A/B testing, also called split testing, is a method for comparing the current version with a change and determining which version produces better results. It is commonly used in marketing, e-commerce, web development, and other fields to assess users' responses to changes.
In software engineering, particularly in databases and transaction management, ACID stands for Atomicity, Consistency, Isolation, and Durability. These properties are a cornerstone of databases and transaction management systems.
An AI agent is an artificial intelligence system that can respond to requests, make decisions, and perform real-world tasks. It uses a large language model to interpret input and choose how to process a request using the information and resources available to it.
Information stored so an agent can use it beyond a single request, such as task state or preferences, according to the application's storage mechanism.
Agile software development is an approach in which software is developed in short cycles, feedback is gathered, and direction is adjusted as needed instead of planning everything in advance.
Artificial intelligence is the ability of technological systems and computers to simulate aspects of human intelligence, including learning, reasoning, problem-solving, and visual or linguistic perception. Unlike software that follows a fixed set of instructions, these systems identify patterns in data and adapt to changing circumstances.
An application programming interface (API) defines how software can request information or actions from other software. In business automation, it can be used, for example, to search for a customer, update an order, or create a task without performing those actions manually on screen.
An API key is a value included with a request to a service to identify the connection, associate usage with an account, or grant access, according to the provider's rules. The key type determines what it allows and whether it must be kept secret.
An artificial neural network (ANN), also called a neural network or connectionist network, is a computational mathematical model developed in inspiration from brain or cognitive processes that occur in a natural neural network and used in machine learning. This type of network usually contains many information units (inputs and outputs) connected to one another, with connections that often pass through “hidden” information units (hidden layers).
A record of significant actions that includes the actor's identity, time, and result, so it is possible to determine who changed or approved something.
Automation uses mechanical or electronic devices to perform a planned sequence of actions without human intervention. It frees people from routine, repetitive tasks by replacing them with independent technological means.
Base64 is a data encoding method that represents binary data in text format. Each sequence of six bits is represented by one ASCII character from a set of 64 possible characters. The term Base64 originates from a particular MIME encoding.
In Meta documentation, partners are entities that deploy value-added solutions as authorized WhatsApp service providers on behalf of clients. The current official name is Solution Partner (the market still says BSP): a Meta partner that provides full platform services (messaging, billing, integration, and support) and may extend a line of credit to a client.
Business intelligence (BI) is a field of information technology concerned with building systems that help an organization extract business-relevant information from the data it collects. Business intelligence systems provide historical and current information, as well as forecasts, about business operations. They can help identify meaningful patterns for managing an organization and create reports, alerts, and signals to support management decisions.
On the WhatsApp Platform, a business portfolio lets an organization manage its Meta assets in one place. It primarily acts as a container for WhatsApp Business Accounts (WABAs), and is required to use the platform. Buyers in Israel still commonly say Business Manager.
A CAPTCHA (short for Completely Automated Public Turing test to tell Computers and Humans Apart) is a challenge-response test in computing intended to ensure that a response is not generated by a computer. Typically, the computer presenting the test asks a random question that a person can answer but that is very difficult for a computer.
In cryptography, a certificate authority (CA) is an entity that stores, signs, and issues digital certificates. A digital certificate confirms that a public key belongs to the certificate holder. The CA acts as a trusted third party between the certificate holder and the party relying on the certificate.
A chatbot is software that specializes in managing text conversations, using natural language or a limited script. A conversation can be intended for marketing, consultation, customer service, or training, and the bot can be based on a language model or predefined rules.
An automated process that checks changes and prepares or performs a deployment according to set rules, instead of relying on manual steps for every update.
The WhatsApp Business Platform Cloud API lets you programmatically send and receive WhatsApp messages and calls, including text, rich media, and interactive messages. It runs on Meta's infrastructure and does not require a business to host its own WhatsApp server.
Cloud computing is a model that provides convenient, on-demand network access to a shared pool of computing resources (processing, storage, connectivity, and services).
In n8n, the Code node lets you write JavaScript or Python code and run it as a step in a workflow. It replaced the Function and Function Item nodes starting with version 0.198.0.
CVE (short for Common Vulnerabilities and Exposures) is a standardized system for identifying and recording known security vulnerabilities, with a unique identifier assigned to each one. The system is maintained by MITRE, with funding from the United States Department of Homeland Security.
Computer vision is a major field of computer science concerned with the automatic processing of images and video to extract and interpret the visual information they contain. Practitioners develop mathematical and software tools for processing visual information.
Configuration management is a process that keeps a system consistently aligned with its requirements, design, and operational information throughout its life.
A Content Delivery Network, or CDN, is a system based on proxy servers and data centers. A CDN improves the speed and quality of data delivery to end users of websites by distributing servers geographically.
In OpenAI documentation, the context window is the model's memory limit for a single generation request: how much data it can consider at one time. It is measured in tokens (pieces of input such as text or images), and its size varies between models.
Microsoft Copilot is a generative AI chatbot developed by Microsoft AI. During 2023, Microsoft unified the Copilot brand across its chatbot products (including Bing Chat and Microsoft 365), as an assistant in Windows, Edge, and Microsoft 365.
Customer relationship management is a field focused on customer service and understanding customer needs through information tools. It includes methodologies and software that help an organization manage customer relationships: acquisition, retention, and customer base development.
Cron is a tool for scheduling jobs on Unix-like operating systems. It is commonly used to schedule tasks, commands, or scripts at fixed intervals or dates, so it is especially common in system maintenance and administration.
XSS (short for Cross-site scripting) is a cyberattack against a web user that exploits a vulnerability in a web application. It allows an attacker to inject a malicious script intended to run in a user's browser and perform actions on the user's behalf on the site.
In digital communications and computer science, a hash function is a function that converts input of any length into output of a fixed length, usually much shorter. It is undesirable for a hash function to produce identical output for different inputs, but this is sometimes unavoidable. Hash functions are therefore measured by the probability of producing identical output.
CSRF (short for Cross Site Request Forgery) is an attack against a web user that forces the user's browser to perform unintended actions in web applications on the user's behalf by exploiting limitations in the HTTP protocol.
CSV is a way to store tables as plain text: each record is on a line, and fields are separated by a character (usually a comma). It is commonly used to transfer information between spreadsheets, databases, and automations.
On the WhatsApp Platform, the customer service window is the period after a user contacts a business during which non-template messages can be sent. Outside this window, businesses generally send template messages, which are subject to sending limits.
Data mining is the use of statistical tools to find correlations and trends in large amounts of information from different sources, even when the information is not organized as a database.
A database is a means of storing data in an organized way on a computer for retrieval and processing. It is stored on a data storage medium, usually a hard drive, that allows direct access to the data.
Deep learning is a set of machine learning methods based on artificial neural networks. “Deep” refers to learning through layered structures that progressively extract features from raw data.
In data communications, a denial-of-service attack (DoS) or distributed denial-of-service attack (DDoS) is a family of attacks intended to disable a computer system by placing an unusual load on its resources. DoS describes an attack originating from a single computer, while DDoS describes an attack originating from multiple synchronized computers.
A digital signature is a cryptographic method intended to verify the authenticity and integrity of a digital message or document. A valid digital signature indicates with a high degree of confidence that the message or document has not been corrupted or forged: it was created or sent by a known party and was not changed by a third party after signing, or otherwise tampered with.
DomainKeys Identified Mail, commonly known by the acronym DKIM, is an email authentication protocol developed to identify spoofing and impersonation. It does this by signing every outgoing email message. The public signing key is published in the Domain Name System (DNS).
DNS is a protocol that lets network endpoints translate human-friendly domain names into numeric IP addresses. It also provides other name-based services, such as listing mail servers.
Docker is open-source software for running applications in an isolated virtual environment called a container. Each container includes its own software, libraries, and configuration files, so the application runs consistently without interfering with other applications.
Embeddings are numerical representations of data using vectors. AI uses them to interpret complex data and relationships, and vector databases store them for retrieval.
Encryption is a cryptographic process for encoding information that converts its original representation, known as plaintext, into an alternative form known as ciphertext. Encryption systems generally use a pseudorandom encryption key generated by an algorithm.
An enterprise resource planning system is an administrative information system that handles an organization's needs (human resources, finance, marketing, logistics, and operations) as a whole. It unifies systems through a consistent interface and shared information.
Data transfer: extract data from a source, transform it into a consistent format, and load it into a destination (such as a spreadsheet, warehouse, or CRM).
According to OpenAI, evaluations (sometimes called evals) check a model's output against predefined style and content criteria. Writing evals is an essential part of building reliable LLM applications, especially when upgrading a model or trying a new one.
Organizing a process so that a change or event triggers responses in other systems, instead of one component managing everything through a continuous sequence of calls.
In n8n, an execution is one run of a workflow. It can be started manually during testing or automatically in production through a trigger, schedule, or polling.
An expert system is a computerized information system that performs the role of a human expert, or assists one, in a specific field of knowledge. It is a computer program that represents knowledge, draws conclusions from it, and can explain how it reached those conclusions.
Fine-tuning is the process of adapting a trained model for another task, usually a more focused one. Additional training on new data is applied to parameters that have already been trained, across the whole network or some of its layers.
In data communications, a firewall is a system for monitoring and blocking unwanted connections to a communications network or individual computer. At its core, a firewall is a "smart" router connecting a secured network to networks whose security cannot be guaranteed, such as the internet.
A freelancer is a self-employed professional who provides services for predefined tasks to multiple clients, with control over how the work is done. There is no employer-employee relationship between the freelancer and the client, and the freelancer chooses clients, workload, and payment terms.
Freemium is a business model in which the product's basic tier is free, while advanced features are sold separately. It is a common model for digital tools.
In OpenAI documentation, function calling (also called tool calling) is a mechanism that lets a model connect to external systems and access data outside its training data. The model returns a tool call with arguments, and the application runs the function and returns the result to the model.
The GDPR is the European Union's General Data Protection Regulation. It governs the processing of personal data and people's rights. Whether it applies, and what requirements apply to a particular process, depends on the activities and circumstances.
Generative artificial intelligence is a term for systems that produce new content (text, images, code, and more) instead of only retrieving existing information.
GPT is the name of a family of OpenAI language models that can process and generate content from input and instructions. Capabilities, pricing, and usage limits vary by model and access method.
A graph database stores information in a graph structure: nodes represent entities, edges represent the relationships between them, and properties describe both. This structure is especially useful for queries about relationships, such as a customer network or knowledge graph.
GraphQL is an open-source query language and data manipulation language for APIs at runtime. It was developed at Facebook in 2012 and released in 2015, and lets a client request exactly the fields it needs.
gRPC (short for gRPC Remote Procedure Calls) is an open-source, cross-platform software framework for RPC. The gRPC software layer was originally created by Google. It uses HTTP/2 for data transport and Protocol Buffers as its interface description language.
According to NVIDIA NeMo, guardrails are a programmable control layer around LLM applications. They block, modify, or validate user input and model output that is unsafe, off-topic, malicious, or policy-violating.
In artificial intelligence, a hallucination occurs when a large language model mistakenly perceives patterns or objects that do not exist. In the context of RAG, an ungrounded answer includes a guess or hallucination unsupported by the source documents.
A helpdesk is a department or person that provides assistance and information, usually for computer or electronics issues. Research in this field also emphasizes the helpdesk's unique position: daily conversations with customers or employees about problems and how to use systems.
Human-in-the-loop (HITL) means incorporating a human decision, review, or approval into an automated process. In business automation, a process can prepare an action, wait for a person, and continue, change, or stop only after that person's decision.
HTTP is a protocol for transferring HTML pages and objects over the web. Communication begins with an exchange between a client and a server, from which requests such as GET or POST follow. REST calls, webhooks, and most cloud APIs are based on these requests.
Infrastructure as a service is a cloud service in which users receive computing resources for their use: they can install virtual machines, operating systems, and applications, allocate storage, and set backup and scaling policies.
Idempotency is a property of an action where repeating the same request does not add an intended effect beyond the first execution. In automation, it is used to prevent a retry from creating the same business action again.
In Make, incomplete executions are a safety mechanism that saves a scenario run that failed because of an error, so the scenario does not stop and data is not lost. The saved run can be handled automatically or manually. The mechanism is disabled by default.
Infrastructure as code (IaC) is the process of managing and provisioning computing data-center resources using machine-readable configuration files rather than physical hardware configuration or interactive configuration tools. The IT infrastructure managed through this process includes both physical equipment, such as bare-metal servers, and virtual machines, along with related configuration resources.
iPaaS is a suite of cloud services that lets customers develop, run, and manage integration processes between different applications. Customers build the connections without installing or managing hardware or middleware themselves.
ISO/IEC 27001 (commonly called ISO 27001) is an international standard for information security management. It was first published by the International Organization for Standardization (ISO) and the International Electrotechnical Commission (IEC) in 2005, then updated in 2013 and again in 2022.
ISO/IEC 27701 is an international standard that extends ISO/IEC 27001 with privacy requirements. Its purpose is to establish and maintain a privacy information management system on top of an existing information security management system.
The Privacy Protection Law, 5741-1981, is Israel's principal law governing the right to privacy. It also covers databases and the use of personal information, so it is relevant to automation processes that collect or transfer customer details. The scope of obligations in each case depends on the circumstances.
IVR systems are automated telephony systems that interact with callers, collect information, and route calls to the appropriate destination. They work with speech recognition and DTMF input from the phone keypad.
JSON is a standard file and data interchange format that uses human-readable text to store and transfer key-value data structures. It is commonly used to exchange information between a server and a client in automations and APIs.
JWT (JSON Web Token) is a standard for transferring signed information between two parties in JSON format. A common use is as an access token for authentication: after the server verifies the client, it issues a token, and the client presents it without sending a password again.
A knowledge graph is a structured representation of knowledge in graph form. Entities such as people, places, organizations, concepts, or events are represented as nodes, and the relationships between them as labeled edges. A knowledge graph aims to make facts and relationships explicit in a form that both people and computer systems can process.
Kubernetes, sometimes abbreviated as k8s and pronounced "koo-ber-NET-eez," is an open-source system for managing and automatically deploying applications in containers. It was originally designed by Google and is now maintained by the Cloud Native Computing Foundation. Kubernetes supports a range of container tools, including containerd and rkt.
A landing page is a single web page visitors arrive at from an advertising link. Its purpose is to focus visitors on one message and prompt an action: submitting contact details (a lead) or making a purchase.
Latency is the time from when a process is initiated until it actually begins to run. In a conversation with an agent or an API call, it is the time elapsed before a response is received.
LDAP is an open protocol for accessing and managing directory services over IP networks. It is used to share information about users, systems, and services over a network, mainly within internal networks.
A large language model is a deep learning architecture for processing, predicting, and generating human language. It is trained on large amounts of text, has many parameters, and uses an attention mechanism that enables it to understand context.
In computer science, load balancing is the distribution of work among multiple computers to reduce workload. In computing, load balancing improves the distribution of workloads across computing resources such as computers, computer clusters, network links, central processing units, or disk drives.
A low-code development platform provides an environment, usually a graphical interface, where little or no code is written. It can generate a working application or require additional code for specific cases, shortening delivery time.
Make (formerly Integromat) is an automation platform that connects apps and services to automate processes and tasks, including complex end-to-end workflows, without requiring coding expertise.
MCP (Model Context Protocol) is an open standard for connecting AI applications to external systems: data sources, tools, and workflows. The official documentation compares it to a USB-C port for AI.
Templates are messages that can be prepared in advance. Approval is usually required before they can be sent. This is the type of message that can be sent to WhatsApp users outside the customer service window, and it is subject to quality ratings and sending limits.
Meta Platforms (trade name Meta) is an American technology company that operates Facebook, Instagram, WhatsApp, and Messenger, among other services. The WhatsApp Business Platform runs on Meta infrastructure, including Graph API, business portfolios, and Cloud API.
Multi-factor authentication is electronic authentication that uses two or more methods to grant access to an account or system. Its purpose is to provide protection even if one authentication method fails.
Middleware is computer software that connects software components or applications. It is generally used to develop and run complex, distributed applications.
Machine learning is a subfield of computer science and artificial intelligence focused on algorithms that enable computers to learn from examples. It is used for tasks where conventional programming with predefined rules is not feasible.
MQTT is a lightweight protocol for transferring messages between devices over a network. It suits environments with limited bandwidth, such as IoT and cellular networks, because it uses few resources.
n8n is an automation platform for workflows and AI agents: build visually, add code when needed, and connect to hundreds of integrations. It can run in its cloud or on your own infrastructure.
Named-entity recognition is a natural language processing task that extracts names of entities from text and classifies them into categories. These entities are usually people, organizations, locations, dates, quantities, monetary values, percentages, and more.
Natural language processing is a subfield of artificial intelligence and linguistics. It focuses on processing natural language and understanding human text or speech so computers can handle things said or written in human languages.
No-code development means building applications or processes through a visual interface and configuration, instead of writing most of the integration in code. Advanced capabilities and customizations depend on the platform and process requirements.
In n8n, nodes are the components used to build a workflow. A node defines when the workflow runs, retrieves or sends data, controls logic, or connects to external services.
Number portability is the ability for phone users to change their phone service provider without also changing their phone number. Before this capability was introduced, keeping a phone number when switching providers was possible only if that same number was available with the new provider.
OAuth is an internet standard for delegated authorization. It lets a service access resources on another server on the owner's behalf, without requiring the user to give the third party their credentials for that server.
Object-relational mapping (ORM) is a programming technique used in object-oriented programming languages to convert data between incompatible type systems: complex types (objects) in object-oriented programming languages and tables in relational databases.
The ability to understand what is happening in a system from signals such as logs, metrics, and request traces, and to investigate why a process behaved as it did.
In cryptography and information security, a one-time password, or OTP, is a password used for a single authentication and then discarded. Because it is used only once, such a system provides what is called strong authentication and is resistant to replay attacks.
Open source refers to software whose source code is accessible and free to use, modify, and redistribute. This development model lets anyone who wishes to participate in developing the software and contribute to improving it.
Swagger is a suite of tools for API developers from SmartBear Software. It was also formerly the name of a specification from which the OpenAPI Specification evolved.
Under the WhatsApp Business Policy, user consent is required before sending template messages. The consent should clearly identify the business and the purpose of the messages.
Operating-system-level virtualization is a virtualization method used in computer systems, especially servers. It relies on the operating system kernel's ability to provide multiple separate user spaces that share a common operating system kernel. Such instances, sometimes called containers or software containers, appear to their owners and users to be real servers.
Overfitting is a condition in which a model is fitted too closely to the data it was trained on, so it performs worse when making predictions on new data.
Platform as a service provides a platform for running or developing applications. Users get both computing resources and the infrastructure software needed to run or develop applications, without managing all the hardware themselves.
A penetration test (also called a pentest or PT) is a planned and controlled attack on a computer system carried out by a tester (a hacker or penetration tester) to find security weaknesses, assess the potential access they could provide, and evaluate the value of that access and the information stored by the system.
A point of sale is the time and place where a retail transaction is completed: the merchant calculates the total, the customer pays, and a receipt may be issued.
In software deployment, the production environment is the environment end users interact with. In an industrial setting, it is separated from the development environment, sometimes with intermediate stages for testing and rollback.
A prompt is the input a large language model uses to generate text. Prompt engineering is the practice of writing effective instructions so a model consistently produces content that meets requirements.
Prompt engineering is a technique in the field of artificial intelligence. A “prompt” in this context is a specific input, usually an instruction, question, or query, given to an AI model or language model to produce a particular output.
An attempt to make a model follow instructions from input or external content that is not an authority, contrary to the task and the application's restrictions.
A proxy server, also known in Hebrew as an intermediary server, primarily provides fast access to external resources on a computer network. It can conceal an IP address by routing all internet connections through the proxy server: data is sent to it and then to the server, and back through it to the user.
The public switched telephone network (PSTN) is the worldwide network of circuit-switched telephone networks. Bezeq's telephone network in Israel is one of the networks included in it.
Public-key encryption, also called asymmetric encryption, is a branch of cryptography in which the encryption key differs from the decryption key. Each user generates a pair of keys: a public key, which is available to anyone and used for encryption, and a corresponding private key, which is kept secret and used for decryption.
A race condition is a situation where a system's result depends on the order in which multiple processes run concurrently, or on events that do not occur at the expected time.
RAG is a technique that gives language models access to new information from external sources to improve their answers. The system retrieves relevant documents and grounds the answer in current or internal knowledge, sometimes through a vector database.
Rate limiting controls the rate at which requests are sent to or received by a network interface. On the WhatsApp platform, API requests count toward the app's request quota in a rolling one-hour window, and there are also message and throughput limits.
Redis is an in-memory NoSQL database that stores information as key-value pairs, making it especially fast. In automation systems, it is often used as a cache and task queue, for example when running n8n in queue mode.
A regular expression (Regex) is a sequence of characters that defines a search pattern in text. In automation, regular expressions are used to find, extract, or validate information, such as a phone number or email address in an incoming message.
Reinforcement learning, also called learning through reinforcement, is a field of machine learning in computer science concerned with how agents act in a dynamic environment to maximize the cumulative reward resulting from their actions. Reinforcement learning is considered one of three machine learning paradigms, alongside supervised learning and unsupervised learning.
Reinforcement learning from human feedback (RLHF) is a method for training a language model that has already undergone pretraining, so that it behaves in accordance with human preferences.
REST is an architectural style for implementing network services (APIs). Its central idea is to define a resource whose state changes through interactions between a provider and a consumer, usually over HTTP.
A reverse proxy is a server that receives requests from clients, retrieves resources from one or more servers behind it, and returns them as if they came from the proxy.
Return on investment (ROI) is a financial metric that assesses whether investing money, time, or other resources in a particular activity is worthwhile.
In computer security, role-based access control, or RBAC, is an approach to managing user permissions. It can implement either mandatory access control (MAC) or discretionary access control (DAC).
Reverting code or a configuration to an earlier version after a problematic change. It does not automatically undo business actions that have already taken place.
Robotic process automation (RPA) uses software robots to perform repetitive actions, often through user interfaces. It can work alongside AI capabilities, but does not require a language model or AI agent.
In deployment environments, a sandbox (or experimental environment) is a place for experiments that are not intended to move into production. It appears alongside environments such as QC and acceptance testing, outside the development-testing-production flow.
In Make, a scenario is a series of modules describing how data should move and change between apps and services. It is the visual automation unit that starts with a trigger.
A software development kit is a set of tools for developing software and extending capabilities for a specific platform. An SDK may include language-specific API files and sometimes examples, documentation, and debugging tools.
SIEM (short for Security Information and Event Management) is a category of information security software that serves as an organization's central technology monitoring hub. It automatically collects and analyzes data and activity logs from the organization's computing, communications, and cloud systems in real time to identify threats, alert on cyberattacks, and manage responses to security incidents.
Sender Policy Framework (SPF) is a simple email authentication system developed to identify spoofing and impersonation. It lets receiving mail servers check whether messages sent from a given domain originated from a host authorized by the domain owner. RFC 7208 defines Sender Policy Framework.
Sentiment analysis, sometimes also called opinion mining or emotional AI, uses natural language processing, text mining, computational linguistics, and biometrics to systematically identify, extract, quantify, and study emotional states and subjective information. Sentiment analysis is widely applied to reviews, survey responses, online networks, social networks, and more.
A service-level agreement is an agreement between a provider and a customer that defines aspects such as quality, availability, and responsibility. A common element is whether the product meets the agreed terms, along with metrics such as the maximum response time for a request.
SQL injection is a method of exploiting a security vulnerability in a computer program by interacting with its database. The name comes from a user entering SQL code into an input field intended for ordinary data.
SSO is an authentication mechanism that lets a user sign in to multiple related but independent software systems with a single identity. True single sign-on grants access to services without requiring authentication factors again.
A staging (or pre-production) environment is a test environment that mirrors production as closely as possible. It is used to test before releasing changes to end users.
A speech recognition system (STT or ASR) is a computer system that decodes human speech and converts it to text. The input may come from a microphone, a phone call, or another audio source.
In n8n, a sub-workflow is a reusable part of a larger workflow. A group of nodes can be turned into a sub-workflow to break logic into reusable components.
Supervised learning is a machine learning technique in which a system learns to solve problems from a large collection of examples that already have the correct answer.
In the OpenAI API, the instructions parameter gives the model high-level guidance about its behavior, such as tone, goals, and examples. These instructions take precedence over the prompt in the input and roughly correspond to a developer or system message.
Technical debt is a software development concept that describes the hidden cost of future work created when choosing a quick solution today instead of a thorough one.
Temperature is a parameter that determines how much randomness there is in a language model's predictions when generating text. A higher value produces more creative and varied responses, while a lower value produces more conservative and predictable responses.
In n8n, templates are ready-made workflows built by n8n and community members that you can import into your instance. After importing, you usually add credentials and adjust the settings. (A WhatsApp message template is a separate term: Message template.)
Terraform is open-source infrastructure-as-code (IaC) software developed by HashiCorp. It lets users define and provision data center infrastructure using a readable configuration file in the declarative HCL language or in JSON.
Time to first token (TTFT) is a performance metric that measures how long it takes from when a language model receives a prompt until it produces the first token of its response.
Transport Layer Security (TLS), or its earlier version Secure Sockets Layer (SSL), are the most popular and important security protocols on the internet. Nearly all websites protected by cryptographic means rely on protocols that are part of the SSL/TLS suite.
In OpenAI documentation, tool calling is another name for function calling: the model decides it needs an external tool and returns a tool call. In n8n, a tool is an additional resource the model can access for information or to take action in an external system.
A transformer is a type of neural network architecture used to process sequential data, such as text, audio, video, biological sequences, and time series. The architecture is based mainly on attention mechanisms, particularly one called self-attention, which lets the model calculate relationships between every pair of units in a sequence.
In n8n, a trigger node runs the process when a condition is met. Production workflows require at least one trigger. In Zapier, the trigger is the event that starts the Zap.
A TTS system converts text in ordinary language into speech. Speech synthesis is the artificial production of human speech in software or hardware. The reverse process is speech recognition.
In computing, a Globally Unique Identifier (GUID) or Universal Unique Identifier (UUID) is a pseudorandom number used to uniquely identify an object in any context in which it may appear.
A VPN (short for Virtual Private Network) is a service for transferring private information over infrastructure that is primarily or entirely publicly owned, or privately owned but publicly accessible. A VPN is intended to transfer information from an organization's network to an end user who is somewhere without the organization's private communications infrastructure, but where public infrastructure, such as the internet or telephone network, is available.
An AI agent, as described by n8n, receives requests, makes decisions, and performs tasks using a language model. A voice agent applies the same pattern to a spoken channel: incoming speech goes through STT, the agent acts, and the response is delivered through TTS. There is no official Hebrew page for the compound term.
A WhatsApp Business Account represents a business and contains phone numbers, usernames, and analytics. On the platform, it sits under a business portfolio.
A WAF (short for Web Application Firewall) is a firewall designed for web applications that monitors, filters, and blocks HTTP traffic to and from an application. Unlike a regular firewall that operates at the network level, a WAF examines the content of requests and can block attacks such as SQL injection and XSS.
A webhook is a way to extend a web application's behavior through custom callbacks. These are "user-defined HTTP callbacks" triggered by an event: the source website sends an HTTP request to a configured address.
To authenticate an incoming webhook message, an HMAC signature in an HTTP header and a shared secret are sometimes used. GitHub, Stripe, and Facebook use this technique to prevent spoofing.
An official WhatsApp business bot sends and receives messages through the WhatsApp Business Platform, including the Cloud API. It is not a personal WhatsApp account or an unofficial library.
In n8n, a workflow is a collection of nodes that automates a process. It starts when a trigger condition is met, then proceeds in sequence to perform complex tasks.
In n8n, a workflow template is a ready-made workflow that you can import, learn from, and adapt to your use case, including connecting credentials and configuring settings.
YAML is a text format for storing data structures, written to be human-readable. It is used mainly for configuration files and transferring data between applications.
A Zap is a process in Zapier that connects apps to automate repetitive tasks. It consists of a trigger that starts it and one or more actions that run after the trigger.
A zero-trust network, also called a zero-trust network or simply zero trust (and also a zero-trust architecture, zero-trust security model, ZTA, or ZTNA), is a term in information technology (IT) and cybersecurity describing an approach to designing and implementing IT networks. The central idea behind zero trust is that devices on a network, such as laptops, should not be trusted by default, even if they are connected to a managed corporate network such as the corporate LAN and even if they were previously authenticated.
A/B testing, also called split testing, is a method for comparing the current version with a change and determining which version produces better results. It is commonly used in marketing, e-commerce, web development, and other fields to assess users' responses to changes.
In software engineering, particularly in databases and transaction management, ACID stands for Atomicity, Consistency, Isolation, and Durability. These properties are a cornerstone of databases and transaction management systems.
An AI agent is an artificial intelligence system that can respond to requests, make decisions, and perform real-world tasks. It uses a large language model to interpret input and choose how to process a request using the information and resources available to it.
Information stored so an agent can use it beyond a single request, such as task state or preferences, according to the application's storage mechanism.
Agile software development is an approach in which software is developed in short cycles, feedback is gathered, and direction is adjusted as needed instead of planning everything in advance.
Artificial intelligence is the ability of technological systems and computers to simulate aspects of human intelligence, including learning, reasoning, problem-solving, and visual or linguistic perception. Unlike software that follows a fixed set of instructions, these systems identify patterns in data and adapt to changing circumstances.
An application programming interface (API) defines how software can request information or actions from other software. In business automation, it can be used, for example, to search for a customer, update an order, or create a task without performing those actions manually on screen.
An API key is a value included with a request to a service to identify the connection, associate usage with an account, or grant access, according to the provider's rules. The key type determines what it allows and whether it must be kept secret.
An artificial neural network (ANN), also called a neural network or connectionist network, is a computational mathematical model developed in inspiration from brain or cognitive processes that occur in a natural neural network and used in machine learning. This type of network usually contains many information units (inputs and outputs) connected to one another, with connections that often pass through “hidden” information units (hidden layers).
A record of significant actions that includes the actor's identity, time, and result, so it is possible to determine who changed or approved something.
Automation uses mechanical or electronic devices to perform a planned sequence of actions without human intervention. It frees people from routine, repetitive tasks by replacing them with independent technological means.
Base64 is a data encoding method that represents binary data in text format. Each sequence of six bits is represented by one ASCII character from a set of 64 possible characters. The term Base64 originates from a particular MIME encoding.
In Meta documentation, partners are entities that deploy value-added solutions as authorized WhatsApp service providers on behalf of clients. The current official name is Solution Partner (the market still says BSP): a Meta partner that provides full platform services (messaging, billing, integration, and support) and may extend a line of credit to a client.
Business intelligence (BI) is a field of information technology concerned with building systems that help an organization extract business-relevant information from the data it collects. Business intelligence systems provide historical and current information, as well as forecasts, about business operations. They can help identify meaningful patterns for managing an organization and create reports, alerts, and signals to support management decisions.
On the WhatsApp Platform, a business portfolio lets an organization manage its Meta assets in one place. It primarily acts as a container for WhatsApp Business Accounts (WABAs), and is required to use the platform. Buyers in Israel still commonly say Business Manager.
A CAPTCHA (short for Completely Automated Public Turing test to tell Computers and Humans Apart) is a challenge-response test in computing intended to ensure that a response is not generated by a computer. Typically, the computer presenting the test asks a random question that a person can answer but that is very difficult for a computer.
In cryptography, a certificate authority (CA) is an entity that stores, signs, and issues digital certificates. A digital certificate confirms that a public key belongs to the certificate holder. The CA acts as a trusted third party between the certificate holder and the party relying on the certificate.
A chatbot is software that specializes in managing text conversations, using natural language or a limited script. A conversation can be intended for marketing, consultation, customer service, or training, and the bot can be based on a language model or predefined rules.
An automated process that checks changes and prepares or performs a deployment according to set rules, instead of relying on manual steps for every update.
The WhatsApp Business Platform Cloud API lets you programmatically send and receive WhatsApp messages and calls, including text, rich media, and interactive messages. It runs on Meta's infrastructure and does not require a business to host its own WhatsApp server.
Cloud computing is a model that provides convenient, on-demand network access to a shared pool of computing resources (processing, storage, connectivity, and services).
In n8n, the Code node lets you write JavaScript or Python code and run it as a step in a workflow. It replaced the Function and Function Item nodes starting with version 0.198.0.
CVE (short for Common Vulnerabilities and Exposures) is a standardized system for identifying and recording known security vulnerabilities, with a unique identifier assigned to each one. The system is maintained by MITRE, with funding from the United States Department of Homeland Security.
Computer vision is a major field of computer science concerned with the automatic processing of images and video to extract and interpret the visual information they contain. Practitioners develop mathematical and software tools for processing visual information.
Configuration management is a process that keeps a system consistently aligned with its requirements, design, and operational information throughout its life.
A Content Delivery Network, or CDN, is a system based on proxy servers and data centers. A CDN improves the speed and quality of data delivery to end users of websites by distributing servers geographically.
In OpenAI documentation, the context window is the model's memory limit for a single generation request: how much data it can consider at one time. It is measured in tokens (pieces of input such as text or images), and its size varies between models.
Microsoft Copilot is a generative AI chatbot developed by Microsoft AI. During 2023, Microsoft unified the Copilot brand across its chatbot products (including Bing Chat and Microsoft 365), as an assistant in Windows, Edge, and Microsoft 365.
Customer relationship management is a field focused on customer service and understanding customer needs through information tools. It includes methodologies and software that help an organization manage customer relationships: acquisition, retention, and customer base development.
Cron is a tool for scheduling jobs on Unix-like operating systems. It is commonly used to schedule tasks, commands, or scripts at fixed intervals or dates, so it is especially common in system maintenance and administration.
XSS (short for Cross-site scripting) is a cyberattack against a web user that exploits a vulnerability in a web application. It allows an attacker to inject a malicious script intended to run in a user's browser and perform actions on the user's behalf on the site.
In digital communications and computer science, a hash function is a function that converts input of any length into output of a fixed length, usually much shorter. It is undesirable for a hash function to produce identical output for different inputs, but this is sometimes unavoidable. Hash functions are therefore measured by the probability of producing identical output.
CSRF (short for Cross Site Request Forgery) is an attack against a web user that forces the user's browser to perform unintended actions in web applications on the user's behalf by exploiting limitations in the HTTP protocol.
CSV is a way to store tables as plain text: each record is on a line, and fields are separated by a character (usually a comma). It is commonly used to transfer information between spreadsheets, databases, and automations.
On the WhatsApp Platform, the customer service window is the period after a user contacts a business during which non-template messages can be sent. Outside this window, businesses generally send template messages, which are subject to sending limits.
Data mining is the use of statistical tools to find correlations and trends in large amounts of information from different sources, even when the information is not organized as a database.
A database is a means of storing data in an organized way on a computer for retrieval and processing. It is stored on a data storage medium, usually a hard drive, that allows direct access to the data.
Deep learning is a set of machine learning methods based on artificial neural networks. “Deep” refers to learning through layered structures that progressively extract features from raw data.
In data communications, a denial-of-service attack (DoS) or distributed denial-of-service attack (DDoS) is a family of attacks intended to disable a computer system by placing an unusual load on its resources. DoS describes an attack originating from a single computer, while DDoS describes an attack originating from multiple synchronized computers.
A digital signature is a cryptographic method intended to verify the authenticity and integrity of a digital message or document. A valid digital signature indicates with a high degree of confidence that the message or document has not been corrupted or forged: it was created or sent by a known party and was not changed by a third party after signing, or otherwise tampered with.
DomainKeys Identified Mail, commonly known by the acronym DKIM, is an email authentication protocol developed to identify spoofing and impersonation. It does this by signing every outgoing email message. The public signing key is published in the Domain Name System (DNS).
DNS is a protocol that lets network endpoints translate human-friendly domain names into numeric IP addresses. It also provides other name-based services, such as listing mail servers.
Docker is open-source software for running applications in an isolated virtual environment called a container. Each container includes its own software, libraries, and configuration files, so the application runs consistently without interfering with other applications.
Embeddings are numerical representations of data using vectors. AI uses them to interpret complex data and relationships, and vector databases store them for retrieval.
Encryption is a cryptographic process for encoding information that converts its original representation, known as plaintext, into an alternative form known as ciphertext. Encryption systems generally use a pseudorandom encryption key generated by an algorithm.
An enterprise resource planning system is an administrative information system that handles an organization's needs (human resources, finance, marketing, logistics, and operations) as a whole. It unifies systems through a consistent interface and shared information.
Data transfer: extract data from a source, transform it into a consistent format, and load it into a destination (such as a spreadsheet, warehouse, or CRM).
According to OpenAI, evaluations (sometimes called evals) check a model's output against predefined style and content criteria. Writing evals is an essential part of building reliable LLM applications, especially when upgrading a model or trying a new one.
Organizing a process so that a change or event triggers responses in other systems, instead of one component managing everything through a continuous sequence of calls.
In n8n, an execution is one run of a workflow. It can be started manually during testing or automatically in production through a trigger, schedule, or polling.
An expert system is a computerized information system that performs the role of a human expert, or assists one, in a specific field of knowledge. It is a computer program that represents knowledge, draws conclusions from it, and can explain how it reached those conclusions.
Fine-tuning is the process of adapting a trained model for another task, usually a more focused one. Additional training on new data is applied to parameters that have already been trained, across the whole network or some of its layers.
In data communications, a firewall is a system for monitoring and blocking unwanted connections to a communications network or individual computer. At its core, a firewall is a "smart" router connecting a secured network to networks whose security cannot be guaranteed, such as the internet.
A freelancer is a self-employed professional who provides services for predefined tasks to multiple clients, with control over how the work is done. There is no employer-employee relationship between the freelancer and the client, and the freelancer chooses clients, workload, and payment terms.
Freemium is a business model in which the product's basic tier is free, while advanced features are sold separately. It is a common model for digital tools.
In OpenAI documentation, function calling (also called tool calling) is a mechanism that lets a model connect to external systems and access data outside its training data. The model returns a tool call with arguments, and the application runs the function and returns the result to the model.
The GDPR is the European Union's General Data Protection Regulation. It governs the processing of personal data and people's rights. Whether it applies, and what requirements apply to a particular process, depends on the activities and circumstances.
Generative artificial intelligence is a term for systems that produce new content (text, images, code, and more) instead of only retrieving existing information.
GPT is the name of a family of OpenAI language models that can process and generate content from input and instructions. Capabilities, pricing, and usage limits vary by model and access method.
A graph database stores information in a graph structure: nodes represent entities, edges represent the relationships between them, and properties describe both. This structure is especially useful for queries about relationships, such as a customer network or knowledge graph.
GraphQL is an open-source query language and data manipulation language for APIs at runtime. It was developed at Facebook in 2012 and released in 2015, and lets a client request exactly the fields it needs.
gRPC (short for gRPC Remote Procedure Calls) is an open-source, cross-platform software framework for RPC. The gRPC software layer was originally created by Google. It uses HTTP/2 for data transport and Protocol Buffers as its interface description language.
According to NVIDIA NeMo, guardrails are a programmable control layer around LLM applications. They block, modify, or validate user input and model output that is unsafe, off-topic, malicious, or policy-violating.
In artificial intelligence, a hallucination occurs when a large language model mistakenly perceives patterns or objects that do not exist. In the context of RAG, an ungrounded answer includes a guess or hallucination unsupported by the source documents.
A helpdesk is a department or person that provides assistance and information, usually for computer or electronics issues. Research in this field also emphasizes the helpdesk's unique position: daily conversations with customers or employees about problems and how to use systems.
Human-in-the-loop (HITL) means incorporating a human decision, review, or approval into an automated process. In business automation, a process can prepare an action, wait for a person, and continue, change, or stop only after that person's decision.
HTTP is a protocol for transferring HTML pages and objects over the web. Communication begins with an exchange between a client and a server, from which requests such as GET or POST follow. REST calls, webhooks, and most cloud APIs are based on these requests.
Infrastructure as a service is a cloud service in which users receive computing resources for their use: they can install virtual machines, operating systems, and applications, allocate storage, and set backup and scaling policies.
Idempotency is a property of an action where repeating the same request does not add an intended effect beyond the first execution. In automation, it is used to prevent a retry from creating the same business action again.
In Make, incomplete executions are a safety mechanism that saves a scenario run that failed because of an error, so the scenario does not stop and data is not lost. The saved run can be handled automatically or manually. The mechanism is disabled by default.
Infrastructure as code (IaC) is the process of managing and provisioning computing data-center resources using machine-readable configuration files rather than physical hardware configuration or interactive configuration tools. The IT infrastructure managed through this process includes both physical equipment, such as bare-metal servers, and virtual machines, along with related configuration resources.
iPaaS is a suite of cloud services that lets customers develop, run, and manage integration processes between different applications. Customers build the connections without installing or managing hardware or middleware themselves.
ISO/IEC 27001 (commonly called ISO 27001) is an international standard for information security management. It was first published by the International Organization for Standardization (ISO) and the International Electrotechnical Commission (IEC) in 2005, then updated in 2013 and again in 2022.
ISO/IEC 27701 is an international standard that extends ISO/IEC 27001 with privacy requirements. Its purpose is to establish and maintain a privacy information management system on top of an existing information security management system.
The Privacy Protection Law, 5741-1981, is Israel's principal law governing the right to privacy. It also covers databases and the use of personal information, so it is relevant to automation processes that collect or transfer customer details. The scope of obligations in each case depends on the circumstances.
IVR systems are automated telephony systems that interact with callers, collect information, and route calls to the appropriate destination. They work with speech recognition and DTMF input from the phone keypad.
JSON is a standard file and data interchange format that uses human-readable text to store and transfer key-value data structures. It is commonly used to exchange information between a server and a client in automations and APIs.
JWT (JSON Web Token) is a standard for transferring signed information between two parties in JSON format. A common use is as an access token for authentication: after the server verifies the client, it issues a token, and the client presents it without sending a password again.
A knowledge graph is a structured representation of knowledge in graph form. Entities such as people, places, organizations, concepts, or events are represented as nodes, and the relationships between them as labeled edges. A knowledge graph aims to make facts and relationships explicit in a form that both people and computer systems can process.
Kubernetes, sometimes abbreviated as k8s and pronounced "koo-ber-NET-eez," is an open-source system for managing and automatically deploying applications in containers. It was originally designed by Google and is now maintained by the Cloud Native Computing Foundation. Kubernetes supports a range of container tools, including containerd and rkt.
A landing page is a single web page visitors arrive at from an advertising link. Its purpose is to focus visitors on one message and prompt an action: submitting contact details (a lead) or making a purchase.
Latency is the time from when a process is initiated until it actually begins to run. In a conversation with an agent or an API call, it is the time elapsed before a response is received.
LDAP is an open protocol for accessing and managing directory services over IP networks. It is used to share information about users, systems, and services over a network, mainly within internal networks.
A large language model is a deep learning architecture for processing, predicting, and generating human language. It is trained on large amounts of text, has many parameters, and uses an attention mechanism that enables it to understand context.
In computer science, load balancing is the distribution of work among multiple computers to reduce workload. In computing, load balancing improves the distribution of workloads across computing resources such as computers, computer clusters, network links, central processing units, or disk drives.
A low-code development platform provides an environment, usually a graphical interface, where little or no code is written. It can generate a working application or require additional code for specific cases, shortening delivery time.
Make (formerly Integromat) is an automation platform that connects apps and services to automate processes and tasks, including complex end-to-end workflows, without requiring coding expertise.
MCP (Model Context Protocol) is an open standard for connecting AI applications to external systems: data sources, tools, and workflows. The official documentation compares it to a USB-C port for AI.
Templates are messages that can be prepared in advance. Approval is usually required before they can be sent. This is the type of message that can be sent to WhatsApp users outside the customer service window, and it is subject to quality ratings and sending limits.
Meta Platforms (trade name Meta) is an American technology company that operates Facebook, Instagram, WhatsApp, and Messenger, among other services. The WhatsApp Business Platform runs on Meta infrastructure, including Graph API, business portfolios, and Cloud API.
Multi-factor authentication is electronic authentication that uses two or more methods to grant access to an account or system. Its purpose is to provide protection even if one authentication method fails.
Middleware is computer software that connects software components or applications. It is generally used to develop and run complex, distributed applications.
Machine learning is a subfield of computer science and artificial intelligence focused on algorithms that enable computers to learn from examples. It is used for tasks where conventional programming with predefined rules is not feasible.
MQTT is a lightweight protocol for transferring messages between devices over a network. It suits environments with limited bandwidth, such as IoT and cellular networks, because it uses few resources.
n8n is an automation platform for workflows and AI agents: build visually, add code when needed, and connect to hundreds of integrations. It can run in its cloud or on your own infrastructure.
Named-entity recognition is a natural language processing task that extracts names of entities from text and classifies them into categories. These entities are usually people, organizations, locations, dates, quantities, monetary values, percentages, and more.
Natural language processing is a subfield of artificial intelligence and linguistics. It focuses on processing natural language and understanding human text or speech so computers can handle things said or written in human languages.
No-code development means building applications or processes through a visual interface and configuration, instead of writing most of the integration in code. Advanced capabilities and customizations depend on the platform and process requirements.
In n8n, nodes are the components used to build a workflow. A node defines when the workflow runs, retrieves or sends data, controls logic, or connects to external services.
Number portability is the ability for phone users to change their phone service provider without also changing their phone number. Before this capability was introduced, keeping a phone number when switching providers was possible only if that same number was available with the new provider.
OAuth is an internet standard for delegated authorization. It lets a service access resources on another server on the owner's behalf, without requiring the user to give the third party their credentials for that server.
Object-relational mapping (ORM) is a programming technique used in object-oriented programming languages to convert data between incompatible type systems: complex types (objects) in object-oriented programming languages and tables in relational databases.
The ability to understand what is happening in a system from signals such as logs, metrics, and request traces, and to investigate why a process behaved as it did.
In cryptography and information security, a one-time password, or OTP, is a password used for a single authentication and then discarded. Because it is used only once, such a system provides what is called strong authentication and is resistant to replay attacks.
Open source refers to software whose source code is accessible and free to use, modify, and redistribute. This development model lets anyone who wishes to participate in developing the software and contribute to improving it.
Swagger is a suite of tools for API developers from SmartBear Software. It was also formerly the name of a specification from which the OpenAPI Specification evolved.
Under the WhatsApp Business Policy, user consent is required before sending template messages. The consent should clearly identify the business and the purpose of the messages.
Operating-system-level virtualization is a virtualization method used in computer systems, especially servers. It relies on the operating system kernel's ability to provide multiple separate user spaces that share a common operating system kernel. Such instances, sometimes called containers or software containers, appear to their owners and users to be real servers.
Overfitting is a condition in which a model is fitted too closely to the data it was trained on, so it performs worse when making predictions on new data.
Platform as a service provides a platform for running or developing applications. Users get both computing resources and the infrastructure software needed to run or develop applications, without managing all the hardware themselves.
A penetration test (also called a pentest or PT) is a planned and controlled attack on a computer system carried out by a tester (a hacker or penetration tester) to find security weaknesses, assess the potential access they could provide, and evaluate the value of that access and the information stored by the system.
A point of sale is the time and place where a retail transaction is completed: the merchant calculates the total, the customer pays, and a receipt may be issued.
In software deployment, the production environment is the environment end users interact with. In an industrial setting, it is separated from the development environment, sometimes with intermediate stages for testing and rollback.
A prompt is the input a large language model uses to generate text. Prompt engineering is the practice of writing effective instructions so a model consistently produces content that meets requirements.
Prompt engineering is a technique in the field of artificial intelligence. A “prompt” in this context is a specific input, usually an instruction, question, or query, given to an AI model or language model to produce a particular output.
An attempt to make a model follow instructions from input or external content that is not an authority, contrary to the task and the application's restrictions.
A proxy server, also known in Hebrew as an intermediary server, primarily provides fast access to external resources on a computer network. It can conceal an IP address by routing all internet connections through the proxy server: data is sent to it and then to the server, and back through it to the user.
The public switched telephone network (PSTN) is the worldwide network of circuit-switched telephone networks. Bezeq's telephone network in Israel is one of the networks included in it.
Public-key encryption, also called asymmetric encryption, is a branch of cryptography in which the encryption key differs from the decryption key. Each user generates a pair of keys: a public key, which is available to anyone and used for encryption, and a corresponding private key, which is kept secret and used for decryption.
A race condition is a situation where a system's result depends on the order in which multiple processes run concurrently, or on events that do not occur at the expected time.
RAG is a technique that gives language models access to new information from external sources to improve their answers. The system retrieves relevant documents and grounds the answer in current or internal knowledge, sometimes through a vector database.
Rate limiting controls the rate at which requests are sent to or received by a network interface. On the WhatsApp platform, API requests count toward the app's request quota in a rolling one-hour window, and there are also message and throughput limits.
Redis is an in-memory NoSQL database that stores information as key-value pairs, making it especially fast. In automation systems, it is often used as a cache and task queue, for example when running n8n in queue mode.
A regular expression (Regex) is a sequence of characters that defines a search pattern in text. In automation, regular expressions are used to find, extract, or validate information, such as a phone number or email address in an incoming message.
Reinforcement learning, also called learning through reinforcement, is a field of machine learning in computer science concerned with how agents act in a dynamic environment to maximize the cumulative reward resulting from their actions. Reinforcement learning is considered one of three machine learning paradigms, alongside supervised learning and unsupervised learning.
Reinforcement learning from human feedback (RLHF) is a method for training a language model that has already undergone pretraining, so that it behaves in accordance with human preferences.
REST is an architectural style for implementing network services (APIs). Its central idea is to define a resource whose state changes through interactions between a provider and a consumer, usually over HTTP.
A reverse proxy is a server that receives requests from clients, retrieves resources from one or more servers behind it, and returns them as if they came from the proxy.
Return on investment (ROI) is a financial metric that assesses whether investing money, time, or other resources in a particular activity is worthwhile.
In computer security, role-based access control, or RBAC, is an approach to managing user permissions. It can implement either mandatory access control (MAC) or discretionary access control (DAC).
Reverting code or a configuration to an earlier version after a problematic change. It does not automatically undo business actions that have already taken place.
Robotic process automation (RPA) uses software robots to perform repetitive actions, often through user interfaces. It can work alongside AI capabilities, but does not require a language model or AI agent.
In deployment environments, a sandbox (or experimental environment) is a place for experiments that are not intended to move into production. It appears alongside environments such as QC and acceptance testing, outside the development-testing-production flow.
In Make, a scenario is a series of modules describing how data should move and change between apps and services. It is the visual automation unit that starts with a trigger.
A software development kit is a set of tools for developing software and extending capabilities for a specific platform. An SDK may include language-specific API files and sometimes examples, documentation, and debugging tools.
SIEM (short for Security Information and Event Management) is a category of information security software that serves as an organization's central technology monitoring hub. It automatically collects and analyzes data and activity logs from the organization's computing, communications, and cloud systems in real time to identify threats, alert on cyberattacks, and manage responses to security incidents.
Sender Policy Framework (SPF) is a simple email authentication system developed to identify spoofing and impersonation. It lets receiving mail servers check whether messages sent from a given domain originated from a host authorized by the domain owner. RFC 7208 defines Sender Policy Framework.
Sentiment analysis, sometimes also called opinion mining or emotional AI, uses natural language processing, text mining, computational linguistics, and biometrics to systematically identify, extract, quantify, and study emotional states and subjective information. Sentiment analysis is widely applied to reviews, survey responses, online networks, social networks, and more.
A service-level agreement is an agreement between a provider and a customer that defines aspects such as quality, availability, and responsibility. A common element is whether the product meets the agreed terms, along with metrics such as the maximum response time for a request.
SQL injection is a method of exploiting a security vulnerability in a computer program by interacting with its database. The name comes from a user entering SQL code into an input field intended for ordinary data.
SSO is an authentication mechanism that lets a user sign in to multiple related but independent software systems with a single identity. True single sign-on grants access to services without requiring authentication factors again.
A staging (or pre-production) environment is a test environment that mirrors production as closely as possible. It is used to test before releasing changes to end users.
A speech recognition system (STT or ASR) is a computer system that decodes human speech and converts it to text. The input may come from a microphone, a phone call, or another audio source.
In n8n, a sub-workflow is a reusable part of a larger workflow. A group of nodes can be turned into a sub-workflow to break logic into reusable components.
Supervised learning is a machine learning technique in which a system learns to solve problems from a large collection of examples that already have the correct answer.
In the OpenAI API, the instructions parameter gives the model high-level guidance about its behavior, such as tone, goals, and examples. These instructions take precedence over the prompt in the input and roughly correspond to a developer or system message.
Technical debt is a software development concept that describes the hidden cost of future work created when choosing a quick solution today instead of a thorough one.
Temperature is a parameter that determines how much randomness there is in a language model's predictions when generating text. A higher value produces more creative and varied responses, while a lower value produces more conservative and predictable responses.
In n8n, templates are ready-made workflows built by n8n and community members that you can import into your instance. After importing, you usually add credentials and adjust the settings. (A WhatsApp message template is a separate term: Message template.)
Terraform is open-source infrastructure-as-code (IaC) software developed by HashiCorp. It lets users define and provision data center infrastructure using a readable configuration file in the declarative HCL language or in JSON.
Time to first token (TTFT) is a performance metric that measures how long it takes from when a language model receives a prompt until it produces the first token of its response.
Transport Layer Security (TLS), or its earlier version Secure Sockets Layer (SSL), are the most popular and important security protocols on the internet. Nearly all websites protected by cryptographic means rely on protocols that are part of the SSL/TLS suite.
In OpenAI documentation, tool calling is another name for function calling: the model decides it needs an external tool and returns a tool call. In n8n, a tool is an additional resource the model can access for information or to take action in an external system.
A transformer is a type of neural network architecture used to process sequential data, such as text, audio, video, biological sequences, and time series. The architecture is based mainly on attention mechanisms, particularly one called self-attention, which lets the model calculate relationships between every pair of units in a sequence.
In n8n, a trigger node runs the process when a condition is met. Production workflows require at least one trigger. In Zapier, the trigger is the event that starts the Zap.
A TTS system converts text in ordinary language into speech. Speech synthesis is the artificial production of human speech in software or hardware. The reverse process is speech recognition.
In computing, a Globally Unique Identifier (GUID) or Universal Unique Identifier (UUID) is a pseudorandom number used to uniquely identify an object in any context in which it may appear.
A VPN (short for Virtual Private Network) is a service for transferring private information over infrastructure that is primarily or entirely publicly owned, or privately owned but publicly accessible. A VPN is intended to transfer information from an organization's network to an end user who is somewhere without the organization's private communications infrastructure, but where public infrastructure, such as the internet or telephone network, is available.
An AI agent, as described by n8n, receives requests, makes decisions, and performs tasks using a language model. A voice agent applies the same pattern to a spoken channel: incoming speech goes through STT, the agent acts, and the response is delivered through TTS. There is no official Hebrew page for the compound term.
A WhatsApp Business Account represents a business and contains phone numbers, usernames, and analytics. On the platform, it sits under a business portfolio.
A WAF (short for Web Application Firewall) is a firewall designed for web applications that monitors, filters, and blocks HTTP traffic to and from an application. Unlike a regular firewall that operates at the network level, a WAF examines the content of requests and can block attacks such as SQL injection and XSS.
A webhook is a way to extend a web application's behavior through custom callbacks. These are "user-defined HTTP callbacks" triggered by an event: the source website sends an HTTP request to a configured address.
To authenticate an incoming webhook message, an HMAC signature in an HTTP header and a shared secret are sometimes used. GitHub, Stripe, and Facebook use this technique to prevent spoofing.
An official WhatsApp business bot sends and receives messages through the WhatsApp Business Platform, including the Cloud API. It is not a personal WhatsApp account or an unofficial library.
In n8n, a workflow is a collection of nodes that automates a process. It starts when a trigger condition is met, then proceeds in sequence to perform complex tasks.
In n8n, a workflow template is a ready-made workflow that you can import, learn from, and adapt to your use case, including connecting credentials and configuring settings.
YAML is a text format for storing data structures, written to be human-readable. It is used mainly for configuration files and transferring data between applications.
A Zap is a process in Zapier that connects apps to automate repetitive tasks. It consists of a trigger that starts it and one or more actions that run after the trigger.
A zero-trust network, also called a zero-trust network or simply zero trust (and also a zero-trust architecture, zero-trust security model, ZTA, or ZTNA), is a term in information technology (IT) and cybersecurity describing an approach to designing and implementing IT networks. The central idea behind zero trust is that devices on a network, such as laptops, should not be trusted by default, even if they are connected to a managed corporate network such as the corporate LAN and even if they were previously authenticated.