feat: add Local RAG Factory workflow (#19)

* feat: add local RAG factory showcase

* fix: use structured RAG activity commands

Author: ovvesley

generative-ai/local-rag/.gitignore

@@ -0,0 +1,2 @@
+outputs/
+

generative-ai/local-rag/README.md

@@ -0,0 +1,10 @@
+# Local RAG Factory
+
+A network-free retrieval-augmented generation pipeline. It normalizes a local
+Markdown knowledge base, chunks documents, builds a hashed TF-IDF vector index,
+retrieves evidence for a benchmark question set, and produces grounded answers
+with citations. The deterministic extractive generator is useful for CI; an
+Ollama adapter can be added without changing the workflow boundaries.
+
+Run `./run.sh`. The corpus and evaluation questions are versioned under `data/`.
+

generative-ai/local-rag/data/corpus/environments.md

@@ -0,0 +1,5 @@
+# Environments
+An AkôFlow environment describes where work may execute. Local, SSH, Slurm,
+Kubernetes and cloud targets expose capabilities through runtime resources.
+Environment versions keep old plans reproducible when infrastructure changes.
+

generative-ai/local-rag/data/corpus/planning.md

@@ -0,0 +1,5 @@
+# Planning
+A workflow defines dependencies without fixing an execution location. A plan
+maps each activity to compatible resources and estimates time, transfer cost
+and queue delays. The same workflow may compare multiple scheduling plans.
+

generative-ai/local-rag/data/corpus/provenance.md

@@ -0,0 +1,5 @@
+# Provenance
+AkôFlow records activities, generated entities and responsible agents. The
+lineage graph follows W3C PROV relationships such as used, wasGeneratedBy and
+wasAssociatedWith. Artifact hashes make outputs verifiable.
+

generative-ai/local-rag/data/questions.json

@@ -0,0 +1,6 @@
+[
+  {"question": "Why are environment versions useful?", "expected": "reproducible"},
+  {"question": "Which standard describes the lineage graph?", "expected": "W3C PROV"},
+  {"question": "What does a plan map to resources?", "expected": "activity"}
+]
+

generative-ai/local-rag/docker/Dockerfile

@@ -0,0 +1,6 @@
+FROM python:3.12-slim
+WORKDIR /app
+COPY data /app/data
+COPY scripts/pipeline.py /app/scripts/pipeline.py
+ENTRYPOINT ["python", "/app/scripts/pipeline.py"]
+

generative-ai/local-rag/environment.yaml

@@ -0,0 +1,6 @@
+id: local-rag-factory-environment
+name: Local RAG Factory environment
+driver: local
+runtimes: [{id: local, type: local, containerRuntime: docker}]
+resources: [{id: workstation, runtimeId: local, cpuCores: 4, memoryBytes: 4294967296}]
+

generative-ai/local-rag/run.sh

@@ -0,0 +1,5 @@
+#!/bin/sh
+set -eu
+root=$(CDPATH= cd -- "$(dirname -- "$0")" && pwd);rm -rf "$root/outputs"
+python3 "$root/scripts/pipeline.py" --stage all --output "$root/outputs";"$root/validate.sh"
+

generative-ai/local-rag/scope.yaml

@@ -0,0 +1,4 @@
+id: local-rag-factory-scope
+name: Local RAG Factory scope
+environmentVersionIds: [local-rag-factory-v1]
+