Features
Vector and OKF together
Link vector and OKF sources to one engine, and it searches both and merges the results.
Some questions are best answered by meaning, and others by following how ideas connect. An engine can read vector sources and OKF sources at the same time, so one question searches both.
Mix both kinds of source when...
- Some content is long and loosely structured, and some is a set of clear concepts.
- Questions sometimes use different words than your content, and sometimes ask how things relate.
How it works
- Search vector sources. The engine finds the chunks closest in meaning to the question.
- Search OKF sources. The engine finds the concepts that match the question and follows the links between them.
- Merge. Results from both are combined into one ranked list.
- Use.
lynk.querywrites its answer from the merged list, andlynk.retrievereturns it.
If the OKF search can't run, the vector results are still returned. OKF results add to an answer and never block it.
Set it up
- Add your sources in Sources, choosing vector for some and OKF for others. See Containers.
- Link both kinds to the same engine on its Sources tab.
OKF sources have no dimension, so you can link them alongside vector sources of any dimension.
Read the results
You don't change your code. query and retrieve return merged results:
chunks = lynk.retrieve_sync("Which plans include the Atlas desk warranty?", top_k=8)Next steps
Your engine searches both kinds of source. Make common questions faster with Precision Mode next.