{"id":33191,"date":"2025-04-18T09:48:42","date_gmt":"2025-04-18T01:48:42","guid":{"rendered":"https:\/\/www.1ai.net\/?p=33191"},"modified":"2025-04-16T13:07:48","modified_gmt":"2025-04-16T05:07:48","slug":"%e6%89%8b%e6%8a%8a%e6%89%8b%e6%95%99%e4%bd%a0%e6%90%ad%e5%bb%ba%e6%99%ba%e8%83%bd%e5%b7%a5%e4%bd%9c%e6%b5%81%e9%ab%98%e7%ba%a7%e7%af%87%ef%bc%8cn8ndeepseek%e5%a4%a7%e6%a8%a1%e5%9e%8b%e4%bf%9d","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/33191.html","title":{"rendered":"Hands-on teaching you to build smart workflow advanced chapter, n8n + Deepseek big model nanny tutorials"},"content":{"rendered":"

Let's pick up where we left off in the last chapter, the basics.<\/p>\n

Hands-on teaching you to build the basics of intelligent workflow, n8n + Deepseek big model nanny tutorials<\/a><\/h4>\n

Moving on to the chat, recognize the AI core nodes in the foundations chapter \"AI Agent<\/a>\"as well as the Memory and Tool socket child nodes below it, know their roles and advantages and disadvantages as well as the use of scenarios, explains how to Deepseek API docking \"AI Agent\" slot, in addition to this method there is another method can be other AI model API docking \"AI Agent\" slot. Other AI model APIs can also be docked to the \"AI Agent\" slot, not only Deepseek, including all platforms that currently provide AI APIs to the outside world, including the domestic byte beanbag, Ali Tongyi Qianqian, Xunfei Xinghuo, Baidu Wenshen Yiyin, and so on. Because there are too many knowledge points, the length should not be too long, the \"AI Agent\" configuration and other AI docking methods in this chapter of the advanced chapter in detail to introduce to you.<\/p>\n

Another way to connect to Deepseek<\/strong><\/p>\n

That's right, in the base article we actually tricked the OpenAI Chat model node to successfully access Deepseek, yes! The reason we know is that Deepseek calls interface 99% which is compatible with ChatGPT API, so it is possible to use OpenAI Chat model node to call Deepseek, that's right, as long as other AIs are compatible with ChatGPT API, they can use OpenAI Chat model node.<\/p>\n

Using the HTTP Request node allows access to all the big AI models, because API is an open web calling interface, we can use the HTTP Request node to initiate a request URL call to any company or cloud platform that provides an AI API, and the necessary parameters for the request can be set directly in the HTTP Request node. It should be noted that using HTTP Request to invoke the model is more complex, and the other is that if we use HTTP Request, we can not use the n8n<\/a> built-in LangChain node. Because HTTP Requests can't be hung as child nodes below a LangChain node, this would make the Workflow<\/a> The orchestration difficulty of the n8n has increased substantially, which also limits the play of n8n in the AI field. For this reason, it is recommended to cheat the OpenAI Chat model node method, which can convert the APIs of domestic big models into OpenAI APIs to access n8n.<\/p>\n

Step 1: Drag and drop the HTTP Request node onto the canvas.<\/strong><\/p>\n

Search HTTP Request node<\/p>\n

\"c1afecedj00susnvp001rd000ly00t8m\"<\/p>\n

Drag and drop to the canvas and double-click to open the HTTP Request node panel.<\/p>\n

\"d2fc881bj00susnvp000bd000b80090m\"<\/p>\n

Step 2: Configure HTTP Request Node API Parameters<\/strong><\/p>\n

Double-click to open and start parameter configuration.<\/p>\n

\"5e6dec43j00susnvp0011d000u000ezm\"<\/p>\n

The HTTP Request node is empty for both the left data input and the right data output until it runs successfully and we start configuring the request parameters:<\/p>\n

Request method<\/strong><\/p>\n

Method: POST<\/p>\n

URL of the Deepseek API<\/p>\n

URL: https:\/\/api.siliconflow.cn\/v1\/chat\/completions<\/p>\n

Request Header Switch<\/p>\n

Send Headers: Open<\/p>\n

Request Parameters<\/p>\n

Header Parameters:<\/p>\n

\u2013 Name: Autoizations<\/p>\n

Fill in your Deepseek API-Key, separated by a space from the preceding \"Bearer\".<\/p>\n

\u2013Value: Bearer sk-4f7e92150feb4368bd677bf6f338330<\/p>\n

\u2013 Name: Content-Type<\/p>\n

\u2013Value: application\/json<\/p>\n

Body request switch<\/p>\n

Send Body: Open<\/p>\n

\u2013Body Content Type: Jason<\/p>\n

\u2013 Spectify Body: Using Jason<\/p>\n

Select expression (expression), we enter Jason with {}, \"Fixed\" only into plain text\u3002<\/p>\n

\u2013 JSON: exception<\/p>\n

\"646b4ca7j00susnvp000ud000je00asm\"<\/p>\n

{{{<\/p>\n

\"model\": \"deepseek-chat\",<\/p>\n

\"messages\":<\/p>\n

\u266a \"role\": \"system\", \"content\" \u266a \"i'm crazy, please write a poem for me\" \u266a i don't know,<\/p>\n

{\"role\": \"user\", \"content\": \"Hello!\"}<\/p>\n

],<\/p>\n

straam: false<\/p>\n

} }}<\/p>\n

Using expression in n8n must wrap the actual content inside two {{ }} (brackets), let's decipher the information in the actual requested json:<\/p>\n

\"model\": \"deepseek-chat\"<\/p>\n

The name of the Deepseek model is requested to be filled in, and the default is \"Deepseek-chat\"\u3002<\/p>\n

\"messages\":<\/strong><\/p>\n

\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u266a \"role\": \"system\", \"content\" \u266a \"i'm crazy, please write a poem for me\" \u266a i don't know,<\/strong><\/p>\n

\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 {\"role\": \"user\", \"content\": \"Hello!\"}<\/strong><\/p>\n

\u00a0 \u00a0 \u00a0 \u00a0 ],<\/strong><\/p>\n

The request parameters sent to Deepseek were the role of \"role\" and the content of the \"content\" dialogue\u3002<\/p>\n

\"messages\" is an array, which means that we can set multiple \"role\" and \"content\" pairs, wrapped in brackets\u3002<\/p>\n

The zoom button in the lower right corner of the Json box makes it easy to edit the Json in the large box, as shown below:<\/p>\n

\"3c4b66caj00susnvp000qd000u000bem\"<\/p>\n

Take a complete look at the screenshots of all the configurations:<\/p>\n

\"8682f61fj00susnvp001rd000ju0136m\"<\/p>\n

\"ada68a1aj00susnvp001yd000kw010am\"<\/p>\n

Step 3: Deepseek API Call Testing<\/strong><\/p>\n

Click on \"Test Step\" to run the HTTP Request node, and if configured correctly we will see the results of Deepseek's answer on the right side of the HTTP Request node.<\/p>\n

\"fbeb3f1dj00susnvp000ld000n20082m\"<\/p>\n

\"77122805j00susnvp000ud000u000f3m\"<\/p>\n

\"d445d827j00susnvp001bd000u000exm\"<\/p>\n

Up to this point, the HTTP Request node has successfully fetched the results returned by Deepseek for us, which is great. Now you can follow suit and utilize the HTTP Request node to request all the big model APIs.<\/p>\n

Configuring AI Agent Prompt Words<\/strong><\/p>\n

The role of the AI Agent and related knowledge was introduced in the Basics chapter, and there was no actual demonstration of how to interface with an Open Chat Model node; our cue words need to be configured in the AI Agent, which is now starting.<\/p>\n

Step 1: Deepseek Chat Model node connects to AI Agent<\/strong><\/p>\n

\"566dd4c8j00susnvp001wd000u000mdm\"<\/p>\n

First of all, connect the Spoof OpenAI Chat model node to the AI AIgent, ha.<\/p>\n

Step 2: Configure API parameters on the HTTP Request node<\/strong><\/p>\n

Double-click AI AIgent to start configuring the parameters.<\/p>\n

\"6b47aedaj00susnvp0026d000km00pmm\"<\/p>\n

Representation<\/p>\n

Agnet: Tools Agent<\/p>\n

Cue word method<\/p>\n

Prompt: Define below<\/p>\n

Enter the prompt word and select Expression.<\/p>\n

Text: Expression<\/p>\n

Note here that we haven't run the workflow and entered the data yet, i.e. we haven't started capturing AI news yet, so let's capture the news first and have the data before we can edit our cue word expressions based on the captured data.<\/p>\n

Step 3: Run the workflow<\/strong><\/p>\n

Click on \"Test workflow\" to run our workflow and get the news data.<\/p>\n

\"f7182bbaj00susnvp001wd000u000rhm\"<\/p>\n

\"931f5071j00susnvp001nd000u000ksm\"<\/p>\n

\"02610787j00susnvp001fd000u000ism\"<\/p>\n

Once done, we double click on the AI Agent node to view the results and set our prompt word. On the left is the list of captured news, we first write the prompt word into the Text box, or we can choose Expression (expression) to input the prompt word.<\/p>\n

\"c635fa6fj00susnvp001vd000u000e2m\"<\/p>\n

\"e34d90a7j00susnvp002fd000u000f6m\"<\/p>\n

Figure conveniently drag and drop with the mouse, you can also manually enter the expression in the above hit box, the key point here is that our prompt word where?<\/p>\n

\"7fba2d3dj00susnvp0029d000u000f3m\"<\/p>\n

The prompt words to be fed to Deepseek are written directly in the expression box, and Deepseek will output results based on your prompt words.<\/p>\n

\"a7f25adaj00susnvp002vd000u000h7m\"<\/p>\n

It should be noted that Deepseek default output content format is Makedown format with line breaks, this format is displayed on the web page when the format is self-contained style format for users to browse. If you do not need the Makedown format can be followed by further modifications to its, here will not go into detail.<\/p>\n

Subsequent expansion, since we have successfully captured the news content, and successfully let Deepseek in accordance with our requirements for the news processing, the results of this processing how to use the follow-up is very flexible and variable, depending on what kind of needs you have, we can continue to add other workflow nodes behind the AI Agent node, such as direct output PDF, image posters, sent to someone's mailbox. Save to the database through the website to provide users with browsing and so on, very many scenarios can be imagined.<\/p>\n

Conclusion: n8n allows us to manually adapt the required features in case the corresponding adaptation node is not officially provided.<\/strong><\/p>\n

One of the significant advantages of n8n over fully encapsulated online service-based workflow platforms like Coze is its semi-open node design. These nodes provide users with the flexibility to manually adapt and implement specific functionality in the absence of official adaptation nodes. In n8n, the Code node and the HTTP Request node are two of the most commonly used of these semi-open nodes, which provide users with powerful customization capabilities to meet a variety of complex workflow requirements.<\/p>","protected":false},"excerpt":{"rendered":"

We continue the previous chapter of the basic chapter to teach you to build intelligent workflow basic chapter, n8n + Deepseek big model nanny tutorials Then chat, in the basic chapter to recognize the AI core node \"AI Agent\" and its following Memory and Tool two socket sub-nodes, know their role and advantages and disadvantages, as well as the use of scenarios, explained how to Deepseek API to the \"AI Agent\" slot, in addition to this method there is another method can also be other AI model API docking to the \"AI Agent\" slot. In the basic chapter, we got to know the AI core node \"AI Agent\" and its two socket sub-nodes, Memory and Tool, knew their roles, strengths and weaknesses, as well as the use of scenarios, and explained how to interface the Deepseek API to the \"AI Agent\" slot, in addition to this method, there is also another method that can be used to interface other AI model API to the \"AI Agent\" slot. In addition to this method, there is another way to connect other AI model APIs to the \"AI Agent\" slot, not only Deepseek, but also all the AI APIs provided by all the current platforms, including the domestic byte beanbag, Ali Tongyi Qianqian, Xunfei Starfire, Baidu Wenshen Yiyin and so on.<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[149,144],"tags":[1481,6300,6296,6244,5145,6302,6303,6301],"collection":[],"class_list":{"0":"post-33191","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"hentry","6":"category-jiaocheng","7":"category-baike","8":"tag-ai-agent","9":"tag-ai","11":"tag-n8n","12":"tag-5145","13":"tag-6302","14":"tag-6303","15":"tag-6301"},"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/33191","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=33191"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/33191\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=33191"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=33191"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=33191"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=33191"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}