Skip to content
AstrBot
Main Navigation HomeBlogRoadmapHTTP API

English

简体中文

English

简体中文

Toggle dark mode

Intro & DeployMessaging PlatformsAI IntegrationUsageDevelopment
Sidebar Navigation

Introduction

What is AstrBot

Community

FAQ

Deployment

Script Deployment

Desktop Client

One-click Launcher

Docker

Kubernetes

BT Panel

1Panel

Source Deployment

Other Deployments

CasaOS

Compshare GPU

Community-provided Deployment

Support Us

Messaging Platforms

Quick Start

QQ Official Bot

Websockets

Webhook

OneBot v11

WeCom Application

WeCom AI Bot

WeChat Official Account

Personal WeChat

Lark

DingTalk

Telegram

LINE

Slack

Mattermost

Misskey

Discord

KOOK

Satori

Connect Satori

Using server-satori

Community-provided

Matrix

VoceChat

AI Integration

✨ Model Providers

MiraRouter

NewAPI

ShengSuanYun

AIHubMix

PPIO Cloud

SiliconFlow

TokenPony

302.AI

Ollama

LMStudio

⚙️ Agent Runners

Built-in Agent Runner

Dify

Coze

Alibaba Bailian

DeerFlow

Usage

WebUI

AstrBot CLI

Built-in Commands

Plugins

Plugins

MCP

Skills

AstrBot Built-in AI

Agent Sandbox

SubAgent Orchestration

Proactive Capabilities

Web Search

SerpApi

DeepSeek Search API

Knowledge Base

Context Compression

Custom Rules

Agent Execution Mode

Unified Webhook Mode

Development

Plugin Development

🌠 Getting Started

Minimal Example

Listen to Message Events

Send Messages

Plugin Configuration

Plugin Views

Plugin Internationalization

AI

Storage

HTML to Image

Session Control

Publish Plugin

Plugin Market Specification

Versions

2026-06-27

Platform Adapter Integration

AstrBot HTTP API

API Scope–Endpoint Reference

AstrBot Configuration File

Others

Diagnostics

Self-hosted HTML to Image

Open Source Summer

OSPP 2025

On this page

DeepSeek Search API ​

DeepSeek Search API uses DeepSeek's own search engine for web search and provides a Fetch API that extracts a page as Markdown, helping AI agents find and read current web content.

This page covers DeepSeek Search API setup and usage. See Web Search for general prerequisites, provider comparisons, and troubleshooting.

Get an API key ​

  1. Open the DeepSeek Platform and create an account or sign in.
  2. Open the API Keys page, create a key, and copy it.
  3. Check your account balance and service availability. Search and Fetch are billed per request; see the official documentation for current prices and billing rules.

Configure AstrBot ​

Your conversation must use AstrBot Built-in AI, and both the chat model and its API must support tool calling. The server or container running AstrBot must be able to reach https://api.deepseek.com.

  1. Open Config in the sidebar and select the profile your conversation actually uses.
  2. Open AI → Capabilities → Web Search and turn on Enable Web Search.
  3. Set Web Search Provider to DeepSeek Search API (deepseek).
  4. Add your key to the DeepSeek API Key list. One key enables both web search and page extraction; you can also add multiple keys for rotation.
  5. Click Save Configuration at the bottom right.

These settings apply only to the selected profile. Search is configured independently of the chat model: you can use a chat model from another provider. Even if your chat model already uses DeepSeek, enter the search key separately here; you may use the same key.

Supported capabilities ​

CapabilityContent returned by AstrBot
Web search (Search)Titles, source URLs, and summaries; 10 results requested by default, with a supported range of 1–50. Available page content is included when requested by the model
Page extraction (Fetch)The URL, title, and Markdown content of a specific page, including links supplied by the user or found through search

Search returns summaries by default. For more detail, ask the model to include page content in its search request or use the extraction tool on a selected source link. The actual number of results and availability of page content depend on the search service and target website.

Use DeepSeek Search API in a conversation ​

After saving, send the bot or ChatUI an explicit request. The model decides whether to call a tool and which parameters to use. Ask it to “search first, then answer” when you want an answer based on search results.

Web search ​

text
Search for AstrBot's latest release before answering. Give the version, summarize the main changes, and include source links.

Narrow the search ​

Search supports site, language, date, and exclusion operators. Describe these requirements in your message:

text
Search GitHub for AstrBot release information published since the start of 2026. Return up to 5 results and include source links.

You can also supply an explicit search expression:

text
Use DeepSeek Search API to search for: AstrBot site:github.com lang:zh,en after:2026-01-01

You can combine site:, -site:, lang:, after:, before:, exact phrases in double quotes, and excluded terms (-keyword). Dates accept YYYY or YYYY-MM-DD. The complete search expression, including operators, must not exceed 128 characters.

Extract page content ​

text
Read this page and summarize its installation steps: https://docs.astrbot.app/en/deploy/astrbot/docker.html

Fetch reads one HTTP or HTTPS URL at a time. Login requirements, paywalls, site access policies, and pages without extractable content can prevent extraction.

Troubleshooting ​

Search reports a missing key even though the DeepSeek chat model is configured ​

Enter the DeepSeek API Key separately under AI → Capabilities → Web Search for the active profile and save. The chat model key is not automatically used for search.

Insufficient balance or rate limits ​

402 indicates insufficient account balance, and 429 indicates a rate limit. Search and Fetch have separate account limits. Multiple keys under the same account do not increase those limits, and Fetch also has a per-domain limit. See the official rate limits for current values.

Multiple keys rotate between requests. AstrBot does not immediately retry a failed request with another key; reduce the request rate before trying again after a rate limit error.

Search works, but page extraction fails ​

Try a public page that does not require a login and inspect the error under Data & Logs → Logs. A Fetch 403 can indicate a security policy rejection; 404 can indicate that no usable page content could be retrieved. See Web Search troubleshooting for general checks.

Edit this page on GitHub

Last updated:

Pager
PreviousSerpApi
NextKnowledge Base

Deployed on Rainyun Logo