AI-Powered Document Chat with Nextcloud Files using LangChain and OpenAI

Main Workflow “AI Nextcloud”

• Entry point : A public chat-trigger greets the user; every incoming chat message starts the flow.
• AI agent : A LangChain agent (“AI Nextcloud”) uses the configured OpenAI model plus short-term memory to continue the dialogue in context.
• Purpose : Answers questions about files stored in a Nextcloud folder. The user simply includes the folder path in their question.
• Tool integration : Calls the sub-workflow “Nextcloud Tool” whenever it needs to read files and pass their text back to the AI.

Sub-Workflow “Nextcloud Tool”

1. Invocation : Triggered by other workflows with the input parameter path (folder path).

2. File listing : Retrieves every file in the specified folder via the Nextcloud API.

3. Filter : Allows only readable formats (PDF, Markdown, DOCX).

4. Download & text extraction

   • PDF → Text via Extract From File
   • Markdown → Raw text
   • DOCX → Text via community node word2text

5. Aggregation : Combines all extracted text into a single output field and returns it.

Outcome : Each call yields the plain content of every supported file in a Nextcloud folder—providing rich context for the AI agent to answer user questions accurately.