1. A personal knowledge system (Obsidian-centric)
You are building a system where:
- Obsidian is your main vault of knowledge
- LiveSync keeps it available across devices
- Content from elsewhere (LLMs, web, RSS) should flow into it
Key need: 👉 No manual copy-paste friction
2. Capture → Organise → Distill → Express pipeline
You want a full thinking workflow:
- Capture → from web, chat, RSS, notes
- Organise → structured in vault (tags, folders, metadata)
- Distill → summarise, extract insights (LLMs)
- Express → publish, share, or reuse
n8n acts as the automation backbone for this.
3. Independence from any single AI tool
You explicitly want:
- A web-based LLM interface
- Freedom to switch providers/models
- Better navigation (outline, prompt history)
So: 👉 You are avoiding lock-in and want portable intelligence
4. Structured memory for AI (vector database)
You’re aiming for:
- Long-term memory for your notes and interactions
- Ability to query past knowledge semantically
This implies: 👉 A RAG-style system (LLM + vector DB + your vault)
5. Workflow automation (n8n as orchestrator)
n8n is your engine to:
- Pull data (RSS, APIs, sheets)
- Apply logic (skip, filter, classify)
- Push outputs (vault, notifications, publishing)
Not just YouTube — but:
- News, Blogs, Email (Gmail) alerts
- Visitor tracking
- Publishing pipelines
5.1 Spreadsheet-driven control layer
Google Sheets acts as:
- A control panel for decisions
- Human-readable and editable
Example:
- Channel list
- Status (Keep / Review / Drop)
- Logic flags (skip, category)
5.2 Intelligent content filtering (YouTube use case)
You want to:
- Automatically fetch videos via RSS
- Apply per-channel logic
- Skip irrelevant sources
- Only process valuable content
5.3 Memory of consumption (state tracking)
A key pain point:
- “I don’t recall the last video I watched”
So the system must:
- Track what has been seen
- Only surface new/unseen items
- Maintain continuity over time
5.4 Reconnection and review loops
Not just passive consumption:
You want to:
- Periodically reassess channels and other media content
- Reconnect with neglected content
- Decide to keep or drop sources
5.5 Better navigation of thinking (LLM UX gap)
You identified a major limitation:
- No outline of conversations
- Hard to navigate past prompts
- Knowledge gets lost in chat history
So you want: 👉 Structured, navigable conversations integrated into your system
6 Tool for the family
Wife and son to have an option to switch from public LLM and onto self hosted solution with much higher usage limits and features not available on free public. They are not very technical and have simpler needs
In plain English
I’m building a personal system that helps me:
- Collect information from places like YouTube, news, and AI chats
- Automatically filter out what isn’t useful
- Store everything important in one place (my notes)
- Use AI to summarise and organise it
- Keep track of what I’ve already seen or learned
- And make it easier to revisit and use that knowledge later
Instead of jumping between lots of apps and losing track of things, it becomes one connected system that does most of the heavy lifting for me. :::