AI Knowledge Filler: Turn any LLM into a structured file generator for Obsidian

Share & showcase
1

I built a tool that solves a specific frustration: LLMs give great answers but terrible file structure.

You ask Claude to “create a note about Docker networking” — you get a wall of prose. Then you spend 10 minutes adding YAML frontmatter, fixing headings, adding tags, linking related notes. Every time.

AI Knowledge Filler flips this. Instead of asking AI to answer questions, you ask it to generate Obsidian-ready files — with validated YAML metadata, correct heading hierarchy, WikiLinks, and domain classification. Same input, same structure, every time.

How it works

The core is a system prompt that transforms any LLM from a conversational assistant into a deterministic file generator. You load it into Claude Projects, GPT custom instructions, or use the CLI — and from that point, every response is a complete .md file ready to drop into your vault.

Example input:

Create a guide on API rate limiting

Output:

---
title: "API Rate Limiting Strategy"
type: guide
domain: api-design
level: intermediate
status: active
tags: [api, rate-limiting, performance, architecture]
related:
  - "[[API Design Principles]]"
  - "[[System Scalability Patterns]]"
created: 2026-02-20
updated: 2026-02-20
---

## Purpose
...structured content with sections, code blocks, conclusion

No manual formatting. No missing fields. Validated automatically.

What’s included

  • System prompt — works with Claude, GPT-4, Gemini, Ollama (any LLM)

  • Metadata standard — 8 file types, 30+ domains, enum validation

  • CLI — akf generate "your prompt" → file in outputs/

  • Validation — catches missing fields, wrong dates, invalid types before you import

  • Update protocol — rules for merging AI updates into existing notes without losing content

Two ways to use it

Option 1 — No setup (Claude Projects): Upload system_prompt.md to a Claude Project. Start generating files immediately. No API keys, no terminal.

Option 2 — CLI:

pip install ai-knowledge-filler
export GROQ_API_KEY="..."   # free tier
akf generate "Create Kubernetes security checklist"

Who it’s for

If you maintain a structured Obsidian vault with consistent metadata — and you use AI to help fill it — this removes the formatting overhead entirely.

Not useful if your vault is freeform notes without frontmatter.

Happy to answer questions about the metadata standard or how the system prompt works. Also curious if anyone has tried similar approaches — what broke for you?

2 Likes