💛 Sponsorware + 开源支持： qq-ai-bot 完全免费，欢迎 扫码支持（¥9.9 起）。也有付费评审服务 ¥29.9–¥199。

Backend / systems engineer focused on Go, Java, distributed systems, protocol / RPC infrastructure, performance-oriented engineering, and selective real OSS fixes.

Current flagship: qq-ai-bot — a self-hosted QQ ↔ AI bot scaffold with OneBot 11, NapCat / LLOneBot, ACP-compatible agents, persistent sessions, progress streaming, Prometheus-style /metrics, and Docker. It now has a public project page, Docker quickstart, a stable public image tag (v0.1.4) plus a moving latest tag, ecosystem references, an official OneBot community discussion, and active upstream docs / app-directory PRs, so reviewers can judge it as a small operator-facing systems asset rather than a one-night chat demo. Recent OSS work spans HashiCorp retry semantics, Prometheus client edge cases, jwt validation, chi routing, urfave/cli, GitHub CLI, OpenTelemetry, and Docker / GitHub Docs improvements.

Reading guide for recruiters / reviewers: the repos that best represent my current direction are the pinned repositories below plus the linked upstream PRs. If you see many public forks under my account, treat them as temporary contribution vehicles for upstream pull requests, not as portfolio centerpieces.

If you only open three repos: start with qq-ai-bot → happydb → go-service-starter.

• Building reliable backend services and reusable developer tooling
• Strong interest in distributed systems, storage, networking, performance, and CS fundamentals
• Shipping practical public assets and selective open-source fixes that are small but real
• Turning rough engineering ideas into public repos that feel closer to products than class projects

• Repo — qq-ai-bot · project page
• Why it exists — keep QQ / OneBot transport, session orchestration, and ACP agent execution clearly separated so the system feels like bot infrastructure rather than a one-off chat demo
• What it already does — supports NapCat / LLOneBot, forward / reverse WebSocket wiring, per-chat session reuse, progress messages back to QQ, Docker quickstart, and public image packaging via both ghcr.io/happysnaker/qq-ai-bot:v0.1.4 and ghcr.io/happysnaker/qq-ai-bot:latest
• External proof — already listed on the public OneBot ecosystem page, posted in the official OneBot Discussions / 应用 / SDK, and included in awesome-agent-client-protocol
• Active ecosystem work — official ACP clients docs PR · official LLOneBot docs PR · NapCat docs integration landed · Docker Compose sample PR · CasaOS app-store PR · ACP protocol discussion
• Why it is on this profile — it best reflects the direction I want to be judged on: protocol integration, transport / agent boundaries, operator-friendly packaging, and reusable engineering assets

• Public OneBot ecosystem placement, so the project is no longer only self-claimed
• Public official OneBot discussion in the protocol community’s own “应用 / SDK” channel, so the repo now has one protocol-native showcase thread instead of only its own outbound links
• Public ACP protocol discussion grounded in a real richer-media / attachment boundary from implementation work
• Active ACP clients docs PR to get qq-ai-bot listed as a Messaging client in the official ACP docs
• Active LLOneBot docs PR to document qq-ai-bot as an integration path in the official LLOneBot docs
• NapCat docs integration landed, so qq-ai-bot is now also present in NapCat community / integration-facing docs rather than only in outbound showcase threads
• Active docker/awesome-compose#781 to place a NapCat + qq-ai-bot local stack into a widely browsed Docker Compose samples repo
• Active CasaOS-AppStore-Play#42 to package qq-ai-bot as a more browser-first homelab install surface rather than only a README + Compose story
• Public recommendation / weekly threads are also live in HelloGitHub, GitHubDaily, 阮一峰周刊, howie6879/weekly, and developer-plus/weekly, so discovery is no longer limited to protocol-native communities

This is the strongest current “public proof” chain on the profile: a repo, a project page, ecosystem references, and upstream protocol / docs work around the same asset.

• qq-ai-bot — self-hosted QQ ↔ AI bot scaffold using OneBot 11, ACP-compatible agent bridging, persistent sessions, progress streaming, Docker quickstart, and a public project page at happysnaker.github.io/qq-ai-bot
• happydb — learning-oriented relational database implementation in Java covering storage, indexing, MVCC-style visibility, recovery, query execution, optimization, and replication experiments
• HRpc — Java 11 / Netty based RPC framework learning project with custom protocol, service registry, dynamic proxy invocation, heartbeats, reconnect, and load balancing
• go-service-starter — minimal production-minded Go HTTP service starter with config loading, structured logging, health endpoints, graceful shutdown, and Docker packaging
• go-http-middleware-kit — reusable net/http middleware for request IDs, structured logging, panic recovery, timeouts, and real IP handling
• CSAPPLabsAndNotes — CS:APP lab notes, systems-learning walkthroughs, and low-level computer-systems study material
• Resume — responsive HTML/CSS/JS resume and portfolio template for GitHub Pages, personal sites, and developer landing pages
• system-design-checklist — practical system-design checklist, answer sheet, and worked examples for backend interviews, architecture reviews, and distributed-systems tradeoffs
• production-readiness-checklist — practical release-review, launch-gate, and on-call handoff checklist with copy-paste templates for production work

• For strongest signal first — start with qq-ai-bot, go-service-starter, go-http-middleware-kit, happydb, and the linked upstream PRs below
• For interview / systems fundamentals — CSAPPLabsAndNotes, happydb, and the checklist repos are the fastest read
• For portfolio / packaging signal — Resume, github-profile-checklist, and the public project pages
• What to de-prioritize — short-lived public forks of external repos usually exist only because I was sending or updating upstream pull requests there

Code / behavior fixes

• hashicorp/go-retryablehttp#288 / #289 / #290 / #291 / #292 / #293 — a run of focused fixes for final-response preservation, typed-nil request bodies, logger safety, readable retry Backoff bodies, deadline-aware retry waits, and zero-value Client safety
• prometheus/client_golang#2040 — clamp out-of-range formatted timestamps so model.Earliest / model.Latest no longer serialize into values that can overflow back on the server side
• golang-jwt/jwt#520 — add a required issued-at validation option without changing the existing WithIssuedAt() behavior
• spf13/pflag#483 / #484 / #485 / #486 / #487 — fix nil-default GetIP() handling, allow hex input in UintSlice, make explicit empty typed slice values round-trip as empty slices instead of parse errors, allow explicit empty StringToString overrides, and restore IsBoolFlag() compatibility for custom bool-like values
• go-chi/chi#1120 — fix Host-based routing in RouteHeaders by using Request.Host, plus tests and doc updates
• urfave/cli#2379 — prevent v2 shell completion after -- from accidentally executing command actions, with regression coverage
• urfave/cli#2381 — reject legacy v1-style Name: "flag, f" alias syntax with an explicit migration error instead of silently dropping the alias
• cli/cli#13766 — fix gh skill install --dir ... so a custom install directory no longer still forces an interactive target-agent selection step
• urfave/cli-altsrc#50 — make config-backed map/object values round-trip correctly into StringMapFlag via the same serialized format used by urfave/cli/v3
• Built and shipped go-http-middleware-kit — a small Go middleware library for request IDs, logging, panic recovery, real IP extraction, and timeouts
• Packaged and published happydb as a public systems-learning project around database internals, storage, recovery, and distributed-systems experiments

Quality / observability / SDK

• prometheus/procfs#836 — tolerate wrapped signed /proc/stat processes counters so Stat() consumers do not fail on long-running systems exposing the wrapped fork count as a negative decimal string
• prometheus/procfs#833 — split proc stat limit coverage by architecture so parser limit checks stay correct on both 32-bit and 64-bit targets
• open-telemetry/opentelemetry-go#8527 — document supported SDK environment variables across resource, trace, metric, and log package docs

Selective docs / developer experience

• prometheus/client_golang#2034 — add an OTLP bridge tutorial for exporting existing Prometheus instrumentation through OpenTelemetry
• github/docs#45002 — add SHA pinning notes to OIDC workflow examples across AWS, Azure, GCP, Vault, and PyPI docs
• docker/docs#25462 — clarified that the Ubuntu noble base-image example is version-specific and should be adjusted for the release being imported
• rclone/rclone#9559 — clarified copyto command documentation with maintainer-aligned wording

Contribution focus

• Small but real behavior fixes, API correctness, retry semantics, parser edge cases, routing behavior, observability/client edge cases, and selective documentation only where implementation ambiguity causes real user error
• Recent work spans HashiCorp libraries, Prometheus client_golang / procfs, golang-jwt, chi, urfave/cli, GitHub CLI, OpenTelemetry, GitHub Docs, Docker Docs, and rclone

Languages:      Go, Java, C/C++, SQL
Interests:      Backend engineering, RPC, distributed systems, storage, networking
Strengths:      CS fundamentals, hands-on implementation, reusable engineering assets
Open to:        Backend / infrastructure / systems engineering opportunities

If my open-source work, reusable templates, code contributions, or engineering assets save you time, you can support ongoing maintenance here.

• Support page: happysnaker.github.io/support — WeChat Pay / Alipay QR codes live there
• If you came here from qq-ai-bot: the cleanest support note is simply qq-ai-bot, and the fastest useful paid path is the ¥29.9 quick read for another bot / repo / landing page
• Fastest low-friction tip: ¥9.9 / ¥19.9 if one repo, checklist, or OSS fix saved you time
• Most useful paid entry: ¥29.9 quick read for a blunt first-impression pass on your profile / repo / landing page — review page
• Best packaging option: ¥99 async review for pinned repos, README cleanup, landing-page positioning, or profile packaging — review page
• Preview first: redacted sample audit
• Current July offer: first paid request gets one extra public page / README in the same pass at no extra charge
• One-click quick-read email: Quick read | profile / repo / page link
• One-click async-review email: Async review | target role | repo / profile link
• If qq-ai-bot helped with OneBot / ACP wiring, the cleanest support note is simply qq-ai-bot
• If Resume, CSAPPLabsAndNotes, github-profile-checklist, or one recent OSS fix helped, direct support is especially appreciated

• Email: happysnaker@foxmail.com
• Portfolio site: happysnaker.github.io/Resume
• Blog: happysnaker.github.io
• Support / tip: happysnaker.github.io/support