About this skill

Problem addressed

When setting up RAGFlow on Windows, the friction is rarely one command; it is the sequence: Docker readiness, admin bootstrap, API key generation, model setup, dataset and document parsing, and retrieval from the right dataset. ragflow-workbench packages these operations into an executable workflow with --json output, reducing state drift caused by copying commands manually.

How the skill works

It prefers the scripts in scripts/ and returns structured fields:
- Environment readiness: it checks .env for RAGFLOW_API_URL and RAGFLOW_API_KEY, then avoids repeating install or admin bootstrap when the environment is already validated.
- Models and knowledge bases: it supports dataset creation and inspection, document upload, parse triggering, parse status polling, and default Embedding, Chat, and Rerank model management.
- Retrieval and chat: it can run search, update documents, and create chats using explicit dataset_id / document_id values, following references/output-format.md.

State tracking is explicit: parse.py starts a task, while parse_status.py reports progress; when parsing fails, progress_msg is surfaced as the primary error rather than a guessed percentage or reason.

Scope and cautions

The skill targets a local Windows + Docker RAGFlow instance and fits ops-style work from initialization through knowledge-base retrieval. Deletion operations must first list candidate objects and receive explicit user confirmation. Upload does not automatically trigger parsing. Admin bootstrap and default model setup depend on Docker container availability.

Screenshots

Use Cases

On a first Windows RAGFlow setup, check Docker readiness, bootstrap admin, and generate an API key.
After creating a knowledge base, upload policy documents, trigger parsing, and track progress or failure messages.
Before enabling QA, configure default Embedding, Chat, and Rerank models, then create a chat for retrieval.
When retrieval seems off, inspect documents in a specific dataset, update or clean candidates, then run search.

Best For

An ops engineer piloting local RAG who needs to configure Windows Docker and RAGFlow API credentials.
An ML engineer shipping a knowledge base who needs to manage document parsing, default models, and retrieval.
An application engineer integrating internal QA who needs stable dataset, document, and chat creation calls.
An IT architect evaluating RAGFlow who needs to verify installation state, model config, and knowledge base CRUD.