AI

Agent

Tasks that run AI agents combining system prompts, user input, tools, content retrievers, and optional memory to produce grounded responses.

Tool

Tasks that expose tools agents can call: Kestra flow/task runners, code execution, MCP clients, and web search to extend LLM reasoning with actions and real-time data.

Embeddings

Tasks that provision embeddings stores (Chroma, PGVector, Pinecone, Redis, Qdrant, etc.) for chunk storage and similarity search powering RAG flows.

Provider

Tasks that configure Large Language Model providers (OpenAI, Gemini, Anthropic, Mistral, Bedrock, etc.) to supply chat/completion models used across the AI plugins.

RAG

Tasks that ingest, chunk, and embed documents, search vectors, and run retrieval-augmented chat so LLM answers stay grounded in indexed content.

Memory

Tasks that supply conversational memory backends (Kestra KV Store, Redis, PostgreSQL) to persist recent messages for agents and chat completions.

Retriever

Tasks that pull external context for Retrieval Augmented Generation: web search via Google or Tavily, or SQL database retrieval feeding results into AI prompts.

Completion

Tasks that call LLMs for chat, classification, image generation, and JSON-structured extraction, with optional tools, memory, and provider flexibility.