Credit Genie：https://creditgenie.com/landing-page-1?pid=googleads_int&c=%7Bcampaign_name%7D&af_c_id=24066887466&af_adset=%7Badgroupname%7D&af_adset_id=202467525767&af_ad=799105170228&af_ad_id=%7Bcreative_id%7D&utm_source=googleads_int&utm_campaign=24066887466&utm_medium=cpc&gclid=Cj0KCQjwh4TVBhCWARIsAG0czmo5diIYTZ2ZlDTnutG77Vmr1A7jSm2hXKizt-Ff8UwkIm-iJPlKrskaAgGeEALw_wcB&utm_content=credit%20genie&utm_term=202467525767_799105170228&gad_source=1&gad_campaignid=24066887466&gbraid=0AAAAABccNIiLOCaB5_V8c7zPTF_WkXr8Y&gclid=Cj0KCQjwh4TVBhCWARIsAG0czmo5diIYTZ2ZlDTnutG77Vmr1A7jSm2hXKizt-Ff8UwkIm-iJPlKrskaAgGeEALw_wcB 是一个以移动优先为理念的金融健康平台，旨在帮助个人掌控自己的金融未来。其 AI 和 ML 工程团队利用 OpenWiki，LangChain 的开源仓库文档代理，构建了一个自助门户，将文档汇总到一个可搜索、可读的人类界面中。
随着 Credit Genie 的 AI 和 ML 工程团队在日益增长的代码库中不断扩展，保持文档的更新变得越来越困难。关于单个项目的知识开始被贡献者孤立起来，并且形成了小规模的“部落知识”。
尝试正式化手动文档流程未能成功，因为撰写和维护良好的文档会占用其他紧迫工作的时间。对于繁忙的团队来说，文档成为一种高频变动的活动，需要持续维护。
为了解决这个问题，Credit Genie 采用了 LangChain 的 OpenWiki：https://github.com/langchain-ai/openwiki，以全面自动化生成和维护高质量的仓库文档，使人类和编程代理都能轻松找到所需的信息。
在使用 OpenWiki 之前，团队维护了项目特定的 Notion 页面、README 文件和 AGENTS.md 文件。然而，在 Credit Genie 快速发展的环境中，这些内容很快就过时了。
文档不仅变得陈旧，而且难以查找——不仅对人类来说是如此。跨代码库工作的编程代理也遇到了同样的问题：没有可靠的真相来源，它们很难理解仓库的功能或子系统的工作方式，并且经常基于过时或不完整的上下文进行操作。
团队意识到重要的知识正越来越集中在个人手中。关于某个仓库的功能、某个特性的位置或某个子系统的工作方式的问题，答案往往很难找到，有时甚至难以获得——无论提问者是队友还是编码代理。
最初作为 Credit Genie 的 AI 和 ML 负责人 Mattia Ciollaro 的一个副项目，它已经演变为他的团队的“机构知识保险政策”。
Mattia 将 OpenWiki 进行了改造，建立了一个自助门户网站，托管在 GitHub Pages 上，能够汇总所有已接入仓库的 OpenWiki 文档，形成一个可搜索的单一网页界面。其结果是一个不仅为团队提供每日可信上下文来源的系统，还通过更快地呈现正确信息、减少猜测并将实现保持在源代码这一事实基础上，从而帮助编码代理提升性能。
文档存放在每个项目仓库的 openwiki/ 文件夹中。OpenWiki 每晚运行，检查提交变更，并在发现有意义的差异时生成文档更新，在仓库中发起拉取请求。接入新仓库只需在门户配置中做一个更改，就会触发一个自动拉取请求，将 OpenWiki GitHub Actions 工作流程的定制版本添加到目标仓库：https://github.com/langchain-ai/openwiki/blob/main/examples/openwiki-update.yml。
为了消除合并瓶颈，Credit Genie 添加了一个操作，以自动批准和合并 OpenWiki 更新的 PR，从文档维护过程中移除人工摩擦。一个清理任务每天运行，遍历所有已接入仓库，合并待处理的 OpenWiki 更新，并重建门户，确保最新内容始终可用。
这为 Mattia 及其团队提供了一种简单、可扩展的方法，将文档转变为开发生命周期的一部分，而不是一个单独的维护负担。
Mattia 说道：“OpenWiki 让我们能够将文档成为代码生命周期的一部分，而不是作为一个单独的任务。”
OpenWiki 在 Credit Genie 的使用方式有两种明显的不同：
正如Mattia所描述的：“为了跟踪我们系统的持续演进，工程师和非技术相关方现在会查看OpenWiki门户。通常，这能避免阅读大量代码所需的精力，也避免追查并采访所谓的‘专家’所需的时间和协调。高质量、自动生成并始终保持最新的文档消除了这两类障碍，使每个人在需要时都能了解我们的系统如何运作。”
Mattia的团队已经看到了强劲的采用和实际价值。迄今为止最大的收益包括：
Credit Genie的下一个目标是通过整合其内部跨代码库知识图，使OpenWiki更加智能，并逐步让所有Credit Genie工程团队和代码库加入。
“来自我们Web团队的Ian Yoder和Anela Trakic，以及应用团队的Patrick VanDusen帮助我们理解了系统可以扩展的方式。他们加入了自己团队的多个代码库，并验证了生成文档的准确性。AI让每个人都感到兴奋，并创造了跨部门协作的良好机会，”Mattia补充道。
这是与Credit Genie工程主管Adil Hashem以及正在构建内部AI工具的工程团队的共同努力——包括LLM网关、自定义CLI和GraphRAG代码图。
如今，Credit Genie的工程团队已经在绘制代码库和项目之间的依赖关系图。下一阶段是将该图引入OpenWiki，使文档不仅能显示一个代码库的功能，还能显示它与其他系统的连接方式以及变更可能产生的影响。这将让所有工程师和编程代理在进行代码更改前，清楚地了解影响范围。
想要保持与代码同步更新的文档吗？试试OpenWiki：https://github.com/langchain-ai/openwiki，并在你自己的代码库上进行设置。
我们的代理工程平台LangSmith帮助开发者调试每一个代理决策、评估变更，并一键部署。
Credit Genie：https://creditgenie.com/landing-page-1?pid=googleads_int&c=%7Bcampaign_name%7D&af_c_id=24066887466&af_adset=%7Badgroupname%7D&af_adset_id=202467525767&af_ad=799105170228&af_ad_id=%7Bcreative_id%7D&utm_source=googleads_int&utm_campaign=24066887466&utm_medium=cpc&gclid=Cj0KCQjwh4TVBhCWARIsAG0czmo5diIYTZ2ZlDTnutG77Vmr1A7jSm2hXKizt-Ff8UwkIm-iJPlKrskaAgGeEALw_wcB&utm_content=credit%20genie&utm_term=202467525767_799105170228&gad_source=1&gad_campaignid=24066887466&gbraid=0AAAAABccNIiLOCaB5_V8c7zPTF_WkXr8Y&gclid=Cj0KCQjwh4TVBhCWARIsAG0czmo5diIYTZ2ZlDTnutG77Vmr1A7jSm2hXKizt-Ff8UwkIm-iJPlKrskaAgGeEALw_wcB is a mobile-first financial wellness platform designed to help individuals take control of their financial future. Its AI and ML Engineering teams use OpenWiki, LangChain's open-source repo-documentation agent, to build a self-serve portal that aggregates documentation into a single, searchable, human-readable interface.
As Credit Genie's AI and ML Engineering teams scaled across an ever-growing codebase, keeping documentation current became increasingly difficult. Knowledge about individual projects started to become siloed with their contributors, and pockets of "tribal knowledge" began building up.
Attempts to formalize manual documentation processes fell short, since writing and maintaining good docs takes time away from other pressing work. For busy teams, documentation becomes a high-churn activity that requires constant upkeep.
To solve this, Credit Genie adopted LangChain's OpenWiki：https://github.com/langchain-ai/openwiki to fully automate the generation and maintenance of high-quality repository documentation, making it easy for both humans and coding agents to find the information they need.
Before OpenWiki, the team maintained project-specific Notion pages, README files, and AGENTS.md files. However, in Credit Genie’s fast-moving environment, that content quickly became obsolete.
Docs didn't just become stale. They were also hard to find – and not just for people. Coding agents working across the codebase ran into the same problem: without a reliable source of truth, they had a hard time understanding what a repository did or how a subsystem worked, and would often act on outdated or incomplete context.
The team recognized that important knowledge was becoming increasingly concentrated in individuals. Answers to questions about what a repository did, where a feature lived, or how a particular subsystem worked were often slow to find and sometimes hard to get at all – whether the one asking was a teammate or a coding agent.
What originally started as a side project for Mattia Ciollaro, Credit Genie’s Head of AI and ML, has evolved into an "institutional knowledge insurance policy" for his teams.
Mattia adapted OpenWiki to build a self-serve portal hosted on GitHub Pages that aggregates OpenWiki docs across all onboarded repositories into a single, searchable web interface. The result is a system that not only gives his teams a daily source of trusted context, but also helps coding agents perform better by surfacing the right information faster, reducing guesswork, and keeping implementation grounded in the source of truth: the code itself.
The documentation lives inside each project repository in the openwiki/ folder. OpenWiki runs nightly, checks for commit changes, and generates documentation updates when it finds meaningful differences, opening a pull request in the repo. Onboarding a new repo requires just one change in the portal config, which triggers an automated pull request that adds a customized version of the OpenWiki GitHub Actions workflow：https://github.com/langchain-ai/openwiki/blob/main/examples/openwiki-update.yml to the target repo.
To eliminate merge bottlenecks, Credit Genie added an action to automatically approve and merge OpenWiki update PRs, removing human friction from the documentation maintenance process. A sweep job runs daily across all onboarded repositories, merges pending OpenWiki updates, and rebuilds the portal so the latest content is always available.
This gave Mattia and his teams a simple, scalable way to turn docs into a part of the development lifecycle rather than a separate maintenance burden.
"OpenWiki allows us to make documentation part of the code lifecycle, instead of treating it like a separate chore," said Mattia.
OpenWiki is used in two distinct ways at Credit Genie:
As Mattia described it, "To keep track of the continuous evolution of our systems, Engineers and non-technical stakeholders now check the OpenWiki portal. Often, this removes the need to read a substantial amount of code, which takes effort, or to track down and interview whoever the 'expert' is, which takes time and coordination. High-quality, autogenerated documentation that's always up to date removes both of those obstacles and allows everyone to learn how our systems work any time the need arises."
Mattia’s teams are already seeing strong adoption and practical value. The biggest gains so far include:
Credit Genie's next goal is to make OpenWiki even more intelligent by incorporating its internal cross-repository knowledge graph and progressively onboarding all Credit Genie Engineering teams and code repositories.
“Ian Yoder and Anela Trakic, from our Web team, and Patrick VanDusen, from our Application team, helped us understand that we could scale the system. They added a number of their team repositories and verified the accuracy of the generated docs. AI gets everyone excited and creates great opportunities to collaborate across departments,” Mattia added.
This is a joint effort with Adil Hashem, Credit Genie's Head of Engineering, and the Engineering teams who are building internal AI tools – including LLM gateways, custom CLIs, and GraphRAG code graphs.
Today, Credit Genie’s Engineering teams already map how repositories and projects depend on one another. The next phase is to bring that graph into OpenWiki so documentation can show not only what a repository does, but how it connects to other systems and how changes could ripple across them. This will give all engineers and coding agents at Credit Genie a clearer view of blast radius before making code changes.
Want docs that stay up to date with your code? Try OpenWiki：https://github.com/langchain-ai/openwiki and set it up on your own repos.
LangSmith, our agent engineering platform, helps developers debug every agent decision, eval changes, and deploy in one click.
扫码直达文章详情