agent.ts 20 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252
  1. import { randomUUID, createHash } from 'node:crypto'
  2. import { DatabaseSync } from 'node:sqlite'
  3. import { realpath,access } from 'node:fs/promises'
  4. import { resolve, sep, join } from 'node:path'
  5. import {ComputerError} from './computers/container.js'
  6. import {ComputerRuntime,ComputerGateway,type ComputerTarget} from './worker/gateway.js'
  7. import { HttpError } from '../server/errors.js'
  8. import { UpstreamClient } from '../server/upstream.js'
  9. import { UpstreamServiceSession } from '../server/localAuth.js'
  10. import { WorkspaceGateway, type GatewayTarget, type GatewayFrame } from '../server/workspaceGateway.js'
  11. import { createWorkspaceToolLease, type WorkspaceToolLease } from '../server/workspaceToolLease.js'
  12. import type { RunnerCommand, RunnerConfiguration } from '../shared/runner.js'
  13. import { LoopbackTransport, isLocalAuthorizationTarget } from '../server/loopbackAuthorization.js'
  14. const commands=new Set(['session.create','session.resume','session.close','session.usage','session.interrupt','session.cwd.set','prompt.submit','session.steer','image.attach_bytes','file.attach','approval.respond','clarify.respond'])
  15. interface Connection {cleanupOnly?:boolean;gateway:WorkspaceGateway|ComputerGateway;sessions:Map<string,string>;running:Set<string>;events:Promise<void>}
  16. interface Lease {connectionId:string;profile:string;lease:WorkspaceToolLease;controller:AbortController;session:{runtimeId:string;storedId:string}}
  17. export class RunnerAgent {
  18. readonly instance=randomUUID()
  19. private active=true
  20. private connected=false
  21. private disconnectRevision=0
  22. private disconnecting?:Promise<void>
  23. private connections=new Map<string,Connection>()
  24. private leases=new Map<string,Lease>()
  25. private closedConnections=new Map<string,number>()
  26. private closedLeases=new Map<string,number>()
  27. private inflight=new Map<string,Promise<unknown>>()
  28. private db:DatabaseSync
  29. private computers?:ComputerRuntime
  30. private target:GatewayTarget
  31. private controlTransport=new LoopbackTransport()
  32. private serverEpoch?:string
  33. private controlAbort=new AbortController()
  34. constructor(readonly config:RunnerConfiguration,home:string,readonly fetchImpl:typeof fetch=fetch,readonly onStatus:(message:string)=>void=()=>{}) {
  35. const web=new URL(config.serverURL),hermes=new URL(config.hermesURL)
  36. if(web.username||web.password||web.search||web.hash||!['http:','https:'].includes(web.protocol))throw new Error('Web 地址无效')
  37. if(web.protocol==='http:'&&!['127.0.0.1','localhost','[::1]'].includes(web.hostname)&&!config.allowInsecureLan)throw new Error('远程执行节点必须使用 HTTPS,或明确配置可信局域网 HTTP')
  38. if(hermes.username||hermes.password||hermes.search||hermes.hash||!['127.0.0.1','[::1]'].includes(hermes.hostname)||!['http:','https:'].includes(hermes.protocol))throw new Error('Runner 必须连接本机 Hermes')
  39. const client=new UpstreamClient(hermes)
  40. this.target={url:hermes,client,session:new UpstreamServiceSession(client,()=>config.hermesCredentials)}
  41. this.db=new DatabaseSync(join(home,'runner-commands.sqlite3'))
  42. this.db.exec('CREATE TABLE IF NOT EXISTS commands(id TEXT PRIMARY KEY,fingerprint TEXT NOT NULL,state TEXT NOT NULL,result TEXT,created INTEGER NOT NULL)')
  43. if(config.computers)this.computers=new ComputerRuntime(this.db,config,home)
  44. this.compactReceipts()
  45. }
  46. private async api(path:string,body?:unknown):Promise<any> {
  47. const url=new URL(`/api/runner/v1/${this.config.runnerId}/${path}`,this.config.serverURL)
  48. if(url.hostname==='localhost')url.hostname='127.0.0.1'
  49. const response=await (isLocalAuthorizationTarget(url)&&this.fetchImpl===fetch?this.controlTransport.fetch.bind(this.controlTransport):this.fetchImpl)(url,{
  50. method:body===undefined?'GET':'POST',redirect:'error',signal:AbortSignal.any([this.controlAbort.signal,AbortSignal.timeout(25000)]),
  51. headers:{Authorization:`Bearer ${this.config.token}`,'x-runner-instance':this.instance,'x-runner-protocol':'1','x-runner-features':this.computers?'computer-worker-v1,artifact-chunks-v1':'',...(this.serverEpoch?{'x-runner-epoch':this.serverEpoch}:{}),'Content-Type':'application/json',...(path==='poll'&&!this.connected?{'x-runner-reset':'1'}:{})},
  52. ...(body===undefined?{}:{body:JSON.stringify(body)}),
  53. })
  54. if(!response.ok)throw new HttpError(response.status,`执行节点连接被拒绝(HTTP ${response.status})`,response.status===403?'runner_unauthorized':'runner_request_failed')
  55. const reader=response.body?.getReader();const chunks:Uint8Array[]=[];let bytes=0
  56. if(!reader)throw new Error('执行节点响应为空')
  57. try{for(;;){const part=await reader.read();if(part.done)break;bytes+=part.value.length;if(bytes>40*1024*1024)throw new Error('执行节点响应超过大小限制');chunks.push(part.value)}}finally{await reader.cancel()}
  58. return JSON.parse(Buffer.concat(chunks).toString())
  59. }
  60. private requireProfile(profile:unknown):string {
  61. if(typeof profile!=='string'||!this.config.allowedProfiles.includes(profile))throw new HttpError(403,'Profile 未授权','runner_profile_forbidden')
  62. return profile
  63. }
  64. private async closeConnection(id:string,interrupt=true) {
  65. this.closedConnections.set(id,Date.now()+60000)
  66. const connection=this.connections.get(id)
  67. if(!connection)return
  68. this.connections.delete(id)
  69. for(const [leaseId,lease] of this.leases)if(lease.connectionId===id){this.closedLeases.set(leaseId,Date.now()+60000);lease.controller.abort();this.leases.delete(leaseId);void lease.lease.dispose()}
  70. if(interrupt){let timer:ReturnType<typeof setTimeout>|undefined;try{await Promise.race([Promise.all([...connection.running].map(session_id=>connection.gateway.rpc('session.interrupt',{session_id}).catch(()=>{}))),new Promise(resolve=>{timer=setTimeout(resolve,3000)})])}finally{clearTimeout(timer)}}
  71. await connection.gateway.close()
  72. }
  73. private disconnect():Promise<void> {
  74. if(this.disconnecting)return this.disconnecting
  75. this.disconnectRevision++;this.connected=false
  76. const leases=[...this.leases.values()];this.leases.clear()
  77. for(const value of leases)value.controller.abort()
  78. const pending=Promise.all([
  79. ...leases.map(value=>value.lease.dispose()),
  80. ...[...this.connections.keys()].map(id=>this.closeConnection(id)),
  81. ]).then(()=>{}).finally(()=>{if(this.disconnecting===pending)this.disconnecting=undefined})
  82. this.disconnecting=pending;return pending
  83. }
  84. private async execute(command:RunnerCommand):Promise<any> {
  85. const p=command.payload
  86. if(!this.active||!this.connected)throw new Error('执行节点已断开')
  87. if(command.kind==='http') {
  88. const path=String(p.path),search=new URLSearchParams(String(p.search??''))
  89. if(path==='/api/computer/capabilities'||(p.computer&&path==='/api/plugins/yaoyao-bot-bridge/capabilities')){
  90. this.requireProfile(search.get('profile')??'default')
  91. if(p.method!=='GET'||!this.computers)throw new HttpError(409,'执行节点未配置隔离电脑','computer_unavailable')
  92. await Promise.all([access(this.computers.config.python),access(this.computers.config.hermesSource),access(this.computers.script)])
  93. return {status:200,headers:{},body:Buffer.from(JSON.stringify({version:1,ready:true,native_tools:true,in_process:true,network:'none'})).toString('base64')}
  94. }
  95. if(p.computer&&path==='/api/files/download')throw new HttpError(409,'隔离电脑产物需通过产物导出接口读取','computer_artifact_required')
  96. if(p.computer&&/^\/api\/sessions\/[^/]+\/messages$/.test(path)){
  97. if(p.method!=='GET'||!this.computers)throw new HttpError(403,'电脑历史不可用','computer_unavailable')
  98. const profile=this.requireProfile(search.get('profile')??'default'),session=this.computers.session(decodeURIComponent(path.split('/')[3]!),p.computer as ComputerTarget,profile)
  99. const offset=Math.max(0,Number(search.get('offset')??0)),history=session.history
  100. return {status:200,headers:{},body:Buffer.from(JSON.stringify({messages:history.slice(Math.max(0,history.length-offset-500),history.length-offset)})).toString('base64')}
  101. }
  102. if(p.method!=='GET'||!(/^\/api\/(?:status|profiles|config)$/.test(path)||/^\/api\/sessions\/[^/]+\/messages$/.test(path)||path==='/api/plugins/yaoyao-bot-bridge/capabilities'||path==='/api/files/download'))throw new HttpError(403,'Runner 不支持此 HTTP 操作','runner_http_forbidden')
  103. if(!['/api/status','/api/profiles'].includes(path))this.requireProfile(search.get('profile')??'default')
  104. if(path==='/api/files/download') {
  105. const requested=await realpath(String(search.get('path')??''))
  106. const roots=await Promise.all(this.config.artifactRoots.map(root=>realpath(root)))
  107. if(!roots.some(root=>requested===root||requested.startsWith(root+sep)))throw new HttpError(403,'文件不在允许的产物目录','runner_artifact_forbidden')
  108. }
  109. const response=await this.target.session.request(path,{search,maxResponseBytes:8*1024*1024})
  110. return {status:response.status,headers:Object.fromEntries(response.headers.entries()),body:response.body.toString('base64')}
  111. }
  112. const connectionId=String(p.connectionId??'')
  113. if(command.kind==='gateway.open') {
  114. if(this.closedConnections.has(connectionId)||this.connections.has(connectionId)||!/^[0-9a-f-]{36}$/.test(connectionId)||this.connections.size>=64)throw new Error('执行通道无效或已达上限')
  115. let gateway:WorkspaceGateway|ComputerGateway
  116. if(p.computer){
  117. const meta=p.computer as ComputerTarget
  118. if(!this.computers||![meta.environmentId,meta.agentId,String(p.workId)].every(value=>/^[0-9a-f-]{36}$/.test(value))||!/^[a-f0-9]{64}$/.test(meta.ownerKey))throw new HttpError(409,'电脑任务配置或授权无效','computer_unavailable')
  119. gateway=new ComputerGateway(this.computers,meta,String(p.workId),async()=>{if(!this.active||!this.connected||(await this.api('check',{connectionId})).allowed!==true)throw new HttpError(403,'电脑任务授权已失效','computer_authorization_revoked')},body=>this.api('artifact',{connectionId,...body}))
  120. }else gateway=new WorkspaceGateway(this.target)
  121. const connection:Connection={cleanupOnly:p.cleanupOnly===true,gateway,sessions:new Map(),running:new Set(),events:Promise.resolve()}
  122. this.connections.set(connectionId,connection)
  123. gateway.onEvent=(frame:GatewayFrame)=>{
  124. if(frame.session_id&&!connection.sessions.has(frame.session_id))return
  125. if(frame.session_id&&['message.complete','run.completed','run.failed'].includes(frame.type))connection.running.delete(frame.session_id)
  126. if(frame.type==='session.info'&&frame.session_id)for(const lease of this.leases.values())if(lease.session.runtimeId===frame.session_id&&typeof frame.payload?.stored_session_id==='string')lease.session.storedId=frame.payload.stored_session_id
  127. connection.events=connection.events.then(()=>this.api('event',{connectionId,frame})).then(()=>{},error=>{if(error instanceof HttpError&&error.status===410)void this.closeConnection(connectionId);else void this.disconnect()})
  128. }
  129. gateway.onDisconnect=()=>{void this.api('event',{connectionId,closed:true}).catch(()=>{})}
  130. try {await gateway.connect();if(!this.active||!this.connected||this.connections.get(connectionId)!==connection)throw new Error('执行节点已断开');return {ok:true}}catch(error){this.connections.delete(connectionId);gateway.close();throw error}
  131. }
  132. if(command.kind==='gateway.close'){await this.closeConnection(connectionId);return {ok:true}}
  133. if(command.kind==='gateway.rpc') {
  134. const connection=this.connections.get(connectionId),method=String(p.method),params=p.params as Record<string,unknown>
  135. if(!connection||!commands.has(method)||!params||typeof params!=='object')throw new HttpError(403,'执行通道或命令无效','runner_command_forbidden')
  136. if(connection.cleanupOnly&&!['session.resume','session.interrupt','session.close'].includes(method))throw new HttpError(403,'清理通道不允许执行新工作','runner_cleanup_forbidden')
  137. if(method==='session.create'||method==='session.resume'){if(!connection.cleanupOnly)this.requireProfile(params.profile)}
  138. else if(!connection.sessions.has(String(params.session_id)))throw new HttpError(403,'会话不属于这个通道','runner_session_forbidden')
  139. if(method==='prompt.submit')connection.running.add(String(params.session_id))
  140. const result=await connection.gateway.rpc(method,params)
  141. if(method==='session.create'||method==='session.resume'){
  142. if(this.connections.get(connectionId)!==connection)throw new HttpError(409,'执行通道已关闭','runner_command_uncertain')
  143. connection.sessions.set(String(result.session_id),String(params.profile))
  144. }
  145. if(method==='session.interrupt')connection.running.delete(String(params.session_id))
  146. return result
  147. }
  148. const leaseId=String(p.leaseId??'')
  149. if(command.kind==='lease.create') {
  150. this.requireProfile(p.profile)
  151. if(this.closedLeases.has(leaseId)||this.leases.has(leaseId)||!/^[0-9a-f-]{36}$/.test(leaseId)||this.leases.size>=64)throw new Error('工具授权无效或已达上限')
  152. const session=p.session as {runtimeId:string;storedId:string}
  153. const source=[...this.connections].find(([,c])=>c.sessions.get(session?.runtimeId)===p.profile)
  154. if(!source)throw new HttpError(403,'工具会话不属于此节点','runner_session_forbidden')
  155. const controller=new AbortController()
  156. const virtual=source[1].gateway
  157. if(virtual instanceof ComputerGateway){
  158. virtual.installTeamLease(leaseId,p.catalog as any[],async(toolId,args,callId)=>(await this.api('tool',{leaseId,toolId,arguments:args,callId})).result)
  159. const lease={bind:async()=>{},dispose:async()=>{virtual.removeTeamLease(leaseId)}}
  160. this.leases.set(leaseId,{lease,controller,session,connectionId:source[0],profile:String(p.profile)});return {ok:true}
  161. }
  162. const lease=await createWorkspaceToolLease({target:this.target,profile:String(p.profile),workId:String(p.workId),session:()=>session,signal:controller.signal,
  163. assertActive:()=>{if(!this.connected||controller.signal.aborted)throw new Error('工具授权已失效')},
  164. catalog:()=>p.catalog as any[],call:async(toolId,args,callId)=>(await this.api('tool',{leaseId,toolId,arguments:args,callId:callId??randomUUID()})).result,
  165. onFailure:()=>{controller.abort()},
  166. })
  167. if(!this.active||!this.connected||this.closedLeases.has(leaseId)||!this.connections.has(source[0])){controller.abort();await lease.dispose();throw new Error('工具授权已停止')}
  168. this.leases.set(leaseId,{lease,controller,session,connectionId:source[0],profile:String(p.profile)});return {ok:true}
  169. }
  170. if(command.kind==='lease.bind') {
  171. const lease=this.leases.get(leaseId)
  172. if(!lease)throw new HttpError(410,'工具授权已关闭','runner_lease_expired')
  173. const update=p.session as {runtimeId?:string;storedId?:string}|undefined
  174. if(update?.runtimeId!==lease.session.runtimeId||!this.connections.get(lease.connectionId)?.sessions.has(update.runtimeId))throw new HttpError(403,'工具不能改绑其他会话','runner_session_forbidden')
  175. if(typeof update.storedId!=='string')throw new Error('存储会话编号无效')
  176. lease.session.storedId=update.storedId;await lease.lease.bind();return {ok:true}
  177. }
  178. if(command.kind==='lease.close') {
  179. this.closedLeases.set(leaseId,Date.now()+60000)
  180. const lease=this.leases.get(leaseId);this.leases.delete(leaseId)
  181. if(lease){lease.controller.abort();await lease.lease.dispose()}return {ok:true}
  182. }
  183. throw new HttpError(400,'未知执行命令','runner_command_unknown')
  184. }
  185. private compactReceipts(){
  186. // Keep fingerprints as replay tombstones; large HTTP/file replies expire.
  187. this.db.prepare("UPDATE commands SET state='retired',result=NULL WHERE state='complete' AND created<?").run(Date.now()-3600000)
  188. }
  189. private handled=0
  190. async handle(command:RunnerCommand):Promise<unknown> {
  191. if(++this.handled%256===0)this.compactReceipts()
  192. for(const map of [this.closedConnections,this.closedLeases])for(const [id,expires] of map)if(expires<Date.now())map.delete(id)
  193. const existing=this.inflight.get(command.id)
  194. const fingerprint=createHash('sha256').update(JSON.stringify(command)).digest('hex')
  195. const previous=this.db.prepare('SELECT * FROM commands WHERE id=?').get(command.id)
  196. if(previous){
  197. if(previous.fingerprint!==fingerprint)throw new HttpError(409,'命令编号已用于其他内容','idempotency_conflict')
  198. if(existing)return existing
  199. if(previous.state==='complete')return JSON.parse(String(previous.result))
  200. throw new HttpError(409,'原命令结果不确定,禁止重复执行','runner_command_uncertain')
  201. }
  202. this.db.prepare("INSERT INTO commands VALUES(?,?,'started',NULL,?)").run(command.id,fingerprint,Date.now())
  203. const pending=Promise.resolve().then(async()=>{
  204. if(command.expiresAt<=Date.now()||!(await this.api('admit',{id:command.id})).allowed||command.expiresAt<=Date.now())throw new HttpError(403,'命令授权已经失效,未执行','runner_command_not_admitted')
  205. return this.execute(command)
  206. }).then(result=>({result}),error=>({error:{message:error instanceof HttpError||error instanceof ComputerError?error.message:'执行节点操作失败',code:error instanceof HttpError||error instanceof ComputerError?error.code:'runner_error'}}))
  207. .then(reply=>{this.db.prepare("UPDATE commands SET state='complete',result=? WHERE id=?").run(JSON.stringify(reply),command.id);return reply})
  208. .finally(()=>this.inflight.delete(command.id))
  209. this.inflight.set(command.id,pending);return pending
  210. }
  211. private lastStatus=''
  212. private status(message:string){if(message!==this.lastStatus){this.lastStatus=message;this.onStatus(message)}}
  213. async run(signal:AbortSignal):Promise<void> {
  214. const deliveries=new Map<string,Promise<void>>()
  215. const abort=()=>{this.active=false;this.controlAbort.abort();void this.disconnect()};signal.addEventListener('abort',abort,{once:true})
  216. try {
  217. while(this.active&&!signal.aborted) {
  218. try {
  219. const revision=this.disconnectRevision
  220. const response=await this.api('poll')
  221. await this.disconnecting
  222. if(revision!==this.disconnectRevision)continue
  223. if(this.serverEpoch&&this.serverEpoch!==response.epoch)await this.disconnect()
  224. this.serverEpoch=response.epoch;this.connected=true;this.status('执行节点已连接')
  225. for(const command of response.commands as RunnerCommand[]){
  226. if(deliveries.has(command.id))continue
  227. const delivery=(async()=>{
  228. let reply
  229. try {reply=await this.handle(command)}catch(error){reply={error:{message:'原命令需要核对',code:error instanceof HttpError?error.code:'runner_error'}}}
  230. await this.api('result',{id:command.id,...reply as object})
  231. })().catch(()=>{}).finally(()=>deliveries.delete(command.id))
  232. deliveries.set(command.id,delivery)
  233. }
  234. }catch(error){
  235. this.status(error instanceof HttpError?error.message:'连接暂时中断,正在重连')
  236. await this.disconnect()
  237. if(error instanceof HttpError&&error.code==='runner_unauthorized')throw error
  238. if(this.active)await new Promise(resolve=>setTimeout(resolve,1000))
  239. }
  240. }
  241. }finally{
  242. signal.removeEventListener('abort',abort)
  243. try{
  244. await this.disconnect().catch(()=>{})
  245. await Promise.allSettled(deliveries.values());await Promise.allSettled(this.inflight.values())
  246. await this.computers?.pool.close()
  247. }finally{this.target.client.close();this.controlTransport.close();this.db.close();this.status('执行节点已停止')}
  248. }
  249. }
  250. }