rag-architect
VerifiedOpen-source profile for principal-level RAG and agent architecture.
What is rag-architect?
rag-architect is a structured agent profile that equips AI assistants to operate as senior technical leads on retrieval-augmented generation projects. It supplies persistent role instructions plus modular skills covering architecture decisions, retrieval strategies, evaluation frameworks, and roadmap translation into GitHub issues.
The profile works by loading a core identity file and specialized skill packs that prompt the agent to produce architecture decision records, namespace matrices, eval plans, and implementation-ready tickets. Users integrate it directly into Hermes or adapt the templates for other coding agents.
It targets engineering teams shipping production RAG or tool-connected agents who need consistent quality gates, latency tracking, and business-aligned tradeoffs rather than ad-hoc code suggestions.
What you can build with rag-architect
Production RAG system design
Define chunking, embedding, hybrid retrieval, and Pinecone namespace strategies while documenting tradeoffs and rollout steps.
Evaluation and observability setup
Create golden datasets, retrieval metrics, LLM-as-judge rubrics, and cost or latency monitoring specifications for ongoing regression testing.
Architecture to execution handoff
Convert high-level plans into detailed GitHub issues with acceptance criteria, agent tool specs, and rollback guidance for implementation teams.
Install rag-architect
git clone https://github.com/greynewell/rag-architect.gitgit clone https://github.com/greynewell/rag-architect.git
cd rag-architect- 1Clone the repository to your local machine.
- 2Create the target Hermes profile directory under ~/.hermes/profiles/.
- 3Copy the SOUL.md file and skills folder into the new profile.
- 4Load the rag-architect profile in your Hermes Agent session.
- 5Start a conversation by referencing one of the included skills such as production-rag-architecture.
rag-architect: pros & cons
Pros
- +Strong focus on production concerns like cost, latency, and rollback planning
- +Ready-made templates for ADRs, eval plans, and implementation issues
- +Modular skills that can be mixed or adapted beyond Hermes
- +Clear emphasis on outputs that downstream coding agents can follow
Cons
- –Tied to the Hermes Agent runtime for full profile support
- –Primarily oriented toward RAG and retrieval workflows
- –Requires manual setup and potential customization for non-standard stacks
Frequently asked questions
It is optimized for Hermes but the templates and checklists can be used independently with other agents or manually.
User reviews
Verified reviews from the community shape this listing's rating.
Loading reviews…