Skip to main content

πŸ“š Local Knowledge Base β€” Your Words Belong to You

Many AI tools require you to upload files to the cloud before you can use "knowledge base" features. GPT AI Flow is different: your notes never leave your computer.

The Local Knowledge Base is the foundation of every intelligent feature in GPT AI Flow β€” semantic search, related note discovery, and knowledge retrieval during AI conversations all depend on it. And its core promise is simple: data is stored entirely on your local machine.

What Does It Do for You?​

You don't need to understand "vector databases" or "text embeddings." You just need to know:

Once the Local Knowledge Base is enabled, your notes become smart.

BeforeAfter
Searching notes requires exact keywordsDescribe what you need in natural language
AI chat only has general knowledgeAI can reference your own writing to answer questions
More notes = harder to manageMore notes = AI finds more connections for you
Must upload to cloud for analysisEverything stays on your computer, no internet needed

What Files Are Supported?​

File TypeSupport Level
Markdown filesβœ… Full support with intelligent chunking (split by heading hierarchy)
PDF filesβœ… Automatic text extraction
Image filesβœ… OCR text recognition (supports English and Chinese)

Intelligent Chunking​

Your notes aren't indexed as one monolithic block β€” that would make search accuracy too low. GPT AI Flow automatically splits content by your Markdown heading structure:

# My Article ← Chunk 1

Body text...

## Chapter One ← Chunk 2

Chapter one content...

### Section 1.1 ← Chunk 3

Section content...

## Chapter Two ← Chunk 4

Chapter two content...

Each chunk is indexed independently, so searches can pinpoint the exact section β€” not just tell you "it's somewhere in that article."

Privacy & Security​

This is the Local Knowledge Base's most important feature:

Your Data, Your Control
  • πŸ“ Data location: Your local computer, inside your Obsidian Vault directory
  • 🚫 No cloud uploads: Index files are never sent to any server
  • πŸ”’ No external dependencies: Indexing and searching happen entirely locally
  • πŸ—‘οΈ Delete anytime: You can delete index files at will β€” your original notes are unaffected

Many AI tools' "knowledge base" features require uploading your files to third-party servers β€” you don't know how your data will be used or whether it'll be fed into model training. GPT AI Flow chose a local-first approach from day one.

The Indexing Process​

Initial Index​

The first time you enable the Local Knowledge Base, the system scans your Obsidian Vault and indexes all supported files.

  • Indexing runs in the background β€” it won't block your normal workflow
  • Speed depends on your note count and hardware performance
  • You can check indexing progress at any time

Incremental Updates​

After the initial index, the system automatically detects file changes:

  • New notes β†’ Indexed automatically
  • Modified notes β†’ Re-indexed automatically
  • Deleted notes β†’ Removed from index automatically

No manual action required.

Resume from Interruption​

If the app closes or your computer restarts during indexing:

  • The system remembers where it left off
  • On next launch, it resumes from the checkpoint β€” no re-indexing of completed files
  • No need to start over

How It Connects to Other Features​

The Local Knowledge Base underpins multiple GPT AI Flow capabilities:

Local Knowledge Base (Index & Storage)
    β”œβ”€β”€ Semantic Search β€” Find notes using natural language
    β”œβ”€β”€ Related Note Discovery β€” Automatically recommend linked content
    β”œβ”€β”€ Forgotten Note Mining β€” Surface inspiration from old notes
    └── AI Chat RAG β€” AI references your notes when answering

Without the Local Knowledge Base, none of the above features work. That's why we recommend enabling indexing as your first step when using Pro Mode.

FAQ​

Q: How much disk space does the index use?

Index files are typically a small fraction of your original note volume. For example, 10,000 notes produce an index of roughly a few dozen MB β€” negligible for any modern computer.

Q: Does indexing affect my computer's performance?

Initial indexing uses some CPU and memory, but the system keeps resource usage within reasonable limits. Incremental updates have virtually no performance impact.

Q: Can I selectively exclude certain files?

Yes. You can configure exclusion rules in settings β€” specify folders or file types to skip during indexing.

Q: Does the index survive switching computers?

Index files are stored within your Vault directory. If you sync your entire Vault (via iCloud, Git, etc.), the index comes along. Otherwise, you'll need to re-index on the new machine β€” but thanks to the incremental mechanism, this is typically fast.

Q: Does the Local Knowledge Base need an internet connection?

Indexing and searching are completely offline. An internet connection is only needed when AI conversations call large model APIs β€” but that's the AI chat module's concern, not the knowledge base itself.

  • Product Feedback:
  • πŸ’¬ Have a question? Check out the FAQ for quick solutions: Click here

Thank you for choosing GPT AI Flow, together building the essential tools for the super individuals of the future!