Mockingbird

What if you could talk to your second brain?

Mockingbird is a local-first AI assistant that lets you have natural conversations with your personal knowledge base. Built on a RAG (Retrieval-Augmented Generation) pipeline using LangChain, ChromaDB, and Ollama, it runs entirely on your machine — no cloud, no API keys, no data leaving your device. Currently supports Markdown files with intelligent chunking that preserves heading structure, metadata-aware retrieval that understands dates, filenames, and note types, and smart routing that distinguishes between personal and technical content. Planned support for PDFs, Word documents, and plain text files.

Run

1. Ensure .env.json contains a valid VAULT path and CHAT_MODEL. (See .env-example.json)
2. Start Ollama and pull required models (for example nomic-embed-text(REQUIRED) and a chat model like qwen2.5:3b).
3. Run:

python main.py

Indexing Behavior

• The app creates chroma.db on first run.
• On later runs, it hashes markdown file content and only re-embeds files whose content changed.
• Unreadable files and embedding failures are logged and skipped instead of crashing the whole run.
• Pass --watch to keep the index fresh in the background while you chat.
• You can force rebuild at any time:

python main.py --reindex

WebUI

Currently in the test branch. Work in progress.