A reference guide to key AI concepts and terms used in the Ajelix interface.
AI Terms
The following terms describe core AI concepts that apply to Ajelix and AI platforms in general.
Prompt
The text instruction or question you type into the chat to communicate with the AI. A well-written prompt leads to more accurate, useful responses.
LLM (Large Language Model)
The underlying AI model that reads and generates text. Ajelix lets you choose from multiple LLMs – each has different strengths in speed, reasoning, and task complexity.
AI Agent / Agentic AI
An AI that can plan, make decisions, and execute multi-step tasks autonomously. In Ajelix, Agent Mode activates this capability.
Token
The unit an LLM uses to process text – roughly 3–4 characters or one short word. Token limits determine how much text a model can read or generate in one request.
Context Window
The maximum amount of text (measured in tokens) a model can “remember” within a single conversation. Text beyond this limit is not visible to the model. Ajelix, for example, automatically summarizes earlier conversation history and carries it forward as context for each new response.
Model
A specific version of an LLM with distinct capabilities, speed, and cost characteristics. Different models are suited to different tasks – see Model Selection in the interface.
Inference
The process of an AI model generating a response based on your input. Each message you send triggers an inference.
RAG (Retrieval-Augmented Generation)
A technique where the AI retrieves relevant external information before generating a response, improving accuracy. Ajelix uses this when web search is enabled.
Hallucination
When an AI generates confident-sounding but factually incorrect information. Always verify critical outputs, especially data, dates, and statistics.
System Prompt
A hidden set of instructions that shapes how the AI behaves throughout a conversation – setting its tone, role, and rules. Ajelix has the option to create a project, wherein the user can provide instructions based on their wants.
Temperature
A setting that controls how creative or predictable an AI’s responses are. Higher temperature = more creative; lower = more focused and deterministic. Model parameters are not user-configurable. Ajelix handles configuration for each model, with settings tuned through internal testing to ensure the best performance.
Ajelix Interface Terms
The following terms describe features and concepts specific to the Ajelix interface.
Agent Mode
An advanced interaction mode where Ajelix plans and executes multi-step tasks, such as creating files, searching the web, or running code. Recommended for complex workflows.
Chat Mode
A lightweight mode for quick questions, research, or content writing that does not require multi-step processing. Responses are faster and conversational.
Asset
A self-contained visual output generated by Ajelix, such as a dashboard, webpage, diagram, or app, that renders directly in the chat and can be published or downloaded.
Branch
A copy of a conversation that starts from a specific message, letting you explore a different direction without losing the original thread.
Prompt Library
A personal library of saved prompts accessible from the chat input area. Useful for repeating tasks or standardizing workflows across sessions.
Skills
Specialized capabilities the Agent can invoke to handle domain-specific tasks – e.g., Excel processing, PDF generation, or web research. Skills are managed, developed and tested by Ajelix to ensure the security and performance of AI. Only available in Agent Mode.
Plan Approval
A step in Agent Mode where Ajelix presents its execution plan before taking any action. You can review, modify, or cancel the plan before the Agent proceeds.
Web Search Toggle
A switch in the input toolbar that grants the AI real-time internet access. Enable it when you need current information, news, or live data.
Mode Switcher
The toggle in the top toolbar that switches between Chat and Agent modes. Switch at any point in a conversation based on task complexity.
Model Selector
A dropdown in the input field for choosing which LLM powers the current conversation. Available models depend on your subscription tier.
Workspace / Sidebar
The side panel that gives access to your conversations, assets, and prompt library. Use it to navigate between sessions and manage saved outputs.
Publish (Asset)
An action available on any asset that generates a public shareable link so others can view the output without needing an Ajelix account.
File Storage
Every Ajelix user has their own private storage space. Files are not shared across accounts unless explicitly published or shared.
Project Files
Each project in Ajelix has its own dedicated file space. Files uploaded or generated within a project remain scoped to that project, keeping your work organized.
File Upload
You can upload files directly into any conversation using the attachment icon in the input area. Uploaded files are available to the AI for analysis, editing, or transformation within that session.
File Actions
From the file manager, you can upload, copy, edit, and delete files. These actions let you manage your project files without leaving the Ajelix interface.
Quick Reference
All terms at a glance.
AI Terms
| Term | Definition |
|---|---|
| Prompt | The text instruction or question you type into the chat to communicate with the AI. |
| LLM (Large Language Model) | The underlying AI model that reads and generates text. |
| AI Agent / Agentic AI | An AI that can plan, make decisions, and execute multi-step tasks autonomously. |
| Token | The unit an LLM uses to process text – roughly 3-4 characters or one short word. |
| Context Window | The maximum amount of text (measured in tokens) a model can “remember” within a single conversation. |
| Model | A specific version of an LLM with distinct capabilities, speed, and cost characteristics. |
| Inference | The process of an AI model generating a response based on your input. |
| RAG (Retrieval-Augmented Generation) | A technique where the AI retrieves relevant external information before generating a response, improving accuracy. |
| Hallucination | When an AI generates confident-sounding but factually incorrect information. |
| System Prompt | A hidden set of instructions that shapes how the AI behaves throughout a conversation – setting its tone, role, and rules. |
| Temperature | A setting that controls how creative or predictable an AI’s responses are. |
Ajelix Interface Terms
| Term | Definition |
|---|---|
| Agent Mode | An advanced interaction mode where Ajelix plans and executes multi-step tasks, such as creating files, searching the web, or running code. |
| Chat Mode | A lightweight mode for quick questions, research, or content writing that does not require multi-step processing. |
| Asset | A self-contained visual output generated by Ajelix – such as a dashboard, webpage, diagram, or app – that renders directly in the chat and can be published or downloaded. |
| Branch | A copy of a conversation that starts from a specific message, letting you explore a different direction without losing the original thread. |
| Prompt Library | A personal library of saved prompts accessible from the chat input area. |
| Skills | Specialized capabilities the Agent can invoke to handle domain-specific tasks – e.g., Excel processing, PDF generation, or web research. |
| Plan Approval | A step in Agent Mode where Ajelix presents its execution plan before taking any action. |
| Web Search Toggle | A switch in the input toolbar that grants the AI real-time internet access. |
| Mode Switcher | The toggle in the top toolbar that switches between Chat and Agent modes. |
| Model Selector | A dropdown in the input field for choosing which LLM powers the current conversation. |
| Workspace / Sidebar | The side panel that gives access to your conversations, assets, and prompt library. |
| Publish (Asset) | An action available on any asset that generates a public shareable link so others can view the output without needing an Ajelix account. |
| File Storage | Every Ajelix user has their own private storage space. |
| Project Files | Each project in Ajelix has its own dedicated file space. |
| File Upload | You can upload files directly into any conversation using the attachment icon in the input area. |
| File Actions | From the file manager, you can upload, copy, edit, and delete files. |
