Skip to main content

Glossary - Term Explanations

This glossary explains all important terms related to the OpenAI Agents Manager plugin.

A​

Agent​

A virtual assistant with a specific task and its own knowledge. Each agent has its own instructions, tools, and settings. Examples: "Product Advisor Agent", "Customer Service Agent".

API (Application Programming Interface)​

An interface through which software programs communicate with each other. The plugin uses the OpenAI API to communicate with AI models.

API Key​

A secret code that authenticates your Shopware system and enables access to your OpenAI account. Looks like: sk-proj-abc123...

Assistant Logs​

The system's knowledge database. Questions, answers, and conversations are stored here to continuously improve agents.

B​

Backend​

The administration area of Shopware where you can create, configure, and test agents.

Backend Chat​

A chat interface in the Shopware backend for direct testing of your agents.

C​

Cached Tokens​

Tokens that were already used in previous requests and are cached by OpenAI. Cached tokens cost only a fraction of normal token costs (e.g., $0.08 instead of $0.15).

Chat History​

The complete conversation between user and agent, including all messages and tool calls.

Context​

The context of a conversation – everything the agent "knows": previous messages, instructions, tool returns, etc.

D​

Display Name​

The display name of an agent – the name users see (e.g., "Product Advisor Max").

F​

Fallback Instructions​

Instructions the agent should use when stuck or when a problem occurs. E.g.: "Apologize politely and offer to contact a staff member."

Function Call / Function Tool​

See Tool.

G​

GPT (Generative Pre-trained Transformer)​

The AI technology behind OpenAI's models (like GPT-4o, GPT-4o-mini, etc.). "Generative" means they can generate text.

I​

Init Instructions​

The initial instructions loaded at the beginning of a conversation. Often contains the greeting message.

Instructions​

Instructions that define how an agent behaves. Consist of:

  • System Instructions: Basic behavior
  • Init Instructions: Greeting/Start
  • Fallback Instructions: Error handling

J​

JSON (JavaScript Object Notation)​

A format for data transmission. Used to exchange structured data between systems.

L​

Log Entry​

A saved entry in the knowledge database (Assistant Logs) containing a question and answer.

LONGTEXT​

A large text field in the plugin configuration where you can store extensive information (e.g., FAQs, meta information).

M​

Metadata​

Additional information about a request or agent (e.g., sales channel ID, user ID).

Model​

The OpenAI AI model that the agent uses. Examples: gpt-4o-mini, gpt-4o, gpt-5.

O​

OpenAI​

The company that develops ChatGPT and the GPT models. The plugin uses OpenAI's Responses API.

Output​

The return/response of a tool or agent.

P​

Pay-as-you-go​

OpenAI's billing model: You only pay for what you use (per API request).

Prompt​

The input sent to the AI (= user message + instructions + tool results).

Prompt Caching​

OpenAI's system for caching frequently used prompts. Saves 50-90% costs.

R​

Rate Limit​

The number of API requests you can send per minute/hour. When exceeded, you must wait.

Reasoning Effort​

The "thinking depth" of a model. Values: low (fast, simple), medium (balanced), high (thorough, slow).

Responses API​

The modern OpenAI API used by the plugin. Replaced the deprecated Assistants API.

Response Format​

The format of the agent response:

  • Auto: HTML (default)
  • JSON Object: Attempts to generate JSON
  • JSON Schema: Enforces JSON structure

S​

Sales Channel​

A Shopware concept: different shops/languages/currencies within one system. Agents can be configured specific to sales channels.

System Instructions​

The main instructions that define the basic behavior of an agent. E.g.: "You are a friendly product advisor..."

T​

Tag​

A keyword for categorizing log entries. Examples: "shipping", "returns", "product".

Technical Name​

The technical name of an agent (only lowercase letters, numbers, underscores). Used in code/APIs. Example: product_advisor.

Temperature​

Setting for creativity/randomness of AI responses:

  • 0.1-0.5: Very consistent, predictable
  • 0.6-1.0: Balanced
  • 1.1-2.0: Creative, variable

Thread​

A complete conversation between user and agent. Each thread has a unique ID.

Token​

The smallest unit that OpenAI processes. One token ≈ 4 characters ≈ 0.75 words in English.

Example:

"Hello, how are you?" = approx. 6 tokens

Tool / Tool Call​

A function the agent can call to retrieve information or perform actions. Examples: product_search, get_order_status.

Top P (Nucleus Sampling)​

Alternative setting to Temperature. Controls the predictability of responses. Values: 0.0-1.0.

U​

Usage​

The number of API calls and tokens consumed. Used for cost calculation.

V​

Vector Store​

A database for documents (PDFs, text files) that the agent can use as a knowledge source. Provided by OpenAI.

Verbosity​

The detail level of agent responses. Higher verbosity = longer, more detailed responses.

Z​

Timezone​

The shop's timezone (e.g., "America/New_York"). Important for time-dependent responses.

Symbols & Abbreviations​

API​

Application Programming Interface (see above)

CMS​

Content Management System (Shopware's system for content management)

FAQ​

Frequently Asked Questions

JSON​

JavaScript Object Notation (see above)

SKU​

Stock Keeping Unit (= product number)

URL​

Uniform Resource Locator (= web address)

USD / EUR​

Currencies (US Dollar / Euro) - OpenAI bills in USD

ZIP​

ZIP Code (used e.g., in get_order_status for verification)

Commonly Confused Terms​

Agent vs. Assistant​

Same meaning in this plugin. Formerly called "Assistants" (OpenAI), now "Agents".

Tool vs. Function​

Same meaning. "Tool" is the modern term, "Function" the old one.

Thread vs. Conversation​

Nearly the same. A thread is technically the ID, conversation is the content exchange.

Prompt vs. Message​

  • Prompt = Entire input to the AI (including instructions, history, etc.)
  • Message = Individual message from the user

Next Steps​

Now you understand all important terms! Continue with:

➡️ API Reference - Technical details

➡️ Troubleshooting - Problem solutions

➡️ Back to Main Documentation