AI Engineering on Mengboy 技术笔记

Claude Code + GitHub Actions CI Self-Healing Pipeline: Error Attribution, Minimal Patches, and Human Approval Gates

Mon, 06 Apr 2026 01:17:34 +0000

If your CI keeps failing and engineers keep babysitting logs, you’re paying an invisible velocity tax. A production-grade AI self-healing pipeline is not “let the agent edit anything”. It’s a controlled loop: attribution, patching, approval, rollback.

This post gives you a deployable baseline: Claude Code proposes a minimal fix patch, GitHub Actions enforces risk gates and regression checks, and humans only approve at high-impact checkpoints.

Claude API Rate-Limit Storm Playbook: Adaptive Concurrency, Jittered Backoff, and Quota Isolation

Fri, 03 Apr 2026 01:15:05 +0000

When Claude API starts returning 429 under high load, most systems don’t just slow down—they collapse: queue buildup, retry storms, upstream timeout chains, and pager noise.

Claude API 高并发限流雪崩应对：自适应并发、退避抖动与配额隔离

Fri, 03 Apr 2026 01:15:05 +0000

当 Claude API 在高并发下开始返回 429，很多系统不是“慢一点”，而是直接雪崩：队列堆积、重试风暴、上游超时、下游告警连锁。

OpenAI Responses Streaming in Production: Backpressure, Chunk Reassembly, and Timeout Budget

Fri, 27 Mar 2026 01:08:00 +0000

Most streaming failures are not about “can it stream”, but “does it stay stable under load”: broken chunks, stuck clients, timeout cascades, and retry storms.

OpenAI Responses 流式输出生产稳态：背压控制、分片重组与超时预算闭环

Fri, 27 Mar 2026 01:08:00 +0000

线上最容易把流式输出做坏的，不是“能不能流出来”，而是流量一上来就抖：token 断片、客户端卡死、超时雪崩、重试风暴。

Claude + OpenAI Model Routing Gateway: Latency Tiers, Cost Caps, and Quality Guardrails

Wed, 25 Mar 2026 01:16:31 +0000

Connecting both Claude and OpenAI in production is the easy part. The hard part is keeping the system stable across the quality-latency-cost triangle.
Without a routing gateway, you usually get latency spikes, runaway bills, and ugly cascading failures.

Claude + OpenAI 模型路由网关实战：延迟分层、成本阈值与质量守门

Wed, 25 Mar 2026 01:16:31 +0000

你把 Claude 和 OpenAI 一起接进生产环境后，真正的难题不是“能不能调通”，而是怎么在质量、延迟、成本三角里稳定跑。
如果没有路由网关，最常见结果就是：高峰期延迟抖动、账单失控、异常时全站雪崩。

OpenAI Responses + Go Stream Recovery: Delta Persistence, Resume Tokens, and Duplicate Chunk Dedup

Mon, 23 Mar 2026 01:13:09 +0000

In production, the painful part is not “streaming is slow.” It’s “streaming breaks and then duplicates output after reconnect.”
This guide gives you a practical recovery loop: delta persistence + resume token + idempotent dedup, so reconnection does not replay garbage.

OpenAI Responses in Go Multi-Tenant Quota Governance: Token Buckets, Budget Circuit Breakers, and Cost Attribution

Fri, 20 Mar 2026 01:08:00 +0000

Most multi-tenant AI platforms fail for two boring reasons: one tenant saturates shared capacity, and finance discovers the burn too late.

This guide gives you a practical Go blueprint: token-bucket throttling, budget circuit breakers, and request-level cost attribution.

OpenAI Responses 在 Go 多租户中的配额治理：令牌桶限流、预算熔断与账单归因

Fri, 20 Mar 2026 01:08:00 +0000

多租户 AI 服务最容易死在两件事：一个租户打爆全局配额，以及月底账单炸了才发现。

这篇给你一套可直接落地的 Go 方案：令牌桶限流 + 预算熔断 + 账单归因，目标是“先活下来，再精细化”。

Go + OpenAI Responses Agent Memory Layering: Short-Term Context, Long-Term Index, and Cost Caps

Wed, 18 Mar 2026 16:33:52 +0000

In production Go agents, the first thing that breaks is usually not model quality. It is memory management: context grows, bills spike, and answers drift.

Use a 3-layer memory design:

L1: short-term conversational window (seconds)
L2: rolling summary (minutes)
L3: long-term retrieval memory (days)

Go + OpenAI Responses Agent 记忆分层实战：短期上下文、长期索引与成本封顶

Wed, 18 Mar 2026 16:33:52 +0000

你在 Go 里做 Agent，最容易翻车的不是推理能力，而是“记忆”失控：上下文越来越长、账单越来越高、回答却越来越飘。

这篇给你一个可落地的三层方案：

L1：短期会话上下文（秒级，强相关）
L2：中期摘要记忆（分钟级，压缩）
L3：长期检索记忆（天级，向量索引）

Go 服务调用 OpenAI 的 429/5xx 风暴应对：令牌桶、指数退避与熔断恢复

Wed, 18 Mar 2026 01:14:00 +0000

你不是被 OpenAI API「偶尔报错」打败的；你是被并发放大后的重试风暴打败的。

Handling OpenAI 429/5xx Storms in Go: Token Bucket, Exponential Backoff, and Circuit Breakers

Wed, 18 Mar 2026 01:14:00 +0000

Most Go teams are not killed by a single API error. They are killed by a retry storm they created themselves.

OpenAI Responses + GitHub Actions PR Risk Gate: Automated Evals, Tiered Blocking, and One-Click Rollback

Mon, 16 Mar 2026 01:08:00 +0000

You don’t need an AI reviewer that “sounds smart.” You need a gate that stops risky PRs before they hit main.

This post shows a production-ready minimum setup: OpenAI Responses generates structured risk output, GitHub Actions enforces tiered policies, and critical failures can trigger a one-click rollback.

OpenAI Responses + GitHub Actions 的 PR 风险闸门：自动评测、分级阻断与一键回滚

Mon, 16 Mar 2026 01:08:00 +0000

你不需要一个“会聊天”的 AI 审查器，你需要一个能阻断坏改动进主干的风险闸门。

这篇给一套可上线的最小方案：OpenAI Responses 负责生成结构化审查结论，GitHub Actions 负责分级阻断，发现高风险时自动回滚到安全提交。

OpenAI Batch API + Go 降本实战：离线拆批、失败重放与成本边界

Fri, 13 Mar 2026 01:08:00 +0000

一句话结论：如果你的调用是可延迟、可批处理、可回放，就该把在线请求下沉到 Batch API；省钱最明显，但前提是你把拆批、失败分流和回放链路先做好。

很多团队把 Batch API 当“便宜版同步接口”来用，结果不是省钱，而是把失败样本堆成事故池。真正的保守做法是：先定义成本边界和SLO，再做离线拆批与失败回放。

OpenAI Batch API with Go: Offline Batching, Failure Replay, and Cost Boundaries

Fri, 13 Mar 2026 01:08:00 +0000

Short answer: if your workload is delay-tolerant, batchable, and replay-safe, move it from online calls to Batch API. The savings are real, but only if you design splitting, failure routing, and replay first.

Many teams treat Batch API as a cheaper sync endpoint. That usually creates a replay mess instead of stable savings. A conservative rollout starts with cost boundaries and SLOs, then implements offline batching and controlled replay.

OpenAI Responses Structured Outputs + Go：Schema 演进、坏样本兜底与灰度回滚

Wed, 11 Mar 2026 01:08:00 +0000

Structured Outputs 最容易翻车的地方，不是“模型不听话”，而是你把 schema 当成了永远不变的圣旨。

线上一旦进入版本演进期，最常见的事故就是：字段新增后老消费端崩、枚举值扩展后校验误杀、坏样本把整条链路拖死，最后只能半夜回滚，像在给自己写惊悚片。

OpenAI Responses Structured Outputs with Go: Schema Evolution, Bad-Case Fallbacks, and Gradual Rollback

Wed, 11 Mar 2026 01:08:00 +0000

The hardest part of Structured Outputs is not getting JSON once. It is surviving schema changes without turning production into a small fire with excellent logs and terrible business results.

Once a Go service starts evolving prompts and response contracts, the usual failure modes show up fast: a new required field breaks older consumers, an enum expands and strict validation kills valid requests, or one bad sample drags the whole chain into retries and rollback panic.

OpenAI Realtime + Go in Production: WebRTC Token Rotation, Interruption Recovery, and End-to-End Latency Budgets

Mon, 09 Mar 2026 01:13:00 +0000

If you plan to put OpenAI Realtime into production, do not let a passing demo fool you.

What usually breaks the system is not the model itself. It is non-rotating short-lived auth, missing interruption state, and zero end-to-end latency budgeting. Miss those three and your voice UX starts sounding like an angry walkie-talkie.

OpenAI Realtime + Go 生产落地：WebRTC 鉴权轮换、打断恢复与端到端延迟预算

Mon, 09 Mar 2026 01:13:00 +0000

如果你准备把 OpenAI Realtime 真上生产，先别被“能跑通 demo”骗了。

真正把系统打爆的，通常不是模型本身，而是 短时鉴权没轮换、打断恢复没状态机、端到端延迟没预算。这三件事不补，语音体验会像在和一台卡顿的对讲机吵架。

Go + OpenAI Responses: Connection Pooling and Timeout Budgets from HTTP/2 Reuse to Error-Budget Control

Fri, 06 Mar 2026 01:13:12 +0000

When Go services call the OpenAI Responses API in production, the real failures are rarely about model quality. Most incidents come from transport instability: weak connection pooling, conflicting timeout layers, and retry storms.

This guide gives you a practical baseline: HTTP/2 reuse, layered timeout budgets, bounded retries, and error-budget driven operations.

Go 调 OpenAI Responses 的连接池与超时预算：HTTP/2 复用到错误预算闭环

Fri, 06 Mar 2026 01:13:12 +0000

线上 Go 服务调用 OpenAI Responses 时，最容易踩的坑不是“模型不准”，而是链路抖动：连接池不稳、超时预算乱配、重试叠加把自己打挂。

这篇给一套可落地的基线配置：HTTP/2 连接复用、分层超时、错误预算和退避重试，目标是把 5xx 与超时比例压到可控范围，并且能快速定位瓶颈。

OpenAI Responses + Go 工具调用重试风暴治理：幂等键、退避抖动与熔断阈值

Wed, 04 Mar 2026 01:10:40 +0000

线上最可怕的不是一次失败，而是失败后被重试放大。

在 OpenAI Responses + Go 的工具调用链路里，如果没有幂等键、退避抖动和熔断阈值，10 个请求很快就能打成 1000 个下游调用，账单和延迟一起爆炸。

OpenAI Responses + Go: Taming Retry Storms with Idempotency Keys, Jittered Backoff, and Circuit Breakers

Wed, 04 Mar 2026 01:10:40 +0000

The most expensive outage is not a single failure — it is a failure amplified by retries.

In an OpenAI Responses + Go tool-calling stack, missing idempotency, jittered backoff, and breaker thresholds can turn 10 failing requests into 1000 downstream calls in minutes.

OpenAI Assistants/Responses 在 Go 里的上下文爆炸治理：截断策略、摘要回填与成本上限

Mon, 02 Mar 2026 12:44:00 +0000

线上 Agent 一跑久了就会遇到同一个坑：上下文越来越长，延迟飙升、费用失控，最后还更容易答偏。

这不是模型“变笨”了，通常是上下文治理没做：该留的没留、该删的没删、该摘要的摘要坏了。

Taming Context Explosion in OpenAI Assistants/Responses with Go: Truncation, Summary Backfill, and Cost Caps

Mon, 02 Mar 2026 12:44:00 +0000

Long-running agent sessions usually fail the same way: context keeps growing, latency spikes, costs blow up, and answer quality gets worse.

That is rarely a model-quality issue. It is almost always missing context governance.

Go + OpenAI API Timeout Troubleshooting: DNS, TLS, Proxy, and Connection Pool

Mon, 02 Mar 2026 01:12:10 +0000

When OpenAI API calls start timing out in production, the real problem is usually not “OpenAI is down.”

The real problem is you don’t know which hop is failing: DNS, TLS handshake, proxy path, or your own connection pool.

GitHub Actions + AI Agent Auto-Fix Pipeline: Failure Tiers, Regression Gates, and Security Guardrails

Fri, 27 Feb 2026 01:18:38 +0000

When CI keeps failing, the real risk is not “slow fixes” — it is “fast bad fixes.” This guide gives you a practical GitHub Actions + AI Agent auto-fix pipeline with failure tiering, strict edit boundaries, and merge-time gates.

GitHub Actions + AI Agent 自动修复流水线：失败分级、回归测试与安全闸门

Fri, 27 Feb 2026 01:18:38 +0000

线上 CI 一旦连续红灯，团队最怕的不是“修得慢”，而是“修坏更多”。这篇给你一套可落地的 GitHub Actions + AI Agent 自动修复流水线：先做失败分级，再限制 AI 修改范围，最后用回归与安全闸门兜底。

OpenAI Agents SDK with Go: Tool Calling, Session Memory, and Error Recovery

Wed, 25 Feb 2026 01:18:00 +0000

Most teams can connect an LLM in a demo. The real pain starts in production: multi-step tasks, flaky tool calls, unclear retries, and rising cost.

This guide gives you a pragmatic Go-first blueprint for shipping an Agent workflow that can survive real incidents.

OpenAI Responses API Streaming in Go: Timeouts, Retries, and Observability

Mon, 23 Feb 2026 01:15:00 +0000

Production streaming fails in two predictable ways: users wait while the stream silently drops, and your logs say “timeout” without telling you where it actually broke.

This guide gives you a practical Go pattern for OpenAI Responses API streaming with strict timeout boundaries, safe retries, and useful telemetry.

Claude Code + Codex for Multi-Model Development: Cost, Speed, and Quality (Practical Workflow)

Sun, 15 Feb 2026 10:30:00 +0800

If you still use one model for everything, you usually pay in one of three ways: higher cost, slower delivery, or more rework.

A better setup is role-based collaboration: Claude Code for planning and quality gates, Codex for fast implementation and batch edits.