2.1 Overview¶
RAG Showcase compares seven retrieval-augmented-generation (RAG) approaches
under identical conditions — same corpus, same embedding model, and the same
generation model for the chunk-based approaches — so the dominant variable is the
retrieval-and-reasoning approach itself. Two deliberate exceptions: graph-rag
generates through LightRAG's QUERY role model, and n8n-adaptive-rag inherits the
generator of whichever approach it routes to (see
Evaluation Methodology §4).
1. How It Runs¶
Each approach is an OpenAI-compatible /rag/<name>/v1/chat/completions endpoint in a
self-contained plugin package (backend_plugins/rag/) that is bind-mounted into
Atlas's FastAPI backend through a generic plugin seam. atlas.consumer.yml
declaratively maps seven base routes and twelve flavor aliases into Atlas's
LiteLLM startup configuration, so all nineteen appear automatically as selectable
models in Open WebUI without runtime registration calls.
flowchart LR
U[Open WebUI<br/>multi-model chat] -->|/v1/chat/completions| L[LiteLLM gateway]
L --> R[RAG plugin seam<br/>backend_plugins/rag]
R --> V[vanilla-rag]
R --> H[hybrid-rag]
R --> C[contextual-rag]
R --> G[graph-rag]
R --> A[agentic-rag]
R --> N[n8n-adaptive-rag]
R --> LG[lazy-graph-rag]
V & H & C & A & LG --> W[(Weaviate)]
H & C -.->|rerank| TEI[(TEI reranker)]
G & A --> LR[(LightRAG<br/>knowledge graph)]
LG --> LC[(Persistent lazy<br/>concept-graph cache)]
N --> WF[n8n workflow<br/>classify → route]
Open a multi-model chat, select the approaches (or flavors) you want, and one prompt fans out — every answer comes back with a uniform answer, retrieved-context, and metrics footer so they are directly comparable.
2. Flavors¶
Named tuning flavors such as graph-rag-wide or hybrid-rag-high-recall can also
appear as model aliases. They route to the same base approach with reproducible
parameter overrides (retrieval depth, rerank on/off, graph query mode, agent step
budget, …). One base approach can therefore be benchmarked at several operating points
without code changes. See Flavor Tuning.
3. Fair-Comparison Guarantees¶
- One declared ingestion profile per dataset — Atlas writes the namespaced plain
collection (
RagBase_<profile>) and LightRAG graph once; the showcase derives the matching context-prefixed collection (RagContextual_<profile>) from those exact chunks. - Shared models — the chunk-based approaches generate through the same LiteLLM
model and every approach embeds with the same embedding model (
graph-rag's generator is LightRAG's QUERY role model by design); LLM roles are local-first (backend_plugins/rag/roles.yaml). - Uniform output contract — every approach returns the same answer/context/metrics shape, which the evaluation harness parses and scores.
4. Further Reading¶
- Quick Start — bring the whole stack up with one command.
- Approach Internals — the exact steps, dependencies, and knobs per approach.
- System Architecture — the full topology and per-approach flow phases.
- Evaluation & Results — how the judge panel and dataset ladder work.