big-rag
Public
Project Files
QUICKSTART.md
Get up and running with the Big RAG Plugin in 5 minutes.
nomic-ai/nomic-embed-text-v1.5-GGUF)cd big-rag-plugin npm install
This will start the plugin and register it with LM Studio.
Required Settings:
/Users/yourname/Documents/MyLibrary)/Users/yourname/.lmstudio/big-rag-db)Optional Settings (use defaults for now):
Create a test directory with some documents:
Set Documents Directory to ~/test-docs in the plugin settings.
First Run: The plugin will automatically:
Need to force a rebuild? Flip Manual Reindex Trigger ON in the plugin settings. Every chat will then kick off indexing again—full rebuilds if Skip Previously Indexed Files is OFF, or incremental updates when it’s ON. Remember to switch the toggle back OFF when you’re finished so normal incremental behavior resumes.
The LLM will respond using information from your indexed documents, with citations like:
Based on your dataset:
Small Dataset (<100 files, <100MB):
Medium Dataset (100-1000 files, 100MB-1GB):
Large Dataset (1000+ files, 1GB+):
Watch the LM Studio interface for:
Solution: Lower the Affinity Threshold to 0.3-0.4
Solution: Set the Documents Directory in plugin settings
Solution:
Solution:
Now that you have the plugin running:
Happy querying! 🚀
Subsequent Runs: The plugin will:
npm run build
npm run dev
mkdir -p ~/test-docs
echo "Artificial intelligence is transforming technology." > ~/test-docs/ai.txt
echo "Machine learning is a subset of AI." > ~/test-docs/ml.txt
Based on the provided documents:
Citation 1 mentions that "Artificial intelligence is transforming technology."
Citation 2 explains that "Machine learning is a subset of AI."
[Model's response incorporating this information...]
rm -rf ~/.lmstudio/big-rag-db/*
Documents: ~/Documents/Notes
Vector Store: ~/.lmstudio/notes-db
Settings: Defaults work well
Documents: ~/Code/project/docs
Vector Store: ~/.lmstudio/project-docs-db
Chunk Size: 1024 (larger for technical content)
Documents: ~/Research/papers
Vector Store: ~/.lmstudio/research-db
Retrieval Limit: 10 (more results)
Affinity Threshold: 0.6 (higher precision)