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. 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:

python3 scripts/datasets.py list --json
python3 scripts/datasets.py info DATASET_ID --json
python3 scripts/datasets.py create "Example Dataset" --description "Quarterly reports" --json
python3 scripts/update_dataset.py DATASET_ID --name "Updated Dataset" --json
python3 scripts/upload.py DATASET_ID /path/to/file.pdf --json
python3 scripts/upload.py list DATASET_ID --json
python3 scripts/update_document.py DATASET_ID DOC_ID --name "Updated Document" --json
python3 scripts/parse.py DATASET_ID DOC_ID1 [DOC_ID2 ...] --json
python3 scripts/stop_parse_documents.py DATASET_ID DOC_ID1 [DOC_ID2 ...] --json
python3 scripts/parse_status.py DATASET_ID --json
python3 scripts/search.py "query" --json
python3 scripts/search.py "query" DATASET_ID --json
python3 scripts/search.py --dataset-ids DATASET_ID1,DATASET_ID2 --doc-ids DOC_ID1,DOC_ID2 "query" --json
python3 scripts/search.py --retrieval-test --kb-id DATASET_ID "query" --json
python3 scripts/list_models.py --json

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.
• parse.py 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:
  • dataset specified: inspect that dataset
  • document IDs specified: pass --doc-ids
  • no dataset specified: list datasets first, then aggregate status across datasets
• 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.