PLGrid PLGrid Documentation

How to use LokAI Portal chat

In this guide, you will learn how to use the Portal LokAI chat interface

Key elements of the interface

  • History and management panel (1): Archive of conversations and access to profile settings.
  • Workspace (2): Interactive chat window with support for Markdown formatting and source code.
  • Model selection (3): Model selection menu.
  • Side panel (4): The side panel includes, among others: a button to start a new conversation, conversation history, and configuration of model generation parameters.

Interfejs_lokai

Description of interface elements

Below is a summary of the key interface elements along with an explanation of how each button works.

opis

Model selection

The portal allows dynamic switching between different classes of models, matching computational power to a specific task:

Model Class Application Characteristics
Analytical (Model Pro Thinking) Complex data analysis, programming, document auditing. Deep reasoning, multi-stage fact verification.
Versatile (Model Pro) Daily communication, drafting letters, translations. Optimal balance between precision and response time.
Fast (Model Standard) Summaries, proofreading short texts, brainstorming. Lightning-fast response to simple queries.
Model Mini Conversation in Polish. Model trained to work with the Polish language.

You can change the model during an ongoing conversation to verify the same hypothesis using different AI engines.

History and Work Organization

All interactions are saved chronologically in the side panel.

  • Search: ability to search in the chat conversation history

szukaj.pl.png

  • Agent Marketplace: search engine for shared agents

marketplace.pl.png

  • Export: conversations can be shared via links. To do this, expand the options menu (1) using the three dots icon, and then select "Share" (2).

eksport.pl.png

Using prepared agents

After selecting the "Agent Marketplace" option, find the agent in the search engine and after selecting it, click "Start Chat".

marketplace_gui.pl.png

marketplace_agent.pl.png

In the chat window, you can now ask questions related to the selected field. Each of the pre-prepared agents uses documents related to the area in which it was defined.

In the presented example, the agent Wirtualny Asystent PLGrid (Beta) uses PLGrid documentation and will provide more precise information than a standard model without an agent.

plgagentchat.pl

Sharing conversations

Conversations can be shared as links or screen captures as images. To share a conversation, select the share icon. Then select:

  • Share - to generate a link to share the conversation
  • Export - allows you to share the conversation as an image

udostepnij_lub_eksportuj.pl

Multiple conversations

The multiple conversations feature is an analytical tool that allows generating responses from several different models simultaneously within a single query. Instead of testing models one by one, you can compare their "train of thought", precision, and style in real time.

This is an ideal solution for tasks where verification of information by different algorithms or the selection of the most optimal technical solution matters.

Why use Multi-Response mode?

  • Benchmarking (Comparison): Lightning-fast comparison of results from Pro Thinking and Pro class models for the same problem.
  • Creativity vs Precision: The ability to simultaneously receive a technical analysis and a creative approach to the topic.
  • Time saving: Eliminating the need to copy the same prompt to different chat windows.

How to start multiple generation?

  1. Open the model selection menu on the top bar of the interface.
  2. Select a model and press the + button next to the selection menu. The model will be added to the chat window.
  3. Select another model for comparison.
  4. Send the query.

wielokrotne_konwersacje.pl

Interface in Multi-Stream mode

When the feature is active, the workspace changes its layout:

  • Parallel streaming: You will see several response windows generated at the same time (side-by-side).
  • Independent tools: Each model can use its unique features (e.g., one can run file_search, while the other focuses on pure text analysis).
  • Individual assessment: You can rate or copy the response from the model that best handled the task.

Temporary Chat

The temporary chat feature allows you to conduct conversations that are completely separated from your main activity in the system. This is an ideal solution for users who care about keeping their history tidy and working with confidential information that does not require permanent recording.

Use temporary chats for quick experiments, ad-hoc analyses, or sensitive topics. These conversations do not clutter the side panel, are invisible in the search engine, and are permanently deleted after 30 days.

Activation of temporary mode

To start a temporary session, click the temporary mode icon in the top panel. Upon activation, the chat window will be highlighted.

czat_tymczasowy.pl

What happens to the conversation in temporary mode?

By selecting a temporary chat, the system applies special data retention and visibility rules:

  • Not in history: The conversation will not appear in the chat list in the side panel.
  • Exclusion from search: The chat content will not be indexed by the system search engine.
  • No bookmarks: Such a conversation cannot be added to favorites.
  • Automatic deletion: Data is stored in the technical database for only 30 days, after which it is permanently deleted.

Last update: July 30, 2026