How to populate the vector store for initial use? 🤔

[dolcevita007]

Hi there! 👋
First of all, thank you for building PaperCortex! I love the idea of connecting my local Paperless-ngx with AI via MCP.
I've been trying to set it up with Docker, Paperless-ngx, and Ollama, and I've made good progress:

• PaperCortex container starts successfully (HTTP mode on port 3100)
• Ollama is running and creates embeddings (nomic-embed-text + qwen2.5:14b)
• Paperless-ngx API is reachable with my token
• MCP client (Cherry Studio) can call tools like papercortex_search

But here is where Iam stuck:

  "No documents found matching... The vector store may need to be populated first"

What I habe tried:

• Running docker compose up -d as described in the README
• Setting TRANSPORT=http (so the server starts in HTTP mode)
• Ensuring the volume is mounted correctly (-v papercortex_papercortex-data:/app/data)
• Manually testing Paperless API and Ollama embeddings (both work)

My question:
Is there a step I'm missing to initially fill the vector store? The README mentions auto-indexing on startup, but I'm not seeing that happen. Should I:

• Run a specific command to seed the database?
• Trigger indexing via a special tool call?
• Or is there a different workflow for the first-time setup?

I'd really appreciate any guidance! I'm happy to test suggestions or provide more details about my setup.
Thanks again for this awesome project! 🙏

Environment (if helpful):
OS: Linux
PaperCortex: v0.1.0 (Docker)
MCP Client: Cherry Studio
Models: qwen2.5:14b, nomic-embed-text (both pulled in Ollama)