- From the zero base to the setup, it's really only 10 minutes. You can create it without writing a code, without drawing a flowchartAgent.
They say 2025 is."AI AgentIt's a year, but most of the smarts on the market are still in the second toy phase -- it's cool and it's of limited practical use..
A really valuable intelligence must have addressed your real pain. It must be based on your real needs and on your work flow。
For my own sake, as a media editor, the most troubling is the firm's bill of lading contract, which requires a manual search of the partner's business information and a more word-and-word examination of the terms of the contract..
So, today I useTencent CloudThe Smart Development Platform 3.0 took 10 minutes to customize a “Contract Review Assistant”. A few seconds after uploading the contract:
- Business background survey report for A
- Contract risk-point analysis
- Professional change proposal
Did you get sweaty when you heard the word "custom"? Can I do it with a white boy
It's true that in the past it was the patent of a few tech players, but now, do itAI AgentThe threshold has been lowered to the ice point
I mean, you don't have to write a line of code or build a logical node. Today, this article is a practical guide for all basic whites, and I will use the most popular language, starting with the registration number, to teach you how to build your own intelligence。
Core strengths: Multi-Agent model
A key concept is presented here: Multi-Agent. In this model, all you have to do is describe mission requirements in big words, so that the system can automatically plan the implementation path, use a variety of tools flexibly, and respond easily to complex scenarios of multi-tool calls and multiAgent synergies。
Does that sound incredible? The "breaker" of all of this is the latest online smart-body development platform 3.0。
This paper focuses on the use of the platform, where you can customize your own applications to help you restore productivity。
Next, you need a computer, a cell phone number, what you want him to do, and, above all, a patient heart to read
Information cloud intelligence development platform
Zero-threshold your "law clerk" intelligence
Believe that "contract review" is the pain of many media people, after allAs an individual organizer, it is the verification of the subject information of the other company and the review of the lengthy contract terms that is the most painful part of the business。
It takes time and effort not only to go manually to the business website for information, but also to check the contract word for word, and it is easy to step on the pit for negligence。
My task today, therefore, is to create a “law clerk” who can review contract terms and perform compliance checks。
Step 1: Register and log on to the telecommunication cloud
website: https://cloud.tencent.com/login
First of all, access to the official Internet site。
Step 2: Entering the Smart Body Development Platform
website: https://adp.cloud.tencent.com/adp
After entering the workstation, select Application Development, click New Application。
Enter the name and function description (with no effect on function if you want to enter it) and apply mode selection "Multi-Agent"。
Step 3: Add required plugins
Next, we need to add a plug-in to the smart body, which is to be equipped with a weapon。
You need to be inPlugin SquareIT LOOKS FOR PLUGINS THAT CAN HELP YOU DO YOUR JOB, NOT ONLY MCP, BUT ALSO A LOT OF SEALED APIS, SO YOU CAN CREATE THEM YOURSELF。
For example, contract risk analysis assistants, I need the following plugins:
Questions and Questions on the Knowledge Bank: A number of contract entries previously archived, legal knowledge documents uploaded to the Platform's knowledge base, are more specialized in combination with analysis。
" Document parsing " : The contract file used to interpret the uploaded PDF or Word format into the Markdown format。
“Waterdrop enterprise search”: used to verify business information and business performance of cooperating companies。
Step 4: Complementing the knowledge base
This step is to remove some of the contracts previously archived from the guidelines and to upload some of the contract notes, legal knowledge documents, so that the intelligent body combines analysis more professionally。
Supports uploading pdf, doc, ppt... with basic coverage of all common document formats, with individual files up to 200MB。
Common smart customer service, corporate knowledge bank question and answer assistants are smart body applications based on the knowledge base。
Step 5: Configure workflow (core process design)
WORKSTREAM IS A “PLAYBOOK” OF INTELLIGENT ACTION, WHICH SETS OUT EVERY STEP OF THE AI PROCESSING MISSION. IN THE PAST, THE DEVELOPMENT OF AN INTELLIGENT BODY NEEDED TO ACHIEVE WORKFLOWS BY DRAGGING THE FUNCTIONAL NODES AND SUPERB BRAIN BURN。
But in the “Multi-Agent” model, all we have to do is describe mission requirements in natural languages, so that the system can automatically plan the implementation path, have flexible access to various tools, and are very good for starters。
Select Basic Model:
The platform for the development of information cloud intelligence provides us with the latest R1, V3.1 of Yooutu-agent and DeepSeek。
If you don't get used to these models, it's okay. Go straight to the model square and define the addition
The best modelers in the country are fit to fill out a Key, even if it's compatible with OpenAI API protocols
Fill in descriptions and hints:
The description of what this Agent is for, what it's going to be used in, is a description that helps the model to determine when it should go to Agent。
The following hint is the top priority, which is the equivalent of how you're going to specify in the hint how this intelligent body will call the plugs you added。
Don't worry about not writing
You can speak your mind in white, then click on the "one-key optimization" in the top right corner, which automatically expands your needs and transforms the format into a clearer and more understandable structure。
For example, my contract risk analysis assistant, the entire workflow is as follows:
Step 1: Document parsing: Hand over the uploaded contract documents to the Document Parsing plugin for processing。
Step 2: Information extraction: extraction of the “Face of A” from the analyzed text。
Step three: Businesses do their best: Background check will be carried out by forwarding the name of the company from the extraction to the drop-down business query plugin。
Step 4: Risk analysisRisk analysis of contract terms, using large modelling capabilities embedded in the platform, in combination with legal knowledge in the knowledge base and results。
Step 5: Smart questions and answers: consolidate all information into a complete review report and wait for my inquiry。
Special attention is needed here, and in order to achieve the correct call for the plugins, we also need to be in the right spot for the hints。
Once all additions are completed, you can upload the contract documents you need to analyse in the right dialogue area。
Analysis of effects presentation:
The subject above has been desensitized
Well, isn't it great
A simple upload of a contract would allow the smart body to produce a professional report in a few seconds, including background checks on the A side, summaries of key terms and risk tips。
The whole process, I just need to add the plug-in to the workflow and tell it how to call, and an automated contract review process is complete
If you're a high-end player and want to control the process more precisely, then you can adjust the mode to Workstream. Through visual drag and pull, you can build complex business logics like building blocks and accurately control each implementation node。
One key to all micro-trust ecological platforms
When you determine that the intelligence has achieved the desired effect, you can click the " Release " button at the top right corner。
The first key is deployed to the WILP, the WILK, the Applet, and the Enterprise WIL. It has significantly reduced the operating costs of business users
Through today's exercise, I believe you've seen that building a powerful intelligent body is no longer the exclusive skill of a technocrat
With RAG and Multi-Agent (Multi-Intelligence Collaboration) technology, you can create smart assistants for various scenarios: It could be your personal financial advisor, or your travel planner, or even your content creation assistant, and more of a multi-Agent super-assist working together in the future。
Gartner, a leading global consulting firm, predicted:BY 2028, APPROXIMATELY 151 TP3T WILL BE MADE AVAILABLE TO AI ' S INTELLIGENCE BODY。 Behind that number is the coming massive productivity emancipation。
If you have any ideas, don't hesitate. Try it yourself to see if it works