> For the complete documentation index, see [llms.txt](https://docs.visdom.virtuslab.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.visdom.virtuslab.com/visdom.md). # Visdom An AI-native SDLC platform, self-hosted, built by VirtusLab. Visdom is an AI-native SDLC platform: what it is for, what it is made of, how the components work together, how a change moves through it, and what a deployment looks like. The platform is fully self-hosted. It runs on your Kubernetes cluster, against your databases, your identity provider, your secret manager and your model endpoints. There is no vendor-side data path and no outbound telemetry. ## The problem AI coding agents attack the one part of software delivery that was never the bottleneck. Writing code was never the constraint. Understanding the system, validating the change, reviewing it and shipping it safely always was. Agents produce code faster than that machinery can absorb it, and they do it without knowing what the organisation knows and without leaving an account of what they did. ## The platform Visdom supplies the missing machinery: organisational context for agents, a defined process to run them inside, validation layers a model cannot talk its way past, and a signed record of every step. It is composable, it runs entirely on your own infrastructure, and it treats the coding agent as a replaceable part rather than the product. * **A production process, not an agent wired to a repository.** Versioned flows, approval gates where you place them, failure handling, and bounded fix-and-review loops. * **Context that is compiled, not rediscovered.** Ownership, dependencies, conventions and decision history served to agents, so a fix matches how your organisation actually works. * **Cost designed in.** Work that does not need a model runs as a deterministic task, review depth follows assessed risk, and spend is attributed per model, project, flow and author. * **Evidence you can hand to an auditor.** Every prompt, model call, tool use and file edit in an Ed25519-signed hash chain, verifiable offline, with attribution down to the line. ## One change, end to end ![One change moving through the platform, from ticket to merged pull request](https://377380473-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F2U8RqIIMUGxclsDYVNc3%2Fuploads%2Fgit-blob-b154e0672799e1ae72d313546a7e2bfa83f68dbf%2Foverview.png?alt=media) A ticket lands. A flow picks it up on its own schedule, reads it, and stops there if the ticket does not say enough to act on, leaving a comment saying what is missing rather than a pull request nobody asked for. If it does say enough, an agent gets what a coding agent normally never has: the repository's conventions, the ownership map, the dependency graph around the code it is about to touch, and the decisions that explain why that code looks the way it does. It works in a sandbox whose egress is an allowlist and whose secrets it never sees the values of. What comes out goes through gates the model cannot argue with: a deterministic pass first, then risk classification, then review lenses proportional to what the change actually touches, then architecture rules and mutation testing that hold regardless of who wrote the code. Trivial findings loop back to the agent for a bounded number of rounds. The rest goes to a person, with the findings attached. Underneath all of it, one signed chain records every prompt, model call, tool use, file edit and commit, with the cost attached and attribution down to the line. ## Where to go next The thesis and the concept explain why the platform is shaped this way. The logical architecture lists the components and what each is responsible for. The data flow follows one issue from webhook to merge decision. Deployment architecture and the security summary cover what running it involves. Every component with its own documentation has a section in the tabs above: **Visdom AI Tracing**, **Visdom Context Fabric**, **Visdom Orchestrator**, **Visdom Code Review**, **Visdom Testing** and **Visdom Security**. The Control Plane, which is administration rather than a product surface of its own, is described here. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.visdom.virtuslab.com/visdom.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.