AI Meeting Notes Automation Pipeline Tools: n8n · Groq (Llama 3.3) · Notion API · Slack Webhooks · JavaScript
Overview Built an end-to-end automation workflow that transforms unstructured meeting notes into structured actionable reports, triggered automatically when raw text notes are added to a Notion database.
Architecture
Notion Trigger : Polls a Notion database every minute for new entries and passes page content downstream
LLM Chain (Groq/Llama 3.3) : Processes raw meeting notes via a structured prompt and returns a JSON object containing extracted summary, action items with owners and deadlines, decisions made, risks and blockers, suggestions, and next meeting agenda
JavaScript Code Node : Parses and sanitizes the LLM response, stripping markdown artifacts and handling malformed JSON input via try/catch error handling
Notion API : Creates a formatted report page in a separate database with labeled sections mapped from the parsed JSON output
Slack Webhook : Posts the structured report to a Slack channel via HTTP POST request for immediate team notification
Key Technical Decisions
Used Groq's Llama 3.3 70B (llama-3.3B-70B-versatile) model for fast, accurate structured extraction
Implemented JSON sanitization to handle real-world LLM output inconsistencies
Separated input and output Notion databases for clean data architecture
Used incoming webhooks for Slack integration to avoid OAuth complexity
Outcome Messy unstructured meeting notes are automatically transformed into a structured report in under 30 seconds with zero manual formatting required.
Video demo below : https://www.loom.com/share/7ac0b5ed34cc43ff82b014b4e297cb7e
