Issues - Mark/vllm_mm - cyguin git: Adlerian Hierarchical Self-Actualization Syllogism

Mark/vllm_mm

New issue

[CRITICAL] Model switch holds _model_switch_lock across 10-30s subprocess shutdown, serializing the entire inference queue (proxy.py:163-220)

triage-approved

#41 by Mark was closed 2026-09-18 14:48:50 +00:00

[CRITICAL] Port reservation leak when start_model fails between reserve and Popen (process_manager.py:127-142, 262-272)

triage-approved

#40 by Mark was closed 2026-09-18 23:13:07 +00:00

[CRITICAL] start_model memory check + status update nested inside 'stop other active models' loop (processes.py:90-140)

triage-approved

#39 by Mark was closed 2026-09-18 23:14:45 +00:00

BUG: proxy layer does not support multimodal (vision) requests

triage-approved

#37 by Mark was closed 2026-09-14 17:16:14 +00:00

BUG: UI has no field for mmproj/vision model configuration

triage-approved

#36 by Mark was closed 2026-09-18 02:08:26 +00:00

BUG: llama.cpp backend start endpoint never spawns llama-server

triage-approved

#35 by Mark was closed 2026-09-18 03:12:55 +00:00

chore: fix remaining lint errors

triage-approved

#34 by Mark was closed 2026-05-21 13:39:34 +00:00

[smoke-test] Test coverage review: 132 source files, 17 test files (12.9%)

feature-request

triage-approved

#28 by Mark was closed 2026-05-19 13:54:18 +00:00

[Security] MEDIUM: 4 issue(s) found in vllm_mm

triage-approved

#27 by Mark was closed 2026-05-19 13:54:12 +00:00

Container naming is misleading — manages subprocesses, not Docker containers

triage-approved

#25 by Mark was closed 2026-05-20 13:58:05 +00:00

model_repository.create() omits multimodal fields — mmproj config lost on creation

triage-approved

#23 by Mark was closed 2026-05-20 14:52:08 +00:00

Gemma-4 chat template uses wrong <|turn> tokens instead of correct <start_of_turn>/<end_of_turn>

triage-approved

#21 by Mark was closed 2026-05-20 15:25:58 +00:00

[BUG] No authentication or authorization on any API endpoint — full control exposed

feature-request

triage-approved

#19 by Mark was closed 2026-05-17 21:11:55 +00:00

[DOC] AGENTS.md incorrectly states no test framework is configured — tests exist

triage-approved

#18 by Mark was closed 2026-05-26 17:44:53 +00:00

[BUG] Settings update via PUT /api/models/{id} overwrites entire settings dict instead of merging

triage-approved

#17 by Mark was closed 2026-05-27 16:36:34 +00:00

[BUG] Frontend progress SSE stream uses wrong ID — never receives loading updates

triage-approved

#16 by Mark was closed 2026-05-26 20:15:17 +00:00

[BUG] Multimodal fields not persisted when creating new models via API

triage-approved

#13 by Mark was closed 2026-05-17 21:11:56 +00:00

[CRITICAL] Frontend silently drops settings fields during model create/update — speculative decoding, cache types, multimodal paths lost

triage-approved

#12 by Mark was closed 2026-05-26 22:18:24 +00:00

Bug: 80% progress is meaningless - hardcoded before health check starts

feature-request

triage-approved

#9 by Mark was closed 2026-05-17 20:23:53 +00:00

Bug: Subprocess logs not captured - cannot debug model start failures

triage-approved

#8 by Mark was closed 2026-06-16 05:45:37 +00:00

1 2