No-code UI
Core workflows via the no-code UI.
Create new index
Navigate to Index
feature via the left navbar.
Click Create a new pipeline
button. You should see a index configuration form.
Configure data source
Click Select a data source
dropdown and select desired data source.
Specify data source credentials and configurations (or upload files).
See full list of supported data sources
Configure data sink
Click Select a data sink
dropdown and select desired data sink.
See full list of supported data sinks
Configure embedding model
Select OpenAI Embedding
and put in your API key.
See full list of supported embedding models
Configure parsing & transformation settings
Toggle to enable or disable Llama Parse
.
Select Auto
mode for best default transformation setting (specify desired chunks size & chunk overlap as necessary.)
Manual
mode is coming soon, with additional customizability.
See parsing & transformation details
Deploy index
After configuring the ingestion pipeline, click Deploy Index
to kick off ingestion.
You should see an index overview with the latest ingestion status.
Manage existing index
Update files
Navigate to Data Sources
tab to manage your connected data sources.
You can upsert, delete, download, and preview uploaded files.
Sync index
Click Sync
button on the top right to pull upstream changes (data sources & files) and refresh index with the latest content.
Edit index
Click Edit
to open up modal for configuring ingestion settings.
Delete index
Click Delete
button to remove the index.
Note that this will not delete the uploaded files and previously registered data sources.
Observe ingestion status & history
Navigate to index overview tab. You should see:
- the latest ingestion status on the
Index Information
card (top right), and - ingestion job history on the
Activity
card (bottom left).
Test retrieval endpoint
Navigate to Playground
tab to test your retrieval endpoint.
Select between Fast
, Accurate
, and Advanced
retrieval modes.
Fast
: Only dense retrievalAccurate
: Use hybrid search with dense & sparse retrieval and re-rankingAdvanced
: Full customizability for tuning hybrid search and re-ranking
Input test query and specify retrieval configurations (e.g. base retrieval and top n after re-ranking) and click Run
button to see preview for retrieved nodes.
Click Copy
from bottom left panel to make direct REST API calls to the retrieval endpoint.