`raise "string"` in Python raises TypeError instead of the intended
error, making error messages confusing and debugging difficult.
Co-authored-by: gambletan <ethanchang32@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
fix: reduce TTFT by caching model lookups in chat completion
Skip expensive get_all_models() calls when models are already cached
in app.state. This significantly reduces Time To First Token (TTFT)
for chat completions and embeddings requests.
Previously, every request called get_all_models() which fetches model
lists from all configured backends. Now we check the cache first and
only call get_all_models() on cache miss.
Affected endpoints:
- openai: generate_chat_completion, embeddings
- ollama: embed, embeddings
Fixes#20069
Co-authored-by: Michael <42099345+mickeytheseal@users.noreply.github.com>
Remove Depends(get_session) from the /v1/completions endpoint to prevent database connections from being held during the entire duration of LLM calls.
Previously, the database session was acquired at request start and held until the response completed. Under concurrent load, this exhausted the connection pool, causing QueuePool timeout errors.
The fix allows Models.get_model_by_id() and has_access() to manage their own short-lived sessions internally, releasing the connection immediately after authorization checks complete.
Remove Depends(get_session) from the /v1/chat/completions endpoint to prevent database connections from being held during the entire duration of LLM calls.
Previously, the database session was acquired at request start and held until the streaming response completed. Under concurrent load, this exhausted the connection pool, causing QueuePool timeout errors.
The fix allows Models.get_model_by_id() and has_access() to manage their own short-lived sessions internally, releasing the connection immediately after authorization checks complete.
Remove Depends(get_session) from the /api/chat endpoint to prevent database connections from being held during the entire duration of LLM calls (30-60+ seconds for streaming responses).
Previously, the database session was acquired at request start and held until the streaming response completed. Under concurrent load, this exhausted the connection pool, causing QueuePool timeout errors for other database operations.
The fix allows Models.get_model_by_id() and has_access() to manage their own short-lived sessions internally, releasing the connection immediately after the quick authorization checks complete - before the slow external LLM API call begins.
- Fix file handle memory leak in download_file_stream by properly closing and reopening files
- Add requests.Session context manager for proper HTTP connection cleanup
- Remove unnecessary file.seek(0) after file reopening
- Add timeout to prevent hanging connections
This prevents memory accumulation during large file downloads and ensures
proper resource cleanup in all scenarios.
Signed-off-by: Sihyeon Jang <sihyeon.jang@navercorp.com>
- Replace Request object with user.id in cache key for get_all_models
- Request objects are new instances per HTTP request, preventing cache hits
- Cache keys now use user.id ensuring proper cache functionality
- Affects both Ollama and OpenAI model list endpoints
Signed-off-by: Sihyeon Jang <sihyeon.jang@navercorp.com>
Fix introduced on #15035 is over quoting headers.
Eg mails instead of user@example.com shown as user%40example.com
Eg names instead of First Last shown as First%20Last
Also we are spending some time quoting ids and roles without required.
Keep quote only on user name, initially had problem based on the discussion
https://github.com/open-webui/open-webui/discussions/14391
Also add space in safe characters, in order remove %20 from names.
