Supported components
This section lists all available components and shows how to select them in the configuration file. If you want to know what a specific component does and what role it plays in the system, take a look at Architecture. And if you don’t find the component you are looking for, feel free to add it and open a pull request!
Readers
Readers parse documents so that they can be added to the sytem’s database. Currently supported is text in the following file formats:
pdf
Splitters
Splitters take the text during ingestion and split it into chunks. Currently supported is a recursive text splitter only, so there is no option to select one in the configuration file yet.
Vector Databases
The database stores your embeddings and (references to) your document chunks. RAG Core supports both local and remote databases. The latter is a database which is hosted, for example, by a cloud provider.
Local Databases
Database |
config value |
requires |
|---|---|---|
|
|
Remote Databases
Database |
config value |
requires |
|---|---|---|
|
|
Embeddings
The embedding model is used to create a vector representation of your document chunks and queries. Currently, the following remote embedding model families are supported.
Embedding model family |
config value |
requires |
|---|---|---|
|
|
|
|
|
LLMs
The Large Language Model generates a response in natural language using the retrieved chunks and a prompt.
LLM model family |
config value |
requires |
|---|---|---|
|
|
|
|
|