo/.codex/config.toml',
|
||
command: 'gonavi-mcp-server',
|
||
args: ['stdio'],
|
||
message: '检测到旧的 GoNavi 路径',
|
||
},
|
||
]),
|
||
},
|
||
});
|
||
|
||
expect(result.success).toBe(true);
|
||
expect(result.content).toContain('"serverCount":2');
|
||
expect(result.content).toContain('"name":"Browser"');
|
||
expect(result.content).toContain('"launchCommandPreview":"uvx mcp-server-browser"');
|
||
expect(result.content).toContain('"serverConfigurationIssueCount":2');
|
||
expect(result.content).toContain('"serversWithConfigurationErrors":1');
|
||
expect(result.content).toContain('"key":"command-missing"');
|
||
expect(result.content).toContain('"displayName":"Codex"');
|
||
expect(result.content).toContain('"launchCommandPreview":"gonavi-mcp-server stdio"');
|
||
});
|
||
|
||
it('returns the builtin mcp authoring guide so the model can explain how to fill command, args, env, and templates', async () => {
|
||
const result = await executeLocalAIToolCall({
|
||
toolCall: buildToolCall('inspect_mcp_authoring_guide', {}),
|
||
connections: [buildConnection()],
|
||
mcpTools: [],
|
||
toolContextMap: new Map(),
|
||
runtime: {
|
||
getDatabases: vi.fn(),
|
||
getTables: vi.fn(),
|
||
},
|
||
});
|
||
|
||
expect(result.success).toBe(true);
|
||
expect(result.content).toContain('"supportsWholeCommandAutoSplit":true');
|
||
expect(result.content).toContain('"fullCommandPasteExample":"$env:GITHUB_TOKEN=...; uvx mcp-server-github --stdio"');
|
||
expect(result.content).toContain('"title":"启动命令"');
|
||
expect(result.content).toContain('"example":"npx / node / uvx / python / docker"');
|
||
expect(result.content).toContain('PowerShell $env:KEY=VALUE;');
|
||
expect(result.content).toContain('"title":"npx 包"');
|
||
expect(result.content).toContain('"exampleLaunchPreview":"npx -y @modelcontextprotocol/server-filesystem --stdio"');
|
||
expect(result.content).toContain('"title":"uvx 工具"');
|
||
expect(result.content).toContain('"exampleLaunchPreview":"uvx some-mcp-server"');
|
||
});
|
||
|
||
it('validates an mcp draft with the real command splitter and server validator', async () => {
|
||
const result = await executeLocalAIToolCall({
|
||
toolCall: buildToolCall('inspect_mcp_draft', {
|
||
fullCommand: '$env:GITHUB_TOKEN="ghp test"; uvx mcp-server-github --stdio',
|
||
timeoutSeconds: 45,
|
||
}),
|
||
connections: [buildConnection()],
|
||
mcpTools: [],
|
||
toolContextMap: new Map(),
|
||
runtime: {
|
||
getDatabases: vi.fn(),
|
||
getTables: vi.fn(),
|
||
},
|
||
});
|
||
|
||
expect(result.success).toBe(true);
|
||
expect(result.content).toContain('"command":"uvx"');
|
||
expect(result.content).toContain('"args":["mcp-server-github","--stdio"]');
|
||
expect(result.content).toContain('"envKeys":["GITHUB_TOKEN"]');
|
||
expect(result.content).toContain('"envHints"');
|
||
expect(result.content).toContain('"label":"GitHub Token"');
|
||
expect(result.content).toContain('"secretLikeCount":1');
|
||
expect(result.content).toContain('"launchCommandPreview":"uvx mcp-server-github --stdio"');
|
||
expect(result.content).toContain('"suggestedServerSeed"');
|
||
expect(result.content).toContain('"name":"mcp-server-github"');
|
||
expect(result.content).toContain('"env":{"GITHUB_TOKEN":"***"}');
|
||
expect(result.content).toContain('"fullCommand":"GITHUB_TOKEN=*** uvx mcp-server-github --stdio"');
|
||
expect(result.content).not.toContain('ghp test');
|
||
expect(result.content).toContain('"recommendedTemplate":{"key":"uvx"');
|
||
expect(result.content).toContain('"canSave":true');
|
||
});
|
||
|
||
it('returns MCP argument hints and redacts sensitive inline argument values', async () => {
|
||
const result = await executeLocalAIToolCall({
|
||
toolCall: buildToolCall('inspect_mcp_draft', {
|
||
fullCommand: 'uvx mcp-server-demo --stdio --api-key=sk-real-secret --directory D:\\Work',
|
||
}),
|
||
connections: [buildConnection()],
|
||
mcpTools: [],
|
||
toolContextMap: new Map(),
|
||
runtime: {
|
||
getDatabases: vi.fn(),
|
||
getTables: vi.fn(),
|
||
},
|
||
});
|
||
|
||
expect(result.success).toBe(true);
|
||
expect(result.content).toContain('"argumentHints"');
|
||
expect(result.content).toContain('"argumentDetailHints"');
|
||
expect(result.content).toContain('"businessHints"');
|
||
expect(result.content).toContain('"argument":"--api-key"');
|
||
expect(result.content).toContain('"label":"API Key"');
|
||
expect(result.content).toContain('"sensitive":true');
|
||
expect(result.content).toContain('"argument":"--directory"');
|
||
expect(result.content).toContain('"label":"授权目录"');
|
||
expect(result.content).toContain('"label":"授权目录的值"');
|
||
expect(result.content).toContain('"argsRedacted":true');
|
||
expect(result.content).toContain('"--api-key=***"');
|
||
expect(result.content).not.toContain('sk-real-secret');
|
||
});
|
||
|
||
it('returns mcp tool input schemas so the model can build arguments from discovered tool metadata', async () => {
|
||
const result = await executeLocalAIToolCall({
|
||
toolCall: buildToolCall('inspect_mcp_tool_schema', {
|
||
alias: 'github_create_issue',
|
||
}),
|
||
connections: [buildConnection()],
|
||
mcpTools: [{
|
||
alias: 'github_create_issue',
|
||
originalName: 'create_issue',
|
||
serverId: 'github-server',
|
||
serverName: 'GitHub',
|
||
title: '创建 Issue',
|
||
description: 'Create a GitHub issue',
|
||
inputSchema: {
|
||
type: 'object',
|
||
required: ['owner', 'repo', 'title'],
|
||
properties: {
|
||
owner: { type: 'string', description: '仓库 owner' },
|
||
repo: { type: 'string', description: '仓库名' },
|
||
title: { type: 'string', description: 'Issue 标题' },
|
||
state: { type: 'string', enum: ['open', 'closed'] },
|
||
},
|
||
},
|
||
}],
|
||
toolContextMap: new Map(),
|
||
runtime: {
|
||
getDatabases: vi.fn(),
|
||
getTables: vi.fn(),
|
||
},
|
||
});
|
||
|
||
expect(result.success).toBe(true);
|
||
expect(result.content).toContain('"alias":"github_create_issue"');
|
||
expect(result.content).toContain('"requiredParameters":["owner","repo","title"]');
|
||
expect(result.content).toContain('"path":"state"');
|
||
expect(result.content).toContain('"enumValues":["open","closed"]');
|
||
expect(result.content).toContain('调用 github_create_issue 前必须提供：owner, repo, title');
|
||
});
|
||
|
||
it('returns the current ai guidance snapshot so the model can inspect active prompts and enabled skills', async () => {
|
||
const result = await executeLocalAIToolCall({
|
||
toolCall: buildToolCall('inspect_ai_guidance', {}),
|
||
connections: [buildConnection()],
|
||
mcpTools: [],
|
||
toolContextMap: new Map(),
|
||
userPromptSettings: {
|
||
global: '回答前先核对上下文。',
|
||
database: '生成 SQL 时只读优先。',
|
||
jvm: '',
|
||
jvmDiagnostic: '',
|
||
},
|
||
skills: [{
|
||
id: 'skill-1',
|
||
name: '结构审查',
|
||
description: '优先核对字段',
|
||
systemPrompt: '先看字段和索引，再给结论。',
|
||
enabled: true,
|
||
scopes: ['database'],
|
||
requiredTools: ['get_columns'],
|
||
}],
|
||
runtime: {
|
||
getDatabases: vi.fn(),
|
||
getTables: vi.fn(),
|
||
},
|
||
});
|
||
|
||
expect(result.success).toBe(true);
|
||
expect(result.content).toContain('"customPromptCount":2');
|
||
expect(result.content).toContain('"scope":"global"');
|
||
expect(result.content).toContain('回答前先核对上下文');
|
||
expect(result.content).toContain('"enabledSkillCount":1');
|
||
expect(result.content).toContain('"name":"结构审查"');
|
||
});
|
||
});
|