Releases

Olla v0.0.29

Olla v0.0.29 ships an embedded read-only admin dashboard at /internal/ui/: fleet status, live throughput, per-endpoint health and your discovered model inventory, served from the proxy itself and loopback-only by default.

• Refreshed 2026 model catalogue: GPT-OSS, GLM, Kimi, Granite, Nemotron and more
• Native Prometheus metrics at /internal/metrics
• Circuit-breaker failover to the next healthy endpoint
• Credentials in endpoint URLs now fail startup instead of leaking

Olla v0.0.28

Olla v0.0.28 is a big release: native oMLX support for fast multi-model inference on Apple Silicon with Anthropic passthrough, per-endpoint authentication for local backends, opt-in CORS for browser clients, and a round of Anthropic translation and reliability hardening.

• Native oMLX support
• Docker Model Runner Anthropic passthrough
• Lemonade models are now routable
• OpenAI alias fixes

Olla v0.0.27

Olla v0.0.27 adds native LMDeploy backend support and ships sticky session fixes that improve KV cache alignment across multi-endpoint deployments.

• Add aimock test harness for sticky sessions
• Add language tags to skill code fences
• Add lmdeploy backend docs
• Add lmdeploy model converter and wire into factory and routes

Olla v0.0.26

Olla v0.0.26 is a focused bugfix release resolving SSL connection failures caused by incorrect Host header handling, plus portable multi-arch Docker image builds.

• Lets you build docker image locally without goreleaser
• Make docker builds portable across arm and amd

Olla v0.0.25

Olla v0.0.25 introduces model aliasing and sticky sessions for KV cache alignment, plus a round of community-contributed inspector fixes.

• Feat: add model alias validation, test coverage, and byte-preserving JSON rewrite
• Sticky sessions implementation
• Fix(inspector): copy buffer bytes before pool return to avoid aliasing
• Fix(inspector): incremental scan with token-level skipping for field-order independence

olla-v0.0.24

• Feature: Anthropic agent fixes and improvements
• Default to Olla proxy engine
• Fix duplicate increment
• Fix anthropic tooling bug

olla-v0.0.23

• New Backends: Docker Model Runner and vLLM-MLX
• Support for Anthropic Passthrough on supported backends (vllm etc) so we don't translate in Olla
• Documentation Refinements based on feedback
• Sensible defaults so you can have a lean config file to overide for most users

olla-v0.0.22

• Fix: ensure model_url is used from endpoint config
• Chore: december 2025
• Fix: Alternative method of resolving profile paths
• Alternative way to join paths for OpenAI compatible profiles

olla-v0.0.21

• Feat: path preservation for routing in Olla
• Add preserve_path to ep configuration
• Initial profile consolidation
• Introduce url_builder to abstract out the target path building

olla-v0.0.20

Olla v0.0.20 brings back native LlamaCpp integration and introduces experimental support for Anthropic message translation, making it easier to run local models and route Anthropic-style requests through your own infrastructure.

• Feat: Backend llamacpp
• Feat: anthropic / message logger (development only)
• Feat: Anthropic Message format Support
• Prepare: v0.0.20