v0.0.52

Full Changelog: v0.0.51...v0.0.52

v0.0.50

Changes

• Update nearaidev/vllm-proxy-rs image digest across GLM-5, small-models, and Qwen3.5-122B configs (#3)
• Remove unrecognized vLLM args (--max-cudagraph-capture-size, --stream-interval) from gpt-oss-120b (#4)

v0.0.49

Changes

• Increase nginx proxy_read_timeout from 300s to 3600s across all CVM configs — fixes timeout errors for long-context inference requests (100K+ tokens) where prefill exceeds 5 minutes (#1)
• Update gpt-oss vLLM image to newer version (#2)

v0.0.48

• Update gpt-oss vllm image digest

v0.0.47

• Update vllm-proxy-rs to d46dd03 (attestation cache, GPU evidence serialization, retry on failure)

v0.0.46

• fix: update all vllm/vllm-openai images to new digest to address CUDA/NCCL crashes
• fix: reduce gpt-oss-120b GPU memory utilization to 0.90

v0.0.45

• fix: reduce gpt-oss-120b GPU memory utilization from 0.95 to 0.90 to address CUDA OOM crashes

v0.0.44

Changes

• feat: Add model-proxy registrar sidecar to all model configs (DeepSeek-V3.1, GLM-5, Qwen3.5-122B, small-models) for automatic endpoint/model registration with the proxy fleet
• fix: Remove prefill_token_shift and num_draft_tokens from Qwen3-30B speculative config — these params were removed in vLLM v0.16.0

v0.0.43

Changes

• Remove LMCache entirely — lmcache image, env vars, and --kv-transfer-config flags removed to fix crashes
• Upgrade all vLLM images to v0.16.0 (sha256:4801151759655c57606c844662e5213403c032a62d149c7ce61d615759a821ef)
• GPT-OSS-120B: --max-num-seqs 128→64, --max-num-batched-tokens 8K→16K
• Qwen3-30B-A3B: --max-num-batched-tokens 16K→24K
• Qwen3-VL-30B-A3B: add --gpu-memory-utilization 0.95, --max-model-len 32768, --max-num-seqs 64, --max-num-batched-tokens 16K (was completely unconfigured)

Add cloud-api usage reporting & JSON logs

• Add CLOUD_API_URL=https://cloud-api.near.ai to all Rust proxy services (small-models, Qwen3.5-122B)
• Fix MODEL_NAME in GLM-5.yaml: zai-org/GLM-5 → zai-org/GLM-5-FP8 (was causing 404 on usage reporting)
• Add LOG_FORMAT=json to all proxy services for structured logging