Open bobbykolev opened 3 weeks ago
We need smarter ingestion support for files and callouts to improve the quality of the results.
[ ] Preserve the context when splitting chunks.
[ ] Gather more context when reading chunks.
[ ] Use elastic search ingestion instead of VectorDB?
[ ] Improve Excel ingestion - extract in separate epic
[ ] ... Investigate multiple embeddings. Context increase, tokens limit?
Description
We need smarter ingestion support for files and callouts to improve the quality of the results.
Must have scope
[ ] Preserve the context when splitting chunks.
[ ] Gather more context when reading chunks.
[ ] Use elastic search ingestion instead of VectorDB?
[ ] Improve Excel ingestion - extract in separate epic
[ ] ... Investigate multiple embeddings. Context increase, tokens limit?