Nasdanika
Data Sources
Store
UserKey Extractor
Extracts key from input. For example, generates an embedding vector.
Flavors
TODO
Implementations
OpenAIEmbeddingsKeyExtractor
TODO
Roadmap
- Caching key extractor filter for String -> double vectors. Use the same store type. Save options - on every modification, on a regular interval, on close. Storage formats - xmi, binary, compressed binary.
- Digest - MD5 and SHA in the
coremodule - Open source models like Llama 2, MPT-7B, …
- Generic REST API in the
coremodule - Bag of words, word2vec and other things provided by Apache OpenNLP and Smile NLP possibly in combination with WordNet