真实 Agent 德州扑克与在线观测交付(Issue 35) #62
Open
peng01.guo
wants to merge 2 commits from
ppanel/c5ce8c73-9e72-4520-be35-cc8453d1521f/r1-4 into deliveries/agent-poker
pull from: ppanel/c5ce8c73-9e72-4520-be35-cc8453d1521f/r1-4
merge into: peng01.guo:deliveries/agent-poker
peng01.guo:main
peng01.guo:evidence/issue-26-postrelease
peng01.guo:docs/issue-25-release-clarity
peng01.guo:build/mvp-integration
peng01.guo:evidence/issue-26-candidate-7a
peng01.guo:evidence/issue-35-control-2e920
peng01.guo:evidence/issue-21-git-helper-3b47015f
peng01.guo:evidence/issue-47-adapter-review
peng01.guo:evidence/issue-47-3b-review
peng01.guo:build/control-plane-acceptance
peng01.guo:build/backend-mvp
peng01.guo:deliveries/lifecycle-git-helper-3b47015f
peng01.guo:evidence/issue-47-slow-baseline
peng01.guo:evidence/issue-26-slow-start
peng01.guo:qa/issue-109-evidence
peng01.guo:deliveries/agent-poker-final
peng01.guo:qa/issue-108-evidence
peng01.guo:evidence/issue-26-control-final
peng01.guo:ppanel/f1aae97e-9494-432a-8575-27d80fcf1041/r1-1
peng01.guo:evidence/issue-47-007-review
peng01.guo:evidence/issue-47-d9ca-review
peng01.guo:ppanel/71c1c0f7-69c9-487e-98a2-776bd0872c4f/r1-12
peng01.guo:evidence/issue-47-0a4-review
peng01.guo:evidence/issue-99-qa-802873d1
peng01.guo:evidence/issue-26-rework-final
peng01.guo:ppanel/71c1c0f7-69c9-487e-98a2-776bd0872c4f/r1-9
peng01.guo:evidence/issue-47-efd-review
peng01.guo:evidence/issue-95-qa-a0737d41
peng01.guo:evidence/qa-93/5b368b52
peng01.guo:acceptance/rework-final-01
peng01.guo:ppanel/71c1c0f7-69c9-487e-98a2-776bd0872c4f/r1-6
peng01.guo:evidence/qa-91/90fdae47
peng01.guo:evidence/issue-21-final-2e920977
peng01.guo:ppanel/6c0ff1a7-b4e3-4c8a-9072-7c21128503db/r1-4
peng01.guo:evidence/qa-86/35cf58e3
peng01.guo:evidence/issue-85-cr-r1-20260912
peng01.guo:ppanel/6c0ff1a7-b4e3-4c8a-9072-7c21128503db/r1-1
peng01.guo:ppanel/qa-evidence-issue-77
peng01.guo:deliveries/lifecycle-final-2e920977
peng01.guo:evidence/issue-47-ff816-review
peng01.guo:evidence/issue-26-verified
peng01.guo:evidence/issue-35-upgrade
peng01.guo:evidence/issue-47-f77-review
peng01.guo:ppanel/71c1c0f7-69c9-487e-98a2-776bd0872c4f/r1-1
peng01.guo:acceptance/verified-03
peng01.guo:ppanel/8f7c7a07-b640-4cd7-b466-b3eb23cefcc6/r1-4
peng01.guo:evidence/issue-67-qa-f3f3c1be
peng01.guo:ppanel/8f7c7a07-b640-4cd7-b466-b3eb23cefcc6/r1-1
peng01.guo:acceptance/verified-02
peng01.guo:ppanel/75eb5a7a-a8ac-4175-845e-71ccb7aaf787/r1-1
peng01.guo:ppanel/c5ce8c73-9e72-4520-be35-cc8453d1521f/r1-3
peng01.guo:acceptance/verified-01
peng01.guo:ppanel/e008c2f9-5283-4f10-b30c-edd757721af3/r1-1
peng01.guo:evidence/issue-47-c19-review
peng01.guo:evidence/issue-47-independent-review
peng01.guo:evidence/issue-35-live
peng01.guo:build/ui-review-evidence-18-r7-20260912
peng01.guo:evidence/issue-26-final
peng01.guo:build/frontend-mvp
peng01.guo:ppanel/c5ce8c73-9e72-4520-be35-cc8453d1521f/r1-1
peng01.guo:ppanel/3f1c4dcd-3685-4ebb-83ba-04e16866f5ef/r1-1
peng01.guo:acceptance/final-01
peng01.guo:acceptance/final-02
peng01.guo:acceptance/final-03
peng01.guo:deliveries/agent-poker
peng01.guo:build/ui-review-evidence-18-r6-20260912
peng01.guo:build/ui-review-evidence-18-r5-20260912
peng01.guo:evidence/issue-21-final-905
peng01.guo:evidence/issue-26-gitea-render
peng01.guo:build/ui-review-evidence-18-r4-20260912
peng01.guo:deliveries/lifecycle-final-905
peng01.guo:evidence/issue-26-live
peng01.guo:evidence/issue-21-final-30dc
peng01.guo:acceptance/live-01
peng01.guo:evidence/issue-33-qa-r1-20260912
peng01.guo:ppanel/4a601137-d95b-43fb-84a3-cd8e44a9304b/r1-1
peng01.guo:build/fixture-review-evidence-24-20260912
peng01.guo:acceptance/live-02
peng01.guo:acceptance/live-03
peng01.guo:deliveries/lifecycle-final-30dc
peng01.guo:evidence/issue-26-preparation
peng01.guo:build/ui-review-evidence-18-r3-20260912
peng01.guo:build/ui-review-evidence-18-r2-20260912
peng01.guo:evidence/issue-21-lifecycle-qa
peng01.guo:build/acceptance-fixture
peng01.guo:build/ui-review-evidence-18-20260912
peng01.guo:deliveries/backend-smoke-20260912
peng01.guo:ppanel/5195c461-2402-4237-9151-f26de5abc9f8/r1-1
peng01.guo:deliveries/backend-cancel-20260912
peng01.guo:deliveries/lifecycle-qa
peng01.guo:probe/e2e-base-1789145412
peng01.guo:probe/e2e-fix-1789145412
peng01.guo:probe/connectivity-base
peng01.guo:probe/connectivity-change
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
No description provided.
Delete Branch "ppanel/c5ce8c73-9e72-4520-be35-cc8453d1521f/r1-4"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
开发返工报告:Issue #57
结论:
changes_requested。已修复快照原子写入并补充真实对局证据,但指定公网源站身份已变化,不能确认本轮候选部署到公网;未宣布最终验收通过。修改
projects/agent-poker/persistence.mjs:快照与管理凭据采用同目录临时文件、文件刷盘、原子替换、目录刷盘。进程中断不会将原快照截断;重启保留历史并默认暂停。projects/agent-poker/persistence.test.mjs:替换前强杀真实子进程,验证旧历史完整、后续替换成功及权限0600。projects/agent-poker/real-check.mjs:通过管理和邀请接口接入三名真实玩家,使用本机 YAML 中的deepseek-v4-flash与glm-5.3,没有固定动作或模型替身。projects/agent-poker/README.md与证据文件记录启动、邀请、观测、规则测试、真实对局复现与生命周期。验证与限制
本轮完成三手真实对局,其中一手真实摊牌(第三手河牌弃牌结束,不计摊牌);17 次成功模型响应,16 次应用为合法动作,另一次因页面暂停未应用。两次超时失败另有错误事件。三手筹码均为
3000;真实服务重启后三手历史完整且默认暂停。规则与持久化测试
6/6通过,构建通过;独立 HTTP 测试验证鉴权、隐私、非法行动与并发重复请求。首次 HTTP 测试与真实模型共用串行服务,固定等待不足而失败,保留原始
api.log;在独立18460服务复验通过。重启测试验证状态一致。第三手河牌阶段 GLM 连续两次调用超时并自动暂停,通过该玩家邀请接口切换为真实 DeepSeek 继续自主决策;切换记录与错误事件保留,没有脚本动作或 mock。
浏览器实际验证本轮本地牌桌、暂停/恢复、邀请加入与底牌保密。截图和 DOM 原始证据已保留。
公网 临时观测入口 直连 HTTP 返回
200,浏览器异步加载完成后显示另一张已有牌桌,停在第 5 手并提示模型调用或动作失败;代理请求返回CONNECT 403。指定 Python PID70638已不存在,cloudflaredPID70644启动身份仍为Sat Sep 12 01:23:26 2026。依据任务的失败分支,未替换未知源站、未改动隧道或控制面,公网不能算本轮候选通过。临时域名依赖既有隧道存活,无稳定域名或可用性保证。本轮服务独立监听
127.0.0.1:18459,进程身份、私有状态及日志绝对路径见evidence/verification.json。单进程 JSON 快照不提供多实例与日志/状态跨文件事务;物理断电保证受操作系统与存储硬件约束。未修改冻结验收文件。
后续独立
reviewer与qa须绑定本报告候选 SHA 与需求版本1,由控制面安排;开发自测不是最终验收。版本与证据
固定基线:
46dca1a8eaf45e67477d2e27d8a3ad44abdd7aba。开发分支:
ppanel/c5ce8c73-9e72-4520-be35-cc8453d1521f/r1-4。目标分支:
deliveries/agent-poker。候选 SHA:
470879a305438fc958acfc6486f1f86bcd94b681。原始 HTTP 响应保留
CRLF和末尾空行,git diff --check对这些证据提示空白;源代码与文档检查通过,未改写原始响应来消除提示。Checkout
From your project repository, check out a new branch and test the changes.