Artifact

The information an agent maintains and updates during execution to track progress, store intermediate results, and inform decisions. State includes: current goal, completed steps, pending actions, acc...

The output generated by the LLM in response to the prompt, representing the "assistant" turn in a conversation. Assistant messages contain the model's answers, generations, or actions (like tool calls...

A standardized dataset and evaluation protocol for measuring LLM performance on specific capabilities or tasks. Benchmarks enable comparison across models and tracking of progress over time. They typi...

A file placed in a project repository that provides persistent context, instructions, and constraints to AI coding assistants. Context files are automatically loaded when an AI assistant works on th...

A structured collection of data used for training, fine-tuning, or evaluating AI models. Datasets can contain text, images, labels, or structured records. In the LLM context, datasets are used for pre...

A dense numerical vector representation of content (text, images, audio) where similar items have similar vectors. Embeddings compress semantic meaning into a fixed-size array of floats (typically 384...

Data capturing human judgments about AI outputs—preferences between responses, quality ratings, safety flags, or corrections. Human feedback is the key input for RLHF and related alignment techniques....

A software component that adds functionality to a host application without modifying its core code. In AI systems, plugins typically expose new tools, data sources, or integrations that agents can u...

The text input provided to an LLM to elicit a response. A prompt can be a simple question, a complex instruction, or a carefully structured template with examples and context. Prompts are the primary ...

A structured prompt with placeholders (variables) that are filled in at runtime with specific values. Templates separate the prompt structure (fixed text, instructions, formatting) from the dynamic co...

A packaged, reusable capability that can be loaded into an AI agent or coding assistant to provide specialized knowledge, workflows, or tool access for a specific domain or task. Skills encapsulate ...

Data generated by AI models rather than collected from real-world sources. Synthetic data is increasingly used to train and fine-tune LLMs, especially when real data is scarce, expensive, or raises pr...

Instructions provided to an LLM that define its persona, behavior, constraints, and capabilities for the conversation. System prompts are typically set by the application developer (not the end user) ...

The basic unit of text that an LLM processes, typically a subword piece rather than a full word or character. Tokens are what models actually see—text is converted to tokens for input and tokens are c...

A structured description of a tool that tells an LLM what the tool does, what inputs it accepts, and what outputs it produces. Tool definitions typically include: name, description (natural language),...

The output returned after executing a tool invocation, which is then provided back to the LLM for incorporation into its reasoning or response. Tool results close the loop in tool use: model requests ...

The text corpus used to train an LLM's weights during pre-training. Training data determines what the model "knows"—its vocabulary, facts, reasoning patterns, and biases all come from the training cor...

A message in the conversation from the human user (or the application on behalf of the user), as opposed to system messages or assistant responses. User messages are the primary input mechanism for LL...

A predefined sequence or graph of steps that executes in a determined order, where the path through the workflow is decided at design time, not runtime. Workflows may have conditional branches, but th...