Glossary

Key terms and environment settings for R2R

Key Terms

  • RAG (Retrieval-Augmented Generation): A technique that combines information retrieval and language model generation to produce more accurate and informative responses.
  • Ingestion: The process of parsing and indexing documents into the R2R system for later retrieval and generation.
  • Fragment: A chunk of text extracted from an ingested document, used for similarity search and context in RAG.
  • Hybrid Search: A search method that combines semantic vector search with traditional keyword search for improved relevancy.
  • Knowledge Graph: A structured representation of entities and their relationships, used for advanced querying and reasoning.
  • RAG Agent: An interactive and intelligent query interface that can formulate its own questions, search for information, and provide informed responses based on retrieved context.
  • Collection: A grouping of documents and users for efficient access control and organization.

Environment Settings

  • OPENAI_API_KEY: API key for OpenAI’s language models and embeddings.
  • ANTHROPIC_API_KEY: API key for Anthropic’s language models and embeddings.
  • R2R_POSTGRES_USER: Username for the Postgres database.
  • R2R_POSTGRES_PASSWORD: Password for the Postgres database.
  • R2R_POSTGRES_HOST: Hostname or IP address of the Postgres database server.
  • R2R_POSTGRES_PORT: Port number for the Postgres database server.
  • R2R_POSTGRES_DBNAME: Name of the Postgres database to use for R2R.
  • R2R_PROJECT_NAME: Defines the tables within the Postgres database where the selected R2R project resides.
  • R2R_PORT: Defines the port over which the R2R process is served.
  • R2R_HOST: Defines the host address over which the R2R process is served.
  • HATCHET_CLIENT_TOKEN: API token for Hatchet orchestration service (required for full R2R installation).
  • UNSTRUCTURED_API_KEY: API key for Unstructured.io document parsing service (required for full R2R installation).

For more information on these terms and settings, please refer to the relevant sections of the documentation: