| 25 | PR | pyannote-audio | Fixed an issue where the inference progress hook reported a completion count greater than the total | sungbin1015 | | Open |
| 26 | PR | pyannote-audio | Fixed KeyError 'uri' when passing a file object to a dict | sungbin1015 | | Open |
| 27 | PR | vllm | Fixed an issue where collect_env would read packages from other environments or crash in a UV environment | sungbin1015 | | Open |
| 28 | PR | pymilvus | Fixed missing escape for VARCHAR primary key when querying or deleting by ids | sungbin1015 | | Open |
| 29 | PR | python-dotenv | Fixed an issue where dotenv run rejected FIFO-style .env files | sungbin1015 | | Open |
| 30 | PR | python-multipart | Fixed a TypeError that occurred when passing UPLOAD_DIR as a string and enabling UPLOAD_KEEP_FILENAME | sungbin1015 | | Open |
| 31 | PR | silero-vad | Fixed an issue where consecutive speech was omitted during forced splitting with max_speech_duration_s | sungbin1015 | | Open |
| 32 | PR | silero-vad | Fixed an issue where the default ONNX model incorrectly recognized a path containing "16k" as 16 kHz-specific | sungbin1015 | | Open |
| 33 | PR | datasets | Modified to encode audio paths and bytes even without torchcodec (audiofolder, from_dict) | sungbin1015 | | Open |
| 34 | PR | litellm | Fixed an issue where hosted_vllm altered the caller’s video file path, causing video to be missing in fallback | sungbin1015 | | Open |
| 35 | PR | speechbrain | Fixed an issue where get_all_files excluded all files following those matched by exclude_and | sungbin1015 | | Open |
| 36 | PR | vllm | Fixed an issue where speech recognition and translation requests ignored the server's default presence/frequency penalty | sungbin1015 | | Open |
| 37 | PR | vllm | /Fixed an issue where top_logprobs=-1 in /v1/responses returned an empty list | sungbin1015 | | Open |
| 38 | PR | vllm | Fixed an issue where run-batch did not import the tool/reasoning parser plugin | sungbin1015 | | Open |
| 39 | PR | faster-whisper | Fixed an issue where detect_language crashes when vad_parameters is in dictionary format | sungbin1015 | | Closed |
| 40 | PR | vllm | Fixed backslash handling in the tool parser make_valid_python | sungbin1015 | fatday, NishchayMahor, Kushal9889 | Closed |
| 41 | PR | whisperX | Fixed to leave unsorted segments in SRT/VTT | sungbin1015 | | Closed |
| 42 | PR | starlette | Added Allow header to StaticFiles 405 response | sungbin1015 | | Closed |
| 43 | Issue | pyannote-audio | Audio.crop's pad mode returns incorrect length and actual audio outside the specified range | sungbin1015 | | Open |
| 44 | Issue | PyMySQL | ValueError occurs during SSLContext creation when only ssl_verify_identity is enabled | sungbin1015 | | Open |
| 45 | Issue | pymilvus | Issue where sorting query/search results in-place causes vector and dynamic fields to appear in different rows | sungbin1015 | | Open |
| 46 | Issue | agent-compose | Issue where the entire composition process would halt because LLM thinking did not complete during the boundary phase | sungbin1015 | | Open |
| 47 | Issue | agent-vision | Allowed a vulnerable version of asyncmy (CVE-2025-65896) | sungbin1015 | | Open |
| 48 | Issue | agent-vision | Issue where callback_url was sent without format validation | sungbin1015 | | Open |
| 49 | Issue | agent-stt | Issue where the stt_svc process crashes due to a Rust panic while extracting web search text | sungbin1015 | | Open |
| 50 | Issue | faster-whisper | Issue where transcribe(chunk_length) permanently changes the window size for subsequent calls | sungbin1015 | Purfview | Closed |
| 51 | Issue | Qwen3-ASR | Issue where Korean and English words become concatenated due to missing spaces when merging long audio chunks | sungbin1015 | | Closed |
| 52 | Issue | whisperX | Issue where segments that failed alignment are omitted from SRT/VTT output | sungbin1015 | | Closed |
| 53 | Issue | agent-compose | Issue where clip timestamps were truncated to integer seconds, causing the start to overlap with the previous shot | sungbin1015 | | Closed |
| 54 | Issue | ui-sbs-viewer | Issue where global locks caused GPU rendering to overlap and pending requests to disappear — replaced with a server queue | sungbin1015 | | Closed |
| 55 | Issue | agent-vision | Issue where forcing the reprocessing of a preceding chunk caused a conflict with the generation of subsequent chunks, halting analysis | sungbin1015 | | Closed |
| 56 | Issue | micromark-extension-mdxjs-esm | Issue where re-exporting a value declared in a previous ESM block caused parsing to fail and the error to point to the wrong line | sungbin1015 | | Closed |
| 57 | Discussion | faster-whisper | In Korean broadcasts, denoise reduces hallucinations but lowers language detection accuracy (95.2% → 93.4%) | sungbin1015 | | Posted |
| 58 | Discussion | speechbrain | Comparison of VoxLingua107 and Whisper for speech recognition in Korean broadcasts with background noise | sungbin1015 | | Posted |
| 59 | Discussion | Qwen3-Omni | Serving Qwen3-Omni as a vLLM on a Blackwell GPU: Configuration and 2 Key Considerations | sungbin1015 | | Posted |