Automation Complete 2025
n8n Morning Briefing
Daily briefing workflow automation
The problem
Starting the day means checking email, calendar, and notes separately and manually piecing together what actually matters - a routine that's automatable but usually isn't, because setting it up across multiple providers and a fallback-safe LLM call is enough friction that people don't bother.
The approach
Built a Dockerized, env-configurable n8n workflow suite that pulls overnight email (Gmail or Outlook via Microsoft Graph), calendar events, and Obsidian notes, and synthesizes them into a single personalized morning briefing, with LLM provider fallback (OpenAI to Anthropic to local) so the workflow keeps working if any one provider is down or rate-limited.
This is an n8n workflow suite that generates a personalized morning briefing by pulling together everything that happened overnight: email, calendar, and notes, synthesized into one digest instead of three separate checks.
sources
- Email - overnight messages from Gmail or Outlook (via Microsoft Graph), filtered and summarized.
- Calendar - the day’s schedule, pulled alongside email so the briefing reflects both what happened and what’s coming.
- Notes - relevant Obsidian notes surfaced for context on ongoing projects or todos.
LLM synthesis with fallback
Aggregated content is synthesized into the final briefing by an LLM call, with a fallback chain - OpenAI first, Anthropic if that fails, a local model as the last resort - so a single provider outage or rate limit doesn’t mean skipping the briefing.
{
"llm_fallback_chain": ["openai", "anthropic", "local"],
"sources": ["gmail", "ms_graph_calendar", "obsidian_vault"]
}deployment
The whole suite runs as Docker containers, configured through environment variables - API keys, source selection, fallback order - so it’s straightforward to deploy on a home server or any machine that stays on overnight.
status
Complete: the aggregation workflows, LLM fallback chain, and Dockerized deployment are all built and running on a daily cadence.
What I took away
- Provider fallback (OpenAI to Anthropic to local) is what makes a daily-cadence automation actually reliable - a single-provider outage shouldn't mean no briefing that morning.
- n8n's workflow model is a good fit for this kind of multi-source aggregation: each source (Gmail, Graph, Obsidian) is an isolated node, easy to swap or extend independently.
- Env-configurable, Dockerized deployment matters for something like this - it needs to run unattended and be trivial to redeploy across machines.
0 comments
Loading…