git官网：https://github.com/mendableai/firecrawl

文档：https://docs.firecrawl.dev/introduction

Firecrawl 是一种 API 服务，它获取 URL、对其进行抓取并将其转换为干净的 markdown。我们会抓取所有可访问的子页面，并为每个子页面提供干净的 markdown。无需站点地图。

前提条件

服务器需要较高的运行内存（部署后至少需要2G空余内存），不然执行复杂的抓取任务会报错

本地运行

1. node.js
2. pnpm
3. redis

自托管

1. Docker环境

安装步骤

本地运行

1. 下载源码

使用git或下载压缩包

git clone https://github.com/mendableai/firecrawl.git

2. 设置环境参数

在根目录中创建一个 .env，可以复制 apps/api/.env.example 中的模板

cd /firecrawl/apps/api
cp .env.example .env

默认不会设置身份验证或任何可选子服务（pdf 解析、JS 阻止支持、AI 功能）

# .env

# ===== Required ENVS ======
NUM_WORKERS_PER_QUEUE=8 
PORT=3002
HOST=0.0.0.0

#for self-hosting using docker, use redis://redis:6379. For running locally, use redis://localhost:6379
REDIS_URL=redis://redis:6379

#for self-hosting using docker, use redis://redis:6379. For running locally, use redis://localhost:6379
REDIS_RATE_LIMIT_URL=redis://redis:6379 
PLAYWRIGHT_MICROSERVICE_URL=http://playwright-service:3000/html

## To turn on DB authentication, you need to set up supabase.
USE_DB_AUTHENTICATION=false

# ===== Optional ENVS ======

# Supabase Setup (used to support DB authentication, advanced logging, etc.)
SUPABASE_ANON_TOKEN= 
SUPABASE_URL= 
SUPABASE_SERVICE_TOKEN=

# Other Optionals
# use if you've set up authentication and want to test with a real API key
TEST_API_KEY=
# set if you'd like to test the scraping rate limit
RATE_LIMIT_TEST_API_KEY_SCRAPE=
# set if you'd like to test the crawling rate limit
RATE_LIMIT_TEST_API_KEY_CRAWL=
# set if you'd like to use scraping Be to handle JS blocking
SCRAPING_BEE_API_KEY=
# add for LLM dependednt features (image alt generation, etc.)
OPENAI_API_KEY=
BULL_AUTH_KEY=@
# use if you're configuring basic logging with logtail
LOGTAIL_KEY=
# set if you have a llamaparse key you'd like to use to parse pdfs
LLAMAPARSE_API_KEY=
# set if you'd like to send slack server health status messages
SLACK_WEBHOOK_URL=
# set if you'd like to send posthog events like job logs
POSTHOG_API_KEY=
# set if you'd like to send posthog events like job logs
POSTHOG_HOST=

# set if you'd like to use the fire engine closed beta
FIRE_ENGINE_BETA_URL=

# Proxy Settings for Playwright (Alternative you can can use a proxy service like oxylabs, which rotates IPs for you on every request)
PROXY_SERVER=
PROXY_USERNAME=
PROXY_PASSWORD=
# set if you'd like to block media requests to save proxy bandwidth
BLOCK_MEDIA=

# Set this to the URL of your webhook when using the self-hosted version of FireCrawl
SELF_HOSTED_WEBHOOK_URL=

# Resend API Key for transactional emails
RESEND_API_KEY=

# LOGGING_LEVEL determines the verbosity of logs that the system will output.
# Available levels are:
# NONE - No logs will be output.
# ERROR - For logging error messages that indicate a failure in a specific operation.
# WARN - For logging potentially harmful situations that are not necessarily errors.
# INFO - For logging informational messages that highlight the progress of the application.
# DEBUG - For logging detailed information on the flow through the system, primarily used for debugging.
# TRACE - For logging more detailed information than the DEBUG level.
# Set LOGGING_LEVEL to one of the above options to control logging output.
LOGGING_LEVEL=INFO

3. 安装依赖项

使用 pnpm 安装依赖项

# cd apps/api # 确保您位于正确的文件夹中
pnpm install # 确保您拥有 pnpm 版本 9+！

4. 运行项目

您需要打开 3 个终端来运行服务。

终端 1 - 设置 redis

在项目中的任何位置运行命令

redis-server

终端 2 - 设置工作程序

现在，导航到 apps/api/ 目录并运行：

pnpm run worker

如果您要使用 llm-extract 功能](https://github.com/mendableai/firecrawl/pull/586/))，您还应该导出 OPENAI_API_KEY=sk-______

这将启动负责处理抓取作业的工作程序。

终端 3 - 设置主服务器

为此，请导航到 apps/api/ 目录。如果您尚未安装 pnpm，可以在此处安装：https://pnpm.io/installation

接下来，使用以下命令运行您的服务器：

pnpm run start

4. 测试

curl -X POST http://localhost:3002/v1/crawl \
    -H 'Content-Type: application/json' \
    -d '{
      "url": "https://docs.firecrawl.dev"
    }'

自托管

1. 下载源码

使用git或下载压缩包

git clone https://github.com/mendableai/firecrawl.git

2. 设置环境参数

在根目录中创建一个 .env，可以复制 apps/api/.env.example 中的模板

cd /firecrawl/apps/api
cp .env.example .env

默认不会设置身份验证或任何可选子服务（pdf 解析、JS 阻止支持、AI 功能）

# .env

# ===== Required ENVS ======
NUM_WORKERS_PER_QUEUE=8 
PORT=3002
HOST=0.0.0.0

#for self-hosting using docker, use redis://redis:6379. For running locally, use redis://localhost:6379
REDIS_URL=redis://redis:6379

#for self-hosting using docker, use redis://redis:6379. For running locally, use redis://localhost:6379
REDIS_RATE_LIMIT_URL=redis://redis:6379 
PLAYWRIGHT_MICROSERVICE_URL=http://playwright-service:3000/html

## To turn on DB authentication, you need to set up supabase. 
USE_DB_AUTHENTICATION=false   

# ===== Optional ENVS ======

# Supabase Setup (used to support DB authentication, advanced logging, etc.)
SUPABASE_ANON_TOKEN= 
SUPABASE_URL= 
SUPABASE_SERVICE_TOKEN=

# Other Optionals
# use if you've set up authentication and want to test with a real API key
TEST_API_KEY=
# set if you'd like to test the scraping rate limit
RATE_LIMIT_TEST_API_KEY_SCRAPE=
# set if you'd like to test the crawling rate limit
RATE_LIMIT_TEST_API_KEY_CRAWL=
# set if you'd like to use scraping Be to handle JS blocking
SCRAPING_BEE_API_KEY=
# add for LLM dependednt features (image alt generation, etc.)
OPENAI_API_KEY=
BULL_AUTH_KEY=@
# use if you're configuring basic logging with logtail
LOGTAIL_KEY=
# set if you have a llamaparse key you'd like to use to parse pdfs
LLAMAPARSE_API_KEY=
# set if you'd like to send slack server health status messages
SLACK_WEBHOOK_URL=
# set if you'd like to send posthog events like job logs
POSTHOG_API_KEY=
# set if you'd like to send posthog events like job logs
POSTHOG_HOST=

# set if you'd like to use the fire engine closed beta
FIRE_ENGINE_BETA_URL=

# Proxy Settings for Playwright (Alternative you can can use a proxy service like oxylabs, which rotates IPs for you on every request)
PROXY_SERVER=
PROXY_USERNAME=
PROXY_PASSWORD=
# set if you'd like to block media requests to save proxy bandwidth
BLOCK_MEDIA=

# Set this to the URL of your webhook when using the self-hosted version of FireCrawl
SELF_HOSTED_WEBHOOK_URL=

# Resend API Key for transactional emails
RESEND_API_KEY=

# LOGGING_LEVEL determines the verbosity of logs that the system will output.
# Available levels are:
# NONE - No logs will be output.
# ERROR - For logging error messages that indicate a failure in a specific operation.
# WARN - For logging potentially harmful situations that are not necessarily errors.
# INFO - For logging informational messages that highlight the progress of the application.
# DEBUG - For logging detailed information on the flow through the system, primarily used for debugging.
# TRACE - For logging more detailed information than the DEBUG level.
# Set LOGGING_LEVEL to one of the above options to control logging output.
LOGGING_LEVEL=INFO

一定要将USE_DB_AUTHENTICATION设置为false

3. 构建并运行 Docker 容器：

docker compose build
docker compose up -d

这将运行 Firecrawl 的本地实例，可通过 http://localhost:3002 访问。

你应该能够在 http://localhost:3002/admin/@/queues 上看到 管理端Bull Queue Manager UI。

4. 测试

curl -X POST http://localhost:3002/v1/crawl \
    -H 'Content-Type: application/json' \
    -d '{
      "url": "https://docs.firecrawl.dev"
    }'

其他

1. 开启AI功能

docker-compose.yaml中有对OpenAI的配置

因此我们可以跟换为兼容OpenAI API的其他大模型。只需要在.env文件中添加这些设置即可

OPENAI_API_KEY=xxx
OPENAI_BASE_URL=xxx
MODEL_NAME=xxx

然后重启

docker compose down
docker compose up -d