All issues

Search results

Vocabulary cache staging file: pid-based name collides across containers, and leaks when the rename fails

Status: Open.

#121 In invoke-ai/InvokeAI-7;

· lstein opened on Aug 19, 2026

Cluster vocabulary caches for old embedding models are never pruned (~5 MB each, forever)

Status: Open.

#120 In invoke-ai/InvokeAI-7;

· lstein opened on Aug 19, 2026

Cluster labels: _vocab_lock is held across the whole vocabulary build, blocking the request threads its docstring says it must not

Status: Open.

#119 In invoke-ai/InvokeAI-7;

· lstein opened on Aug 19, 2026

14 bundled default workflows pin stale invocation versions; the registry test only covers the video ones

backend

bugSomething isn't workingSomething isn't working

services

Status: Open.

#116 In invoke-ai/InvokeAI-7;

· lstein opened on Aug 19, 2026

[bug]: Can't modify and save a workflow under a new name

bugSomething isn't workingSomething isn't working

Status: Open.

#109 In invoke-ai/InvokeAI-7;

· lstein opened on Aug 18, 2026

apply_ti constructs nn.Embeddings and installs skip_torch_weight_init without MODEL_LOAD_LOCK

Status: Open.

#105 In invoke-ai/InvokeAI-7;

· lstein opened on Aug 17, 2026

Two merge-time invariants have no test: params-slice migration chain integrity, and vendored MiniMax-H3 drift from upstream

Status: Open.

#104 In invoke-ai/InvokeAI-7;

· lstein opened on Aug 17, 2026

MiniMax H3 VAE working-memory docstring states ~1.3-1.4x its measured peaks; the 243-frame case is 1.44x

Status: Open.

#103 In invoke-ai/InvokeAI-7;

· lstein opened on Aug 17, 2026

extract_audio_pcm detects "no audio stream" by matching ffmpeg's stderr text, so a wording change turns silent clips into hard errors

Status: Open.

#102 In invoke-ai/InvokeAI-7;

· lstein opened on Aug 17, 2026

Cluster labels: uncapped annotations, and a shape contract label_clusters assumes rather than checks

Status: Open.

#101 In invoke-ai/InvokeAI-7;

· lstein opened on Aug 17, 2026

Image index search cache holds ~600 MB after stop(), and misses entirely with 3+ active scopes

Status: Open.

#96 In invoke-ai/InvokeAI-7;

· lstein opened on Aug 16, 2026

Semantic search URL fetch: allow-list is bypassed by HTTP proxy env vars, and by NAT64

Status: Open.

#95 In invoke-ai/InvokeAI-7;

· lstein opened on Aug 16, 2026