n8n / Case study
Company Knowledge Base Agent.
Ask in Slack. Get an answer grounded in company documents.
The problem
Why I built it
Company information can be scattered across documents, leaving people to search manually or ask the same questions again. I built a workflow that makes those documents searchable from Slack and keeps the knowledge base in step with changes in Google Drive.
The workflow
From input to handoff
- Input
A question or document change
A Slack message starts the answer path. A new or updated Google Drive file starts the indexing path.
- Process
Retrieve relevant context
Files are extracted, split, embedded, and stored in Supabase. The RAG agent searches that store when answering a question.
- Output
Reply where the question was asked
The agent formats its response and sends it back to Slack, using the indexed knowledge base as context.
Example path
A teammate asks about the blog delivery process in Slack. The agent retrieves relevant passages from the indexed SOP and returns an answer in the same channel. If that SOP is updated in Drive, the indexing path refreshes its stored content.
Engineering choice
A detail that matters
Document ingestion and question answering are separate paths. When a file changes, the workflow can remove its previous vector rows and index the updated content before the next question uses it.