RAGFlow Dataset And Retrieval

Use only the bundled scripts in scripts/.
Prefer --json so returned fields can be relayed exactly.
Follow reference.md for all user-facing output.

Use This Skill When

• - the user wants to create, list, inspect, update, or delete RAGFlow datasets
• the user wants to upload, list, update, or delete documents in a dataset
• the user wants to start parsing, stop parsing, or check parse progress
• the user wants to retrieve chunks from one or more datasets
• the user wants to list configured RAGFlow models

Core Workflow

1. 1. Resolve the target dataset or document IDs first.
2. Run the matching script from scripts/.
3. Use --json unless a script only needs a simple text response.
4. Return API fields exactly; do not guess missing details.

Common commands:

CODEBLOCK0

Guardrails

• - For any delete action, list the exact items first and require explicit user confirmation before executing.
• Delete only by explicit dataset IDs or document IDs. If the user gives names or fuzzy descriptions, resolve IDs first.
• Upload does not start parsing. Start parsing only when the user asks for it.
• INLINECODE5 returns immediately after the start request; use parse_status.py for progress.
• For progress requests, use parse_status.py on the most specific scope available:

• - If a parse status result includes progress_msg, surface it directly. For FAIL, treat it as the primary error detail.
• Use --retrieval-test only for single-dataset debugging or when the user explicitly asks for that endpoint.

Output Rules

• - Follow reference.md.
• Use tables for 3+ items when possible.
• Preserve api_error, error, message, and related fields exactly as returned.
• Never fabricate progress percentages or inferred causes.