Add a local RAG skill using Ollama for document Q&A

[twishapatel12]

Feature Proposal

I’d like to propose adding a new local RAG (Retrieval-Augmented Generation) skill to OpenWork, using Ollama as the LLM backend for document-based question answering.

Motivation

A common use case for OpenWork is experimenting with workflows locally and privately. Many users want to:

• Run RAG pipelines fully locally
• Avoid reliance on external APIs
• Query private documents securely

Ollama makes local LLM usage simple and accessible, which feels aligned with OpenWork’s workflow- and skill-based approach.

Proposed Addition

Add a new skill folder, tentatively named:

rag-workflows/

This skill would include:

• SKILL.md describing the capability, requirements, and usage
• A prompt file for document Q&A
• One or more example workflows demonstrating local document retrieval + Q&A using Ollama

Scope (Initial Version)

• Local document Q&A only
• Ollama as the LLM provider
• Minimal, focused implementation
• No breaking changes to existing workflows

The goal is to keep this small and consistent with existing skill conventions.

Implementation

If this proposal sounds reasonable, I’m happy to implement it and submit a focused PR following the current repository structure and conventions.

Before starting, I wanted to confirm:

• Whether Ollama is an acceptable dependency for a skill
• Whether the proposed structure/naming makes sense

Feedback welcome — happy to adjust the approach.

[benjaminshafii]

@twishapatel12 thx for suggesting.

I'm happy to accept this change (it should not be installed by default, but easily accessible in the skills settings)

but we will add //review-2026-02-15 @twishapatel12 in the comments where it is used.

this tag will be picked up by one of our agents and will trigger a github issues. to review and see if this has been useful for you and others.

[twishapatel12]

Thanks for the clarification! That makes sense.

I’ll proceed with implementing the skill so that it’s not installed by default, but easily discoverable via the skills settings, and I’ll add the //review-2026-02-15 @twishapatel12 comment where the skill is used.

I’ll submit a focused PR once ready.