CLI Overview
Every
rag-forgecommand at a glance.
RAG-Forge’s CLI is organized around pipeline stages. Most projects need only a handful of commands to get going: init, index, query, audit. The rest are there when you need them.
Project lifecycle
| Command | Purpose |
|---|---|
init | Scaffold a new RAG-Forge project from a template |
add | Add an optional feature module as editable source code (guardrails, caching, reranking, and more) |
parse | Preview document extraction without indexing |
chunk | Preview chunking output and stats without indexing |
index | Embed chunks and write to the vector store |
query | Run a retrieval + generation query against the indexed pipeline |
inspect | Inspect a specific indexed chunk by ID |
Evaluation and quality
| Command | Purpose |
|---|---|
audit | Run evaluation against telemetry or a golden set and score against RMM |
assess | One-shot assessment of an existing pipeline |
golden | Manage the golden question/answer set |
drift | Detect retrieval drift against a baseline |
report | Generate a pipeline health HTML report |
Operations
| Command | Purpose |
|---|---|
cost | Track LLM and embedding spend |
cache | Inspect the semantic cache |
guardrails | Run guardrail tests and PII scans |
serve | Run the MCP server |
n8n | Export a project as an n8n workflow |