独立运营 · 不出售排名位置  · 我们怎么评估工具 →
Anclum 提交收录
P

Pizzylee001/tracemend

本站为独立目录,非该工具官网;下方信息取自其公开仓库。

Turn blocked agent outputs into verified DPO preference pairs on Reppo and SERV Reasoning (原文为英文)

装之前:没有可直接复制的命令 · 许可证未标注 · 链接可访问

第一步:这条暂时没有可执行的下一步,可先看原仓库 https://github.com/Pizzylee001/tracemend

checked 2026-11-20 — 命令里的包名与条目不一致,未作为可复制命令展示。

更新 2026-11-15
在 GitHub 查看 ↗链接已于 2026-11-17 核查 · HTTP 200 · 打开的是厂商自己的官网

装完之后怎么算装好了

  1. 装完该看到什么 — 命令正常结束时退出码为 0,终端里没有 "command not found" 或 "permission denied"。
  2. 一行自检 — 按该仓库 README 里给出的最小用法跑一次;能按 README 的描述输出就是通了。
  3. 答不出来怎么办 — 这个项目没有写排错说明,去仓库的 Issues 搜 install。

安装命令取自该仓库;以上判据只描述可观察到的结果,不替代项目自己的文档。

★ 数取值
0 ★ · Pizzylee001/tracemend · 取自 GitHub 搜索 API · (2026-11-19 20:00:48 核对)
来源明细
从该仓库的 GitHub 页面直接读取 · 条目类型:开源项目
收录与活跃
收录于 2026-11-16 · 最近见到 2026-11-20 · 共见到 5 次 · 被 2 个来源列出
评分构成
综合分 203 · 2 个来源 · repo
安装
安装命令见上方「安装与配置」。
常见问题(答案都来自我们的记录)
它适合谁 / 用来做什么?
我们把它归在这些标签下: 开源项目。标签来自条目自身,不是我们强加的分类。
谁维护它、还在更新吗?
仓库 Pizzylee001/tracemend;我们收录于 2026-11-16;最近见到 2026-11-20。这里没有"官方认证"或"质量评级"——我们只记录来源与时间。
README 原文(英文,未翻译) · 在 GitHub 读完整 README →

# TraceMend

以下摘录自该仓库的 README(片段)

Turn blocked or flagged AI agent outputs into grounded, citation-verified DPO preference pairs using Reppo DataNet evidence pods and SERV Reasoning (Gpt 6 Luna).
-1F2421?style=for-the-badge)
## Languages & Tech Stack
| Layer | Technologies | Usage |
| Languages | TypeScript 5.9, CSS, HTML5, JavaScript | App Router pages, API route handlers, mineral paper stylesheets, design specimen, and build configs |
| Framework & UI | Next.js 16.4, React 19.2, Tailwind CSS v4 | Multi-route workbench (/mend, /datanets, /vault), server endpoints, and Canvas/SVG telemetry |
| Motion & Icons | Motion 12, GSAP 3, Phosphor Icons | Reduced-motion-aware panel choreography and hairline iconography |
| Validation & Tests | Zod 4.1, Vitest 4.0, ESLint 9 | Strict API schema validation, DPO compiler verification, and linting |
| Protocols | Reppo DataNets & Eval API, SERV Reasoning (Gpt 6 Luna) | Evidence pod retrieval, evaluation polling, and citation-grounded repair |
## What TraceMend Does
AI agents fail evaluation gates when claims cite outdated dates, unverified numbers, or missing sources. Raw evaluation APIs return a score and a critique, leaving the engineer to hunt for evidence, rewrite the draft, re-verify the lift, and format training data by hand.
TraceMend closes that loop across three linked routes:
1. Mend Bench (/mend): Load a failed or flagged agent draft (answer, plan, trace, or artifact), run an instant SERV Pre-Screen (Gpt 6 Luna) or submit directly to live Reppo Evaluation Markets (eval.reppo.xyz), pin verifiable Reppo DataNet pods, generate a grounded repair via SERV Reasoning (serv_prompt_guard + serv_shadow_agent + static BRAID prompt caching + strict JSON schema), scrub clause-by-clause diffs with the Kintsugi Optical Seam comparator, re-audit score lift, and save the (rejected, chosen) pair.
2. DataNet Evidence Registry (/datanets): Browse all 19 live Reppo public DataNets, filter by market name, token symbol, or CUID, inspect up-vote vs down-vote ratios and epoch emissions, review IPFS pod provenance, and pin records straight into /mend.
3. DPO Preference Vault (/vault): Filter verified remediation runs (BLOCK -> AUTO-SHIP, REVIEW -> AUTO-SHIP, or your own saved runs), inspect compiled OpenAI/TRL-compatible DPO JSONL records, copy records to clipboard, or download .jsonl datasets and Markdown audit receipts.
## Screenshots
### Mend Bench (/mend)
### Kintsugi Seam Diff Comparator & SERV BRAID Trace
### Reppo DataNet Evidence Registry (/datanets)
### DPO Preference Pair Vault (/vault)
### Mobile Responsive View (375px with Reduced Motion)
## Architecture & Protocol Integration
### 1. Reppo DataNets & Evaluation Markets (src/lib/reppo.ts)
Public Subnets: GET https://reppo.ai/api/v1/public/subnets?page=1&limit=30 fetches active DataNets (19 live subnets) with automatic normalization and captured snapshot fallback if an upstream feed is unreachable.
Public Evidence Pods: GET https://reppo.ai/api/v1/public/pods?page=1&limit=80 retrieves recent pod records, sanitizes external links (http: / https: only), and applies server-side subnetId and search filtering.
Live Evaluation Submit & Poll: POST https://eval.reppo.xyz/api/v1/agents/eval-requests/submit and GET https://eval.reppo.xyz/api/v1/agents/eval-requests/{evalId} dispatch answer, plan, trace, or artifact payloads with bounded settlement windows (10 to 1440 minutes) and surface clear UI feedback on 402 (credit balance), 409 (active pending evaluation conflict), and PENDING settlement states.
收录徽章(可选)
想把收录标记放进 README 就复制这行,可选,不影响收录。
[![Listed on Anclum](https://anclum.com/badge/listed.svg)](https://anclum.com/zh/s/pizzylee001-tracemend/)
来源
github · https://github.com/Pizzylee001/tracemend
按领域浏览
开源项目
收录来源
2 个独立来源
收录于
今天

更多条目

全部 1,875 →

所属领域与标签

每个标签都能点进去看同类的全部条目。数字是当前库里的真实条数。

浏览全部领域 →

← 返回索引