0.4.22

What's Changed

• fix(anthropic-stream): avoid bare/duplicate message_stop on OpenAI upstream by @adilhafeez in #898
• fix: prevent index-out-of-bounds panic in signal analyzer follow-up by @adilhafeez in #896
• Add claude-opus-4-7 to anthropic provider models by @adilhafeez in #901
• Fix request closures during long-running streaming by @adilhafeez in #899
• Improve obs model name matching, latency metrics, and error reporting by @adilhafeez in #900
• Add Prometheus metrics endpoint and Grafana dashboard for brightstaff by @adilhafeez in #904
• add overrides.disable_signals to skip CPU-heavy signal analysis by @adilhafeez in #906
• signals: feature parity with the latest Signals paper. Porting logic from python repo by @syedhashmi in #903
• add jemalloc and /debug/memstats endpoint for OOM diagnosis by @adilhafeez in #885
• Add first-class ChatGPT subscription provider support by @Spherrrical in #881
• feat(providers): add Vercel AI Gateway and OpenRouter support by @Spherrrical in #902
• docs: align signals page with paper taxonomy by @syedhashmi in #910
• fix(routing): auto-migrate v0.3.0 inline routing_preferences to v0.4.0 top-level by @Spherrrical in #912
• signals: restore the pre-port flag marker emoji (🚩) by @syedhashmi in #913
• Bump version to 0.4.21 by @Spherrrical in #911
• fix(cli): drop function-local import yaml shadowing module global in up by @Spherrrical in #916
• fix(config): accept vercel and openrouter as provider_interface values by @Spherrrical in #915
• Bump version to 0.4.22 by @Spherrrical in #917

Full Changelog: 0.4.20...0.4.22

release 0.4.20

What's Changed

• add Plano agent skills framework and rule set by @Spherrrical in #797
• Add DigitalOcean as a first-class LLM provider by @adilhafeez in #889
• Zero-config planoai up: pass-through proxy with auto-detected providers by @adilhafeez in #890
• planoai obs: live LLM observability TUI by @adilhafeez in #891
• fix: passthrough_auth accepts Anthropic x-api-key and normalizes to upstream format by @adilhafeez in #892
• fix: truncate oversized user messages in orchestrator routing prompt by @adilhafeez in #895
• release: bump version to 0.4.20 by @adilhafeez in #897

Full Changelog: 0.4.19...0.4.20

release 0.4.19

What's Changed

• Redis-backed session cache for cross-replica model affinity by @Spherrrical in #879
• use plano-orchestrator for LLM routing, remove arch-router by @adilhafeez in #886
• release 0.4.19 by @adilhafeez in #887

Full Changelog: 0.4.18...0.4.19

0.4.18

What's Changed

• Add first-class Xiaomi provider support by @Spherrrical in #863
• Model affinity for consistent model selection in agentic loops by @adilhafeez in #827
• release 0.4.18 by @Spherrrical in #878

Full Changelog: 0.4.17...0.4.18

0.4.17

What's Changed

• feat(web): merge DigitalOcean release announcement updates by @Spherrrical in #860
• feat(web): merge DigitalOcean release announcement updates by @Spherrrical in #862
• fix: resolve all open Dependabot security alerts by @adilhafeez in #866
• Publish docker images to DigitalOcean Container Registry by @adilhafeez in #868
• support developer role in chat completions API by @adilhafeez in #867
• ci(e2e): stabilize preference demo CI execution by @Spherrrical in #865
• release 0.4.17 by @adilhafeez in #869

Full Changelog: 0.4.16...0.4.17

0.4.17-rc1

What's Changed

• feat(web): merge DigitalOcean release announcement updates by @Spherrrical in #860
• feat(web): merge DigitalOcean release announcement updates by @Spherrrical in #862
• fix: resolve all open Dependabot security alerts by @adilhafeez in #866
• Publish docker images to DigitalOcean Container Registry by @adilhafeez in #868
• support developer role in chat completions API by @adilhafeez in #867

Full Changelog: 0.4.16...0.4.17-rc1

0.4.16

What's Changed

• Update black hook for Python 3.14 by @Spherrrical in #857
• Polish planoai up/down CLI output by @Spherrrical in #858
• replace production panics with graceful error handling in common crate by @adilhafeez in #844
• fix: route Perplexity OpenAI endpoints without /v1 by @Spherrrical in #854
• Handle null prefer in inline routing policy by @Spherrrical in #856
• release 0.4.16 by @Spherrrical in #859

Full Changelog: 0.4.15...0.4.16

0.4.15

What's Changed

• expand configuration reference with missing fields by @Spherrrical in #851
• model routing: cost/latency ranking with ranked fallback list by @adilhafeez in #849
• restructure model_metrics_sources to type + provider by @adilhafeez in #855
• release 0.4.15 by @adilhafeez in #853

Full Changelog: 0.4.14...0.4.15

release 0.4.14

What's Changed

• the orchestrator had a bug where it was setting the wrong headers for… by @salmanap in #839
• release 0.4.14 by @adilhafeez in #840

Full Changelog: 0.4.13...0.4.14

release 0.4.13

Highlights

Key Features

• Output Filter Chain — Adds support for output filter chains, enabling guardrails and processing on LLM responses in addition to the existing input/prompt filters (#822)
• Kubernetes Deployment Support — New K8s manifests and docs for self-hosted Arch-Router (vLLM) with GPU support, including deployment YAMLs and in-cluster routing config (#831)

Improvements

• Brightstaff Refactor — Extracted AppState struct, cleaner error propagation, graceful shutdown (ctrl-c), and reorganized handlers into agents/ and llm/ modules (#736)
• Reliable Container Startup — Separated config generation from process startup in supervisord, ensuring configs are fully rendered before services launch (#838)

Security

• CVE-2026-0861 — Fixed glibc vulnerability via apt-get upgrade in Docker image (#832)

List of changes

• add k8s deployment manifests and docs for self-hosted Arch-Router by @adilhafeez in #831
• fix CVE-2026-0861: upgrade glibc via apt-get upgrade in Dockerfile by @adilhafeez in #832
• add output filter chain by @adilhafeez in #822
• refactor brightstaff by @adilhafeez in #736
• separate config generation from process startup in supervisord by @adilhafeez in #838
• release 0.4.13 by @adilhafeez in #837

Full Changelog: 0.4.12...0.4.13