Rename 'RAG instance' to 'GPU instance' for accuracy, and add a References
section crediting the LinuxCapable Fedora NVIDIA guide that mentors shared and
the RPM Fusion Howto, per review.
Signed-off-by: gtfrans2re <francoisgonothitoure@gmail.com>
Document the RPM Fusion NVIDIA driver + CUDA install used on the GPU instance
(Fedora 44 Cloud, A10G), for reproducibility and to inform the containerized
workflow. Verified: driver 610.43.03, CUDA 13.3, 24GB.
Assisted-by: Claude Opus 4.8
Signed-off-by: gtfrans2re <francoisgonothitoure@gmail.com>
Script-driven corpus staging for the packaging-guidelines RAG pipeline.
Sources declared in data/sources.yaml, fetched into data/source/ for
ramalama rag to ingest. Corpus is a gitignored build artifact.
Assisted-by: Claude Opus 4.8
Signed-off-by: gtfrans2re <francoisgonothitoure@gmail.com>
Reviewed-on: #2
Reviewed-by: dominik kawka <dominikkawka@noreply.forge.fedoraproject.org>
Reviewed-by: Justin Wheeler <jwheel@redhat.com>
Remove duplicate top-level Forgejo issue templates, keeping the standard
issue_template/ layout. Change the default RAG image from the personal Quay
namespace to quay.io/fedora/packaging-guide-ramalama across the app, CLI, and
RAG config; the personal namespace remains available via the PGR_RAG_IMAGE
override. Increase the question text area height for larger displays.
Assisted-by: Claude Opus 4.8
Signed-off-by: gtfrans2re <francoisgonothitoure@gmail.com>
Implement the output design discussed in issue #160: lead with a prioritized,
actionable summary of issues (blockers first, each naming the guideline it
violates and the fix), expandable into a full fedora-review style checklist.
The model returns structured JSON so both views render from a single set of
findings rather than needing two model calls that could disagree. The schema
forces the model to cite a guideline and propose a fix for every finding,
which is where a RAG assistant adds value over the mechanical checks that
fedora-review and rpmlint already cover.
Delivered via a CLI (scripts/review_spec.py, with --full and --json) and the
Streamlit app, sharing one output layer. The parser tolerates messy model
output, with 19 airgapped unit tests.
Assisted-by: Claude Opus 4.8
Signed-off-by: gtfrans2re <francoisgonothitoure@gmail.com>
Add the RAGAS vs DeepEval evaluation harness (eval/) with a Fedora Packaging
Guidelines question set and a tool recommendation, the RamaLama RAG config with
a packaging system prompt, Forgejo issue/PR templates, a Streamlit spec-review
app, and an exploration notebook. Addresses the remaining issue #160 deliverables.
Assisted-by: Claude Opus 4.8
Signed-off-by: gtfrans2re <francoisgonothitoure@gmail.com>
These `.pyc` files were accidentally committed alongside the docs
changes. They are build artifacts that belong in `.gitignore` (which
already excludes `__pycache__/`), not in the repository.
Assisted-by: Claude Opus 4.6 (1M context)
Signed-off-by: Justin Wheeler <jwheel@redhat.com>
The `ai-ml/packaging-guide-ramalama` repo had code and data but not any
documentation explaining how to run the pipeline, how to contribute, or
how AI agents should interact with the project. These three files fill
that gap.
`GUIDE.md` walks a user from a fresh clone through corpus staging to a
built RAG vector store and a live query — the full pipeline end to end.
It is written to eventually move to the AI/ML SIG docs site once the
project stabilizes. `CONTRIBUTING.md` codifies the PR workflow, DCO
sign-off expectations, and AI-assisted contribution disclosure per
Fedora policy. `AGENTS.md` gives coding agents the project layout,
RamaLama API constraints, and safety guardrails they need to contribute
without inventing flags or committing secrets.
All three docs reflect the current state of the data-prep pipeline and
encode review feedback: one-sentence-per-line convention, corpus
terminology defined on first use, Fedora Packaging Guidelines
acknowledged as the primary and heaviest-weight data source, and the AI
disclosure exemption scoped strictly to superficial language help.
Assisted-by: Claude Opus 4.8
Signed-off-by: gtfrans2re <francoisgonothitoure@gmail.com>
Signed-off-by: Justin Wheeler <jwheel@redhat.com>
