dotnet-rag-chatlisted
Install: claude install-skill vedats/RagSkill-Dotnet
# .NET RAG Chat (Blazor + Ollama + SQLite vectors)
This skill packages a tested, working RAG chat app so a new project starts from something that already runs,
instead of from the stock `aichatweb` template, which has several bugs that took real debugging to find
(see `references/pitfalls.md`).
What the app does: users upload PDF/Markdown files → text is chunked and embedded locally → stored in a
SQLite vector DB → questions are answered by an Ollama (cloud or local) chat model that searches the docs via a
tool call and cites sources. Extras: documents page (upload/delete), document list on the start page, only
new/changed files are re-indexed, background indexing at startup, chat history saved per browser and restored,
last-N-messages window, retries for flaky embeddings, friendly error when Ollama is down.
Talk to the user in the language they're using.
## Decide the mode
1. **New app** (most common): use the scaffold script — step "Create a new app".
2. **Add to an existing ASP.NET Core / Blazor project**: run `scripts/integrate.py <project-dir>` for the
mechanical part, then do the manual steps (Program.cs, a `<script>` tag in App.razor, nav links) in
`references/integrate-existing.md`. Read that file first — the App.razor script tag is easy to miss and
without it answers silently don't render.
3. **Questions / debugging an app built from this**: `references/architecture.md` explains every file;
`references/pitfalls.md` lists known failures and their fixe