Files
ragflow/docs/guides/agent/agent_workflow/tool_components.md
YanZhang 7f85a5776e docs: update user guide (#17358)
add model config guide word and picture
add teams guide words
2026-07-29 20:46:04 +08:00

17 KiB

sidebar_position, slug, sidebar_custom_props
sidebar_position slug sidebar_custom_props
4 /tool_components
categoryIcon
RagAiAgent

Tool Components

Tool components extend an Agent's ability to retrieve information, query data, call external systems and generate deliverables. Select tools according to the business scenario and configure only the parameters required by the workflow.

Tool Selection Suggestions

  • Use web search or extraction tools when the Agent needs public web information.
  • Use academic retrieval tools when the Agent needs papers, medical literature or research references.
  • Use SQL and financial query tools when the Agent needs structured data or market information.
  • Use HTTP Request and Email for communication with external systems.
  • Use Document Generator and Browser for content generation and automation tasks.

Web Page and Information Retrieval

Tavily Search (TavilySearch)

Tavily Search retrieves web results through Tavily. It is suitable for public web search, news search and supplementing knowledge base retrieval with external information.

Parameter Description

Parameter Type Required Default Description
query string Yes {sys.query} Search keywords or question.
topic string No general Search topic. Common values are general and news.
search_depth string No basic Search depth. Common values are basic and advanced.
top_n integer No 10 Maximum number of returned results.

Supported Parameter Values

  • topic: general, news
  • search_depth: basic, advanced

Configuration Example

{
  "query": "RAGFlow open source",
  "topic": "general",
  "search_depth": "basic"
}

Output Result

The component returns search result items, including titles, URLs, snippets and related metadata. Downstream Agent or Message components can reference the result content for answer generation.

Tavily Search

Tavily Extract (TavilyExtract)

Tavily Extract extracts content from specified web pages. It is suitable for reading known URLs, article pages, documentation pages or web pages found by a previous search step.

Parameter Description

Parameter Type Required Default Description
urls array/string Yes Empty One or more URLs to extract.
extract_depth string No basic Extraction depth. Common values are basic and advanced.
format string No markdown Output format, such as text or markdown.

Supported Parameter Values

  • extract_depth: basic, advanced
  • format: text, markdown

Configuration Example

{
  "urls": ["https://ragflow.io/docs"],
  "extract_depth": "basic",
  "format": "markdown"
}

Output Result

The component returns extracted page text and metadata. The extracted text can be passed to Agent, Text Processing or Data Operation components.

Tavily Extract

Google Search (Google)

Google Search retrieves public web results through Google. It is suitable for broad web search scenarios where a Google-compatible search capability has been configured.

Parameter Description

Parameter Type Required Default Description
query string Yes {sys.query} Search keywords.
top_n integer No 10 Maximum number of returned results.

Configuration Example

{
  "query": "RAGFlow Agent workflow"
}

Output Result

The component returns search result titles, URLs and summaries for downstream processing.

Google Search

DuckDuckGo (DuckDuckGo)

DuckDuckGo is a privacy-focused search engine component. It does not require a separate API key and can be used for general web and news retrieval.

Parameter Description

Parameter Type Required Default Description
query string Yes {sys.query} Search keywords.
channel string No general Channel exposed by tool metadata, such as general or news.
top_n integer No 10 Maximum number of returned results.

Configuration Example

{
  "query": "RAGFlow open source",
  "channel": "general"
}

Output Result

The component returns web search result items. Confirm parameter mapping in the current version before using news retrieval.

Duckduckgo

Duckduckgo

SearXNG (SearXNG)

SearXNG searches through a configured SearXNG instance. It is suitable for teams that need self-hosted or privacy-oriented meta-search.

Parameter Description

Parameter Type Required Default Description
query string Yes {sys.query} Search keywords.
top_n integer No 10 Maximum number of returned results.
base_url string Yes Empty SearXNG service address.

Configuration Example

{
  "query": "retrieval augmented generation",
  "base_url": "https://searxng.example.com"
}

Output Result

The component returns normalized search results from the configured SearXNG service.

Keenable (KeenableSearch)

Keenable Search retrieves information through the configured Keenable service. Use it when the workflow requires that search provider.

Parameter Description

Parameter Type Required Default Description
query string Yes {sys.query} Search keywords.
top_n integer No 10 Maximum number of returned results.

Configuration Example

{
  "query": "enterprise RAG deployment"
}

Output Result

The component returns search results that can be used by downstream Agent or Message components.

Keenable Search

Wikipedia (Wikipedia)

Wikipedia searches encyclopedia entries and extracts summaries. It is suitable for querying clear entities, concepts and historical events.

Parameter Description

Parameter Type Required Default Description
query string Yes {sys.query} Specific entry topic or keyword.
top_n integer No 10 Maximum number of entries.
language string No en Wikipedia language code, such as zh, en or ja.

Configuration Example

{
  "query": "culture"
}

Output Result

The component returns matching entries, summaries and related metadata.

Wikipedia

Academic Literature Retrieval

Google Scholar (GoogleScholar)

Google Scholar retrieves papers, theses, books, abstracts and other academic materials. It is suitable for preliminary literature discovery and should not replace verification of original papers and citation details.

Parameter Description

Parameter Type Required Default Description
query string Yes {sys.query} Paper topic or keywords.
top_n integer No 12 Maximum number of papers.
sort_by string No relevance Sort method: date or relevance.
year_low integer/null No null Earliest publication year.
year_high integer/null No null Latest publication year.
patents boolean No true Whether to include patents.

Configuration Example

{
  "query": "retrieval augmented generation evaluation"
}

Output Result

The component returns paper titles, authors, abstracts, links and related metadata when available.

Google Scholar

ArXiv (ArXiv)

ArXiv retrieves preprints from arXiv. It is suitable for computer science, mathematics, physics and related research scenarios.

Parameter Description

Parameter Type Required Default Description
query string Yes {sys.query} Paper topic or keywords.
top_n integer No 10 Maximum number of returned papers.

Configuration Example

{
  "query": "large language model agents"
}

Output Result

The component returns paper metadata such as title, authors, abstract, publication time and URL.

Arxiv

PubMed (PubMed)

PubMed retrieves biomedical and life science literature. It is suitable for medical, pharmaceutical and biological literature searches.

Parameter Description

Parameter Type Required Default Description
query string Yes {sys.query} Biomedical literature query.
top_n integer No 10 Maximum number of returned papers.

Configuration Example

{
  "query": "diabetes treatment guideline"
}

Output Result

The component returns PubMed records, abstracts and publication metadata when available.

Pubmed

BGPT (BGPT)

BGPT is used for literature or biomedical-related retrieval scenarios supported by the configured BGPT service.

Parameter Description

Parameter Type Required Default Description
query string Yes {sys.query} Query content.
top_n integer No 10 Maximum number of returned results.

Configuration Example

{
  "query": "gene therapy review"
}

Output Result

The component returns retrieval results from the BGPT service for downstream reference.

Bgpt

Data and Financial Queries

Execute SQL (ExeSQL)

Execute SQL runs SQL queries against configured databases. It is suitable for structured data query, report generation and business data analysis.

Parameter Description

Parameter Type Required Default Description
query string Yes Empty SQL statement or natural-language query converted into SQL by the workflow.
db_type string No Empty Database type.
database string No Empty Target database or connection name.

Supported Values

Supported database types depend on the current deployment and configured connections.

Configuration Example

{
  "query": "select count(*) from orders"
}

Output Result

The component returns query results, error information or structured rows for downstream components.

Execute SQL

Yahoo Finance (YahooFinance)

Yahoo Finance queries market and financial information. It is suitable for stocks, funds, indexes and other market data supported by the service.

Parameter Description

Parameter Type Required Default Description
ticker string Yes Empty Stock, fund or index symbol.
query string No {sys.query} Natural-language financial query.
top_n integer No 10 Maximum number of returned results.

Supported Values

Symbols and markets depend on Yahoo Finance data availability.

Configuration Example

{
  "ticker": "AAPL"
}

Output Result

The component returns financial data, market summaries or related metadata.

WenCai (WenCai)

WenCai performs financial data screening through natural-language conditions, covering stocks, indexes, funds, Hong Kong stocks, U.S. stocks, futures and other query types.

Parameter Description

Parameter Type Required Default Description
query string Yes {sys.query} Stock selection or financial condition, such as "A-shares with P/E ratio below 20".
top_n integer No 10 Maximum number of results.
query_type string No stock Query type.

Supported Query Types

stock, zhishu, fund, hkstock, usstock, threeboard, conbond, insurance, futures, lccp, foreign_exchange.

Configuration Example

{
  "query": "A-shares with P/E ratio below 20 and net profit growth above 20%"
}

Output Result

The component returns matched financial instruments and related fields. In some versions, the runtime may return an empty report if the actual WenCai call is disabled.

Communication and External Systems

Email

The Email component sends emails through configured mail service parameters. It is suitable for notification, approval, report delivery and workflow result distribution.

Parameter Description

Parameter Type Required Default Description
to string/list Yes Empty Recipient address or addresses.
subject string Yes Empty Email subject.
content string Yes Empty Email body.
attachments array No Empty Attachment variables or file references.

Configuration Example

{
  "to": "user@example.com",
  "subject": "Workflow result",
  "content": "Please check the generated result."
}

Output Result

The component returns send status and error information when sending fails.

Email

HTTP Request (Invoke)

HTTP Request calls external HTTP interfaces. It is suitable for querying order status, writing work orders, triggering third-party services and integrating business systems.

Parameter Description

Parameter Type Required Default Description
url string Yes Empty Request URL.
method string No GET HTTP method.
headers string/object No {} Request headers.
variables array No Empty Variables passed into the request.
timeout integer No 60 Request timeout in seconds.
proxy string No Empty Optional HTTP/HTTPS proxy address.
clean_html boolean No false Whether to clean HTML tags from the response.
datatype string No json Python request body type: json or formdata.
body string No Empty Raw request body supported by the Go runtime.
content_type string No Empty Content-Type in the Go runtime; POST/PUT defaults to application/json.

Supported Values

  • method: GET, POST, PUT, DELETE, PATCH
  • datatype: json, formdata

Configuration Example

{
  "url": "https://1.1.1.1/cdn-cgi/trace",
  "method": "GET",
  "timeout": 30,
  "headers": "{}",
  "proxy": "",
  "clean_html": false,
  "datatype": "json",
  "variables": []
}

Output Result

The component returns HTTP response content, status information and error details for downstream processing.

HTTP Request

Content Generation and Automation

Document Generator (DocGenerator)

Document Generator creates documents from workflow inputs and templates. It is suitable for reports, contracts, summaries and structured document output.

Parameter Description

Parameter Type Required Default Description
content string Yes Empty Document content or template-filled content.
output_format string No Empty Output file format.
filename string No Empty Generated file name.

Supported Values

Supported output formats depend on the current deployment and component configuration.

Configuration Example

{
  "content": "Generate a weekly report based on the workflow result.",
  "output_format": "docx",
  "filename": "weekly_report"
}

Output Result

The component returns generated file information and downloadable output variables.

Document Generator

Browser (Browser)

Browser performs browser-based automation tasks. It is suitable for opening pages, collecting page information and executing browser workflows that require visual interaction.

Parameter Description

Parameter Type Required Default Description
task string Yes Empty Browser automation task description.
url string No Empty Initial page URL.
timeout integer No Empty Task timeout.

Supported Values

Supported browser actions depend on the current deployment and browser automation capability.

Configuration Example

{
  "url": "https://example.com",
  "task": "Open the page and extract the main content."
}

Output Result

The component returns browser execution results, extracted content or error information.