Architecture Deep Dive

FASCLAW × NOTION

THE BRAIN BEHIND THE BOT

How a self-hosted Telegram AI agent connects to 7 Notion databases to manage Content Clinic, DVC, and Fasai's entire business workflow.

7
Notion Databases
4
Core Workflows
∞
Session Memory
Architecture

SYSTEM OVERVIEW

VPS → Telegram → Notion — the high-level signal path

📱
Telegram
@fasclaw_bot
→
🤖
FasClaw / OpenClaw
VPS: srv1494598.hstgr.cloud
→
📝
Notion API
7 Databases
Claude Sonnet 4.6 Google Workspace Brave Search Haiku Heartbeat
📧
Gmail
📅
Calendar
📁
Drive
📊
Sheets
📄
Docs
👥
People
All Google services accessed via GOG skill on the VPS
The 7 Databases

NOTION DATABASE MAP

The structured data layer that powers every FasClaw action

🧠
AI Use Case
Project ideas from Idea → Showcase. Tracks status, business unit, tools, and dependencies.
Status Pipeline Self-Relation
📜
Prompt Library
Reusable prompts, system instructions, templates — the knowledge base for all agents.
6 Types Versioned
📓
Session Log
Memory system. Every agent session logged with decisions, blockers, and next steps.
Memory Continuity
👤
Lead Database
All clients and leads with aliases, status pipeline, contacts, and relations to Content + Tasks.
Hub Entity 2 Relations
The 7 Databases

DATABASES CONTINUED

Content production, task management, and DVC calendar

🎬
Content Database
Every piece of content from Idea → Posted. Tracks editors, deadlines, channels (FSYT, CCMETA, CLIENT, etc.).
10 Channels ← Lead DB
✅
Tasks Database
Action items and deliverables. Linked to clients via Lead DB. Tracks assignees, deadlines, areas.
Action Items ← Lead DB
📅
DVC Content Calendar
DVC brand's posting schedule across 3 channels + Instagram. Tracks post type, priority, captions.
DVC Brand Standalone
Business Coverage
Content Clinic DVC Administrative FJ
All 4 business units tracked across these 7 databases — FasClaw queries the right DB based on context.
Entity Relationships

DATABASE ERD

How the 7 databases connect — Lead DB is the central hub

LOGS ALL
🧠 AI Use Case
↺ Depends On (self)
📜 Prompt Library
📓 Session Log
Memory Layer
👤 LEAD DATABASE
Central Hub
🎬 Content DB
Client → Content
✅ Tasks DB
Client → Tasks
📅 DVC Calendar
Standalone
— — → Relation █ Hub entity ↺ Self-relation ● Memory layer
Hub Entity

LEAD DB: THE CENTRAL HUB

Every client flows through Lead DB to Content and Tasks

🎬
Content DB
Status flow:
Idea → Confirmed → ✂️ Editing → 👀 รอลง → Done → Scheduled → Posted
FSYT CCMETA CLIENT
←
👤
Lead Database
Key fields:
Name, Aliases, Status, Payment Type, Source, Contact Info, Login?, Buffet, Contract End
Status Pipeline
Old Leads → New Leads → Closing → Partner / Paid / Expired
→
✅
Tasks DB
Tracks:
Name, Status, Type, Area, Subject, Assign, Deadline, Next Action
Person Area Type
Aliases in Lead DB enable the Context Router to resolve Thai nicknames to the correct client record
Memory System

SESSION LOG: HOW FASCLAW REMEMBERS

Every significant agent session is logged for cross-session continuity

📓 What Gets Logged
TitleSession Title — what was the goal
DoneWhat Got Done — concrete outputs
DecisionsKey Decisions — architectural choices
BlockedWhat Got Blocked — issues hit
NextNext Steps — carry-forward tasks
SkillsSkills Learned — for growth tracking
Tool Used Field
Distinguishes between Openclaw (FasClaw), Claude Cowork, Claude Code, or Both — enabling filtered views per agent type.
🔄 Memory Loop
1
Session Starts
FasClaw reads recent Session Logs for context
2
Work Happens
Agent performs tasks with full prior context
3
Session Ends
Session-Closer logs everything to Notion
4
Next Session Reads
Continuity achieved — no context lost
Workflow 1

CALENDAR WORKER: 📌🤖 TASKS

Google Calendar events with 📌🤖 emoji trigger automated Notion actions

📅 Google Calendar
📌🤖 emoji events
→
🤖 FasClaw
Reads & parses events
→
🔍 Context Router
Resolves client aliases
→
📝 Notion
Tasks / Content DB
↓
📓 Session Log
Logs the action
Example Input
"📌🤖 ส่งคลิปพี่เบิร์ด 3 คลิป"
Context Resolution
พี่เบิร์ด → Lead DB alias → "Bird Thongchai" record
Notion Action
Creates 3 Tasks in Tasks DB linked to Bird's Lead record
Workflow 2

TELEGRAM → NOTION QUERY

Ask a question in Telegram, get structured data back from the right database

📱 Telegram
Thai or English query
→
🤖 FasClaw
Parses intent
→
🗄️ Correct DB
Auto-routed
→
📱 Telegram
Formatted answer
"สถานะ project FasRecap?"
→ AI Use Case DB
Returns: Status, Stage, Last Updated, Next Steps from the FasRecap project page
"ลูกค้าใหม่มีใครบ้าง?"
→ Lead DB
Filters: Status = "New Leads", returns list with contact info and next steps
"content ที่กำลังตัดอยู่?"
→ Content DB
Filters: Status = "✂️ Editing", returns clips with editor name and deadline
Workflow 3

FULL LOOP: CALENDAR + NOTION + TELEGRAM

A single Telegram command triggers multi-system orchestration

"เพิ่มนัดกับพี่ปลา วันศุกร์ บ่าย 2"
1
Resolve Client
พี่ปลา → Lead DB alias lookup → "Pla Nakamura" record found
→
2
Create Event
Google Calendar → New event: "Meeting with Pla" on Friday at 14:00
→
3
Update Lead
Lead DB → Last Contact updated → Next Step set to "Meeting Friday"
→
4
Confirm
Telegram → "✅ นัดกับพี่ปลา วันศุกร์ 14:00 เรียบร้อย"
One message → Four systems updated
Telegram + Lead DB + Google Calendar + Confirmation — all in seconds
Workflow 4

SESSION AS MEMORY

How FasClaw achieves persistent memory across independent sessions

After Significant Work
FasClaw creates a Session Log entry via Notion API with all fields populated.
Next Session Starts
Agent reads recent Session Logs to restore context — blockers, decisions, and next steps carry forward.
Prompt Library Stores
Reusable prompts discovered during sessions are saved for future use — the knowledge layer grows.
Why This Matters
Without Session Log, each FasClaw conversation starts from zero. With it:

→ Decisions don't get relitigated
→ Blocked items get addressed
→ Next steps become current steps
→ Skills compound over time

Tool Used = "Openclaw" filters for FasClaw-specific sessions.
Real Examples

SESSIONS FROM THE LOG

Real entries proving the system works — each one used multiple databases

Phase 12
OpenClaw VPS Setup — Model Routing, Cron & Verification
Proved FasClaw can autonomously log sessions to Notion. Full pipeline verified: Telegram → Claude → Notion write.
Deep Dive
OpenClaw Deep Dive — Setup, Security & Platform Comparison
Compared Claude Cowork, Code Desktop, Code VS, and OpenClaw capabilities. Documented in Session Log + Prompt Library.
Research
OpenClaw & Vibe Hacking — Cybersurveillance Research
Multi-tool research session demonstrating OpenClaw's Brave Search + Notion API + visual output pipeline.
Foundation
System Improvement Sprint — Prompt Library & Dependencies
Built the foundational databases (Prompt Library, Session Log) that FasClaw now uses daily. Cross-DB relations established.
Context
Context Engineering Foundation — Phases 2-4 + Deployment
Established Lead DB ↔ Content DB and Lead DB ↔ Tasks DB relations. Context Router skill designed from this work.
Automation
n8n + Notion API Integration for Client Onboarding
Automated Content Clinic's client onboarding workflow: n8n triggers → Lead DB creation → Template pages generated.
Permissions

READ VS WRITE RULES

What FasClaw can do with each database — capabilities and constraints

Database Read Create Update Primary Use
🧠 AI Use Case ✅ Full ✅ Via skill ✅ Status, fields Project tracking & status queries
📜 Prompt Library ✅ Full ✅ New prompts ✅ Version, notes Store & retrieve reusable prompts
📓 Session Log ✅ Full ✅ After sessions ⚠️ Append only Memory persistence across sessions
👤 Lead Database ✅ Full + aliases ⚠️ With care ✅ Last Contact, Status Client lookup, alias resolution
🎬 Content DB ✅ Full ✅ New content ✅ Status, dates Content pipeline tracking
✅ Tasks DB ✅ Full ✅ New tasks ✅ Status, assignee Action items from calendar/chat
📅 DVC Calendar ✅ Full ✅ Schedule posts ✅ Status, times DVC brand posting schedule
✅ Full Unrestricted access
⚠️ Careful Requires validation before write
Knowledge Layer

AI USE CASE + PROMPT LIBRARY

The twin knowledge databases that make FasClaw smarter over time

🧠
AI Use Case
The project intelligence database
Status Pipeline
Idea → Backlog → Planned → In Progress → Done → Showcase
Self-Relation: "Depends On"
Projects can reference other projects they depend on — enabling dependency mapping
Business Segmentation
Content Clinic / DVC / Administrative / FJ — each project tagged to its business unit
Impact × Feasibility
Prioritization scoring built into each project record
📜
Prompt Library
The reusable intelligence vault
6 Prompt Types
System Prompt, Skill Instruction, Workflow Step, Template, Agent Persona, Few-Shot Example
Version Tracking
Each prompt has Version + Last Tested + Performance Notes — enabling iterative improvement
Cross-Reference
"Used By" field tracks which skills/agents consume each prompt
Business-Tagged
Prompts tagged by business unit and tool — enabling filtered retrieval
Summary

THE FULL PICTURE

7 databases, 4 workflows, persistent memory — FasClaw isn't just a chatbot. It's a business operating system accessed through Telegram.

7
Databases
4
Workflows
6+
Google Services
∞
Memory Loops
What's Next
Use this ERD as reference for FasClaw's MEMORY.md
Build automated query routing from this ERD
Add DVC Calendar to main workflow loop
←
1 / 15
→