 Jason Kidd
|
fced3efd98
|
fix: Trailing slash was never removed from request.base_url because it's not a string but rather a starlette.datastructures.URL
|
2025-06-12 12:29:08 -07:00 |
|
|
Jason Kidd
|
210dc746f0
|
feat: Add OpenTelemetry Metrics Support via OTLP Exporter
|
2025-06-10 10:52:10 -07:00 |
|
|
Jason Kidd
|
c5b67ea430
|
fix: Remove unnecessary decode statement
Since we create our Redis instance with , we don't
need to worry about decoding this value here. Plus this doesn't work
in python3 anyway
|
2025-01-17 13:43:56 -08:00 |
|
|
Jason Kidd
|
b3a52be401
|
fix: Pgvector vector column size check was failing on initialization of database
|
2025-01-07 09:15:13 -08:00 |
|
|
Jason Kidd
|
70b74b5217
|
feat: Allow setting the initial vector length on pgvector document_chunk table
|
2025-01-03 09:18:59 -08:00 |
|
|
Jason Kidd
|
03e48de1a9
|
fix: Issue in some environments running in dev mode with redis where periodic cleanup takes longer than TIMEOUT_DURATION*2 seconds to be called and the lock expires.
|
2024-12-19 14:15:02 -08:00 |
|
|
Jason Kidd
|
8f51681801
|
feat: Make ENABLE_WEBSOCKET_SUPPORT disable polling entirely to allow multiple replicas without sticky sessions.
See https://socket.io/docs/v4/using-multiple-nodes/ for details why this was done.
Also create a redis key to track which replica is running the cleanup job
|
2024-12-18 07:54:12 -08:00 |
|
|
Jason Kidd
|
319ea8cb7f
|
feat: Add ability to set URI for pgvector
|
2024-11-04 14:27:33 -08:00 |
|
|
Jason Kidd
|
701f40aedd
|
feat: Initial support for pgvector
|
2024-11-04 14:26:41 -08:00 |
|