Why Knowledge Bank?
AI assistants forget everything between sessions. The Knowledge Bank solves this by providing persistent storage that survives restarts.
How It Works
The AI can save any information as key-value pairs:
// Save a fact
await fetch("http://localhost:5000/api/knowledge", {
method: "POST",
body: JSON.stringify({
key: "user-preferences",
value: "User prefers dark mode, uses Python primarily"
})
});
// Recall later (even after restart)
const res = await fetch("http://localhost:5000/api/knowledge?key=user-preferences");
// → "User prefers dark mode, uses Python primarily"
Use Cases
- Project State: Track what phase of development the AI is in
- User Preferences: Remember coding style, language, framework choices
- Debugging Context: Store known bugs and their fixes
- Architecture Notes: Keep track of system design decisions
Anti-Hallucination
The flowork_mapping.md directive instructs the AI to use Knowledge Bank to prevent hallucination. Instead of guessing, the AI should:
1. recall_knowledge to check if it already knows something
2. If not found, research and then save_knowledge for future reference