Point Secrin at any repo. It builds a knowledge graph, summarizes every function and class, and generates a complete Markdown wiki — then keeps it in sync on every commit.
Every engineering team knows the feeling. The code works. The docs don't exist or they're six months out of date.
01
New engineer? Two weeks wasted.
Every new hire spends their first two weeks reading code and interrupting senior engineers with questions that should have been documented months ago.
02
Wikis go stale instantly.
The moment someone writes a wiki page, it starts decaying. Code changes. The doc doesn't. In six months no one trusts what they read so they read the code anyway.
03
Knowledge lives in Slack.
Architecture decisions get made in Slack threads and DMs. Someone leaves the team and takes critical context with them. There's no record of why things were built the way they were.
04
Senior engineers are the bottleneck.
Every question that could have been answered by good documentation lands in the inbox of your most senior and most expensive engineers. They answer the same questions on repeat.
Secrin is not a docs generator. It's a living knowledge system understands your code structurally, and keeps that understanding current on every commit.
01
Code knowledge graph
Parses every file into a structured graph of modules, classes, and functions with typed edges for calls, imports, and inheritance. Ask “what calls this?” and get an instant, accurate answer.
02
LLM summaries
Every function, class, and file gets a plain-English summary written by an LLM. Works with Ollama locally no API key needed. Or plug in OpenAI or Anthropic via .secrin.yml.
03
Domain extraction
Identifies the business concepts your codebase implements — Authentication, Billing, User Management and maps which code owns each domain. Instantly shows the big picture.
04
Wiki generation
Generates a structured Markdown wiki at docs/wiki/ module pages, domain pages, architecture diagrams. Committed to your repo, versioned with your code.
05
Chat Q&A
Ask architecture questions in plain English. Answers are grounded in the knowledge graph not hallucinated. Cites the exact files and functions behind each answer.
06
Always in sync
A git post-commit hook detects changed files, re-analyzes them, and regenerates affected wiki pages automatically. Your wiki is always one commit behind at most.
From a fresh clone to a complete, searchable knowledge base in minutes. Then it stays current on its own.
$ secrin init
Connect your tools
Interactive wizard. Pick your LLM provider (Ollama, OpenAI, or Anthropic), connect Neo4j, set your wiki output path. Creates .secrin.yml commit it so teammates inherit the config.
$ secrin graph build --repo .
Build the knowledge graph
Parses every Python, TypeScript, and JavaScript file using tree-sitter. Writes files, classes, functions, and their relationships into Neo4j as a structured, queryable graph.
$ secrin analyze
Generate summaries and embeddings
Your LLM writes a plain-English summary for every node. An embedding model converts each summary to a vector. Both live in the graph ready for hybrid semantic and structural search.
$ secrin generate
Write the wiki
Generates a full Markdown wiki: one page per module, one per business domain, an architecture overview with Mermaid diagrams. Everything cross-linked and committed to docs/wiki/.
Be first when the hosted version ships.