thainguyen-city.vn

I think it's a very useful mechanic - to take a photo of something, upload it to the website, and ask the agent to explain what's what there. The difference from regular communication with AI here is that there is a personal information space, a shared knowledge base with the AI, coordinates, etc. That is, all this is very well suited for exploring the world around us.

Add image and photo upload and recognition features

Нужна база знаний гостиниц, хоумстеев, кемпингов, а так же туристических мест и т.п.

По сути сейчас все наполнение ложится на наши плечи. Сейчас это не больгая проблема, зато у нас полное управление присутствием и качеством информаци.

Добавить личный кабинет управления основными гео-объектами

Most of the time, the interaction process will take place precisely through the chat and the map, and this is a two-way mechanism:
1\. You can write to the agent "Add this object here", attach a description and a photograph, and along with this, the user's current coordinates or manually refined ones are sent.
2\. You can place a cursor on the map at the desired location and go through the addition procedure. In turn, you can add clarifying information to the agent.&#x20;

Add the functionality to add new objects directly from the map

As practice has shown, AI is very unstable when it comes to uncontrolled work with information. It is almost impossible to reliably control it so that the AI agent reliably performs the same actions according to the same logic in similar situations. But not involving the agent in information management is also not cool, otherwise what are they needed for.

Here I am thinking of testing a new interesting mechanic: when something needs to be updated, instead of letting the AI perform the final database update itself, it outputs a card with proposed changes into the chat. It should turn out to be quite convenient. That is, I tell it something like "Find the coordinates and other useful data for this object," the AI finds it, but instead of sending an update mutation, it simply outputs all this in the form of a separate card with proposed changes, and through a personal interface I can see what kind of entity this is and what fields are proposed for change, and simply apply all suggestions or just specific ones. It should be convenient.&#x20;

Add editing of information entities via AI chat

Now information can come through several channels, at least these:

* From me
* From the Vietnamese manager
* From the AI agent (read: any other user, just through an agent)

At least the following tasks arise here:

* Translation. That is, basic information should be stored in English (Vietnamese and others should be stored in translation). This must be embedded into the concept creation/update processors themselves so that the AI determines the language and, if it is not English, automatically performs the translation.&#x20;
* Moderation. Legislation requires a very careful approach to published information, so every concept and fact must have an approved mark.

Develop the information moderation lifecycle

I have a huge number of my own photos with geo-data (I specifically enabled camera GPS access).&#x20;

We need to work out the following scenario:

Drop a large batch of photos (100-200+) into a separate interface all at once.&#x20;

Directly in the browser, without uploading to the server, we need to filter the photos with geo-data, sort them by distance from each other, and display everything as cards with previews. At the same time, the existing list of already uploaded files needs to be loaded to the frontend so we don't process the same ones again. Basically, the uploaded list also needs to be filtered by the "Already uploaded" attribute.

Next, we select any photo and create a new concept based on it (using geo-data, a lot can be easily googled via AI, and it will also prepare a description; all that remains is to double-check and confirm or adjust it).&#x20;

All of this will be a very good starting asset for natural and other attractions.

Add batch photo processing feature

In all main concepts (geo-objects, dishes, flora/fauna, etc.), it must be possible to collect and display a set of photos.

Add photo gallery

This is especially critical for hotels, etc.

Minimum procedure:

1\. Analyze all images attached to the object (and add this information to each image)

2\. Read the current information that already exists

3\. Read the descriptions of all images

4\. Search the web for additional fresh information if available.

5\. Write comprehensive information into the concept with all important data.

Add content generation procedure for the Concept

Very often images are uploaded in PNG format, but it makes no sense because these are not interface elements, they are all opaque, and at the same time they take up several times more space. We need to convert them to JPEG right away.

We just need to check the frontend multi-uploader, since hashes are calculated right on the frontend there, so this conversion will need to be done on the frontend as well.&#x20;

Right now, all this might be unnecessary development overhead, and it will be enough to provide conversion for the frontend in the resizer + Varnish caching.