Find the answer
in your own files.
Search your PDFs, notes, code, and saved exports. See the matching passage and its source before you decide what to trust.
CLIOptional browser UILocal HTTP API1. Ask a question
Choose a recorded query, or run the release locally to ask your own questions.
2. Inspect the passage
Review the matching text rather than accepting a generated claim.
3. Open the source
Read the fictional document and verify what it actually says.
Explore a real retrieval run
MiniLM embeddings + hybrid retrieval. No generation provider is called.
Search can return passages that do not answer a question. These three examples are a small workflow demonstration, not a retrieval-quality benchmark.
Run the same sample on your machine
Use Python 3.10–3.12. Installation and the first embedding-model download need internet and disk space. Retrieval-only search needs no generation API key or Ollama.
git clone --branch v1.5.1 --depth 1 https://github.com/seanebones-lang/personal-RAG.git cd personal-RAG ./scripts/setup.sh --ui ./scripts/try-demo.sh
The source archive includes setup scripts and the sample; the wheel installs the application. This is a Python application, not a standalone native executable or managed cloud service.
Watch the 54-second overview
Captioned throughout, with no narration. You can follow it on mute.
The video shows the synthetic sample. For current pinned installation commands, use the guide above.
Keep your Obsidian vault. Search beyond it.
PersonalRAGVault reads Obsidian Markdown notes, selected frontmatter fields, tags, and wikilinks. Its local index can combine those notes with PDFs, code, and supported saved exports.
Use Obsidian to write and organize. Use PersonalRAGVault when you want a retrieval tool across several sources, with a CLI and local HTTP API.
This is file-based vault integration, not an Obsidian plugin or an endorsement by Obsidian. The showcase above does not test Obsidian graph behavior. Read the released Obsidian integration guide.
Choose when to generate an answer
Retrieval-only search lets you inspect matching source text directly. The installed app also supports optional local Ollama generation, or an explicitly configured external provider.
An external generation provider receives retrieved context. Generated answers and citations need review. Evaluate retrieval and generation on your own collection before relying on them.