Skip to content
View vicfei's full-sized avatar
🎯
Focusing
🎯
Focusing
  • Zhipu AI
  • ShangHai

Block or report vicfei

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
vicfei/README.md
Hi

Currently focused on agent orchestration & LLM tooling.

📍 Shanghai, China 🧭 RAG · DeepSeek Harness Ecosystem · Jev Decision Models

💼 Work Experience

Zhipu AI (智谱) — LLM & Agent Engineering — 4 yrs · 2022 → Present
Tencent · WeChat (微信) — NLP & Infrastructure — 5 yrs

Nine years across China's largest social platform and its frontier LLM lab — NLP & infrastructure at WeChat, then agent infrastructure at Zhipu (GLM).

💻 Open Source Experience

  • 🔐 DeepSeek Harness ecosystem — author of dsh-super-agent (the independent, executor-agnostic admission-control implementation for any CLI coding agent) and the plugin family — dsh-weknora-ops (the WeKnora write path) · dsh-trustquant (oracle-pinned finance tools) — live installs: dsh-super-agent weknora-ops trustquant
  • 🗺️ Author of dsh-plugin-atlas — a live, daily-refreshed map & ranking of the whole DSH plugin ecosystem
  • 🏆 WeKnora contributor — 15 merged PRs; #4 by both commits and merged PRs in Sep 2026 (Tencent's open-source LLM knowledge platform, 30k+ stars)
  • ✨ Contributor — deer-flow · Octop

Featured

  • dsh-super-agent — admission control, observe mode & tickets for any CLI coding agent: admission tables, explicit/inferred strength, off → observe → auto, per-ticket worktrees, PR delivery with approval-gated auto-merge, GitHub App identity. Works with claude / codex / opencode / anything. Stars
  • awesome-jev-prompts — question patterns, anti-patterns & calibration notes for typed decisions: 63 patterns / 10 categories, bilingual, an Engine Cookbook that runs one question across five engines, and a live leaderboard of the ecosystem. Stars

Focus areas: agent infrastructure & admission control, RAG in production, knowledge platforms, decision-layer models, question design & calibration, embodied agents on cheap hardware.


中文:9 年工程经验——腾讯微信 5 年(NLP 与基础设施),智谱 4 年(LLM 工具链与 Agent 基础设施)。现在主推两个项目:dsh-super-agent(DeepSeek Super-agent 准入控制的独立实现,任意 CLI 编码智能体可用,npm 持续装机)与 awesome-jev-prompts(63 条双语问题模式 + 引擎手册 + 生态排行榜);同时深耕 DeepSeek Harness 生态,插件矩阵 dsh-weknora-ops、dsh-trustquant 等在 npm 持续装机;另为 WeKnora 贡献者(15 个被合并 PR;2026 年 9 月单月提交与合并双榜第 4)。

Pinned Loading

  1. awesome-jev-prompts awesome-jev-prompts Public

    Question patterns, anti-patterns & calibration notes for typed decision models — Jev & every System One-compatible engine. 60 entries / 10 categories + linter + engine cookbook. EN/中文.

    Python 14 1

  2. dsh-super-agent dsh-super-agent Public

    Independent implementation of the DeepSeek Super-agent design: admission control, observe mode & tickets for any CLI coding agent (Claude Code, Codex, opencode…) — no DSH required

    TypeScript 26 4

  3. RoboHarness RoboHarness Public

    A general robotic harness — LLM agents operate real arms via deterministic primitives with a current-feedback safety layer. Zero training. Hardware: CreateBlock 5-DOF today, mainstream arms next.

    Python