(Open Source) Chat2DB 3.0: AI SQL Client now with Claude3.7 & DeepSeek

--- type: transcript date: 2025-03-21 youtube_id: nivAShLBA2U --- # Transcript: (Open Source) Chat2DB 3.0: AI SQL Client now with Claude3.7 & DeepSeek In this video, I'm going to be showing you the latest updates within Chat toDB. If you're not familiar with ChatDB, it's a great platform that allows you to effectively have a natural language interface for your database. And this is regardless of whether you have an SQL database, a NoSQL database. And basically what it allows you to do is to write queries to your database to be able to create visualizations, dashboards, all the while have access to the latest and greatest state-of-the-art large language models that are out there. To get started, it is very straightforward. There is an open- source version of it that you can pull down to get started. Alternatively, they do have a hosted version that you can access within the web app or you can download it onto your machine. The first thing that I'm going to do is I'm going to connect to a new database here. And what you can do is you can go to new connection. You'll see within the list there are dozens of different databases that you can choose from. So the odds are that there will be support for the database that you are using and they're continually adding support for more and more databases over time. I'm going to set up a connection to a MySQL database. Once you've connected to your database, you'll see all of your different databases within here. You can go within each of them. You can see the tables, you can see the functions, procedures, so on and so forth. The first thing that I want to point out is on the right hand panel here, they now have this chattoDB AI assistant. This is what you can think of as your dev partner or co-pilot. And what the copilot allows you to do is to have a contextaware AI assistant that supports multi-turn dialogue making the data queries smarter and more efficient. So within here you'll see models from OpenAI. You'll see models from Alibaba. And if you go down the list, you'll see models like DeepSeek as well as models from Anthropic. All that you need to get started is just choose the database that you want to query. In this case, I can go over to data sources. I can select that connection that we just established. I can select our ERP database. And then within here, I can ask specific poignant questions. I can say, what are the 10 most recent purchases? And you will see at each step everything that it's doing. We see the SQL statement that it wrote and it's going to invoke that query. And here is the table of data here. So the other great thing with this is I can say create a bar chart based on the prices of all of the different products that we have. And you'll see again it will work through the process of everything that it needs to do to generate the response for us. Here we go. We have all of the different products. We have from the chocolate bar all the way up to the diamond necklace. We have the visualization in the style that we asked for. What you can do here is you can collapse and expand all of the different steps of what the AI assistant did on your behalf. The great thing with this, it goes without saying, is retrieving the information very quickly within seconds of whatever you might be asking, but also being able to generate these visualizations within seconds. There's a really quick feedback loop if you're trying to research your data, trying to gather insights of whatever it might be. You're going to be able to create very rich visualizations as well as generated potentially complicated SQL statements for the information that you're asking to be retrieved. The other great thing with the update here is it has the contextual awareness of everything that's within this conversation. If you want to reference pieces that are higher up within the conversation, you can do that. For instance, if I ask based on the most recent purchases, create a pie chart grouping the order status within the pie chart. If I go down, I can see that it will go and grab the data and then we have that nice pie chart. The other great thing with this is you can change the settings of the chart. For instance, if you want to change the way that the chart looks, you can go ahead and do that. And also, you can go ahead and export it as an image if you want to include this in something like a presentation or whatever it might be. All in all, it is really helpful. you'll be able to see the chain of thought as well as the chain of actions of everything that it takes. So within here, you can see the different context that it's passing into the large language model to be able to actually perform those SQL queries and ultimately generate that data with whether it's within tables or within some sort of chart that you might want to generate. All in all, it is helpful to see the different steps that the AI is taking on your behalf. The great thing with this is you can go and experiment with different models and figure out what model works best for whatever use case you're exploring. So the other great thing with this is I can say I want to connect to my Hano server that is hosted on a Cloudflare worker. I want to gather the information for the latest purchase orders to show within a dashboard. And that's another new feature of the platform is now it can actually even write code for you. You're going to be able to leverage the state-of-the-art models while also having the context of your database. It goes without saying how that can be helpful because whether you're using Java, Python, C, C++ or whatever it might be, it will have the knowledge and be able to write the relevant SQL statement for whatever you already have within your database. So, say for instance, you have a new feature from a product manager and they say, "Okay, we're going to build a dashboard based on all of the recent purchases." And what you'll have to do within the feature is write out all of the different requests that might be in the PRD or what have you. And you can go within here and specify exactly what you need. And the nice thing with this is it already has the context of how your data is structured. This part along with being able to write out all of the different relevant pieces that is going to be agnostic of the coding language is going to be really helpful because if you were to use something like chat GPT, DeepSeek, Quad, Gemini, whatever it might be, it wouldn't have the context of how your data is structured. So the nice thing with this is you're able to tie into your data and it will be able to generate the relevant queries based on the information that we have. Now, the last thing that I want to show you is the AISQL editor. So, if I go and I click create console here. So, what it will do is since we're connected to our ERP database here, what I can do is if I click forward slash, that will invoke the AI, I can select from all of the different models similar to how I just showed you within the co-pilot and say in this case I want to leverage Quinn 2.5 coder 32B. What I'll say here is I want to improve the efficiency within the supply chain. And what it will do is since it has the contextual awareness of everything that's within your database, it will go ahead and create some novel insights for us within here. If I go and I run this query here, what we'll see here if we just read through the query here is it's going to go through and it's going to look up the average lead time. We're going to see the total purchase value and we will see the ontime delivery rate. So you could imagine the time it would have taken to write out these different statements. And with the platform, it was able to generate all of this within a number of seconds here. The other great thing with this is if you begin to write out different pieces within your SQL here, it will have contextual awareness of what's within your database here, you'll be able to have all of the different autocomplete where you can just go and select everything that's within your database to be able to easily write out and reference what you have. So there's much less friction in writing out your SQL by being able to have that autocomplete and that reference of that context. Just to demonstrate how this works, if I begin to write out our query here, we'll be able to see the correct syntax as well as the data type for what I am looking up. And as you start to write this out or edit different queries, this just creates much less friction within the process here. By it having the correct contextual awareness of what we have within our database, it makes writing out these statements really easy because we'll be able to see all of the different relevant pieces. Overall, it makes writing out SQL statements really simple because you'll be able to reference all of the different data and you can just go and click through. You can see, okay, here's our table and then once we've established that, we can say order by and from there we can see the different data types within here. to say if we're going to be ordering by the date time, we can go ahead and just select that and then we can go and write out everything just like that. So last one great thing is the ability to fix SQL execution errors by simply clicking fix in chat. The AI analyzes the errors and generates the correct SQL statements, saving a lot of debugging time. They also have a table copilot. All that you need to do is input the table name and column names and it will automatically generate the SQL statements for creating the table. So if you need to modify the table, you can directly use natural language. Additionally, in this latest release, they have also implemented our diagrams making database design even more intuitive and efficient. So here is the dashboard. You can easily create the charts you need by asking for an AI assistant that saves these charts directly to your dashboard. This process makes it incredibly convenient for data analysts to monitor and present their data effectively. That's pretty much it for this video. If you found this video useful, please comment, share, and subscribe. Otherwise, until the next