v0.4.0

Changelog

• dfb5171 chore(release): bump version to v0.4.0 with 35 CLI tools, A2A, and multi-account pools
• bf45cbb feat(dashboard): add +Add Account modal, 1-click CLI import, and live Fair-Share pool manager
• 48d3b63 feat(parity): port CLIProxy import, domain budgets, responses continuation, and admin APIs
• ceb1cd5 feat(parity): port Ollama tags/version, direct provider routing, and VS Code path tokens
• d554d00 feat(parity): port free-models catalog, chaos test engine, and db-backups manager
• 6dc90ed feat(parity): port full CLI-tools, web-search, A2A protocol, agent-skills and VNC sessions from OmniRoute
• f0dd93b feat(pool): implement Fair-Share Token Leveling pool across multi-account tiers
• f2c9c29 feat(pool): support multiple named token pools (pool1, pool2) with separate leveling and UI cards
• e77ab6d feat(ui): add high-end provider details view, live model catalog, interactive playground, and 1-click Claude Code exposition
• 5b11eed feat(ui): complete overhaul of dashboard to Vercel obsidian design system
• e5268a4 feat(ui): dynamic multi-target combo chain builder and clean topology graph
• 99b15bf feat(ui): expand provider catalog to 45+ providers with dynamic BaseURL and optgroups
• 9f83792 feat: async batch jobs engine and compliance audit logging (/api/batches, /api/compliance/audit-log)
• aa88f31 feat: automated provider discovery scanner and results engine (/api/discovery/scan, /api/discovery/results)
• b8b31b2 feat: call log filter options aggregator API (/api/usage/call-logs/filters)
• 8439075 feat: call log purge cleanup engine and /api/settings/purge-call-logs endpoint
• 5d4b2d7 feat: call logs filter engine and /api/usage/call-logs endpoint
• b9e6caf feat: conductor orchestration fleet, tasks, and skills registry engine (/api/conductor/*, /api/skills)
• c5d4c4f feat: connection resilience state and cooldown monitor (/api/resilience/cooldowns, /api/resilience/connections)
• 7cc5437 feat: embedded SQLite call log persistence with WAL journal mode
• 22796fe feat: evaluation suites, run scorecards, and local corpus RAG settings (/api/evals, /api/settings/local-corpus)
• 0b4f570 feat: event webhooks engine and /api/webhooks CRUD endpoints
• 4c36adb feat: fallback policy chains engine and /api/fallback/chains API
• d215de2 feat: live provider radar status, intelligence catalog, and sync engine (/api/radar/*)
• aea6b13 feat: model alias rewrite engine and /api/settings/model-aliases API
• 19ace80 feat: model capability overrides engine and /api/model-capability-overrides API
• b132a8d feat: omniroute provider catalog, account fallback priority reordering, and connection combo UI
• 64f4ab5 feat: persistent system settings API with SQLite key_value backend
• 9056cc4 feat: reasoning thinking budget policy engine and /api/settings/thinking-budget API
• 930b24b feat: sensitive data guardrails and AgentBridge DNS hosts manager
• a27eeac feat: shared quota pools engine and /api/quota/pools API
• 3600550 feat: system prompt injection, IP filtering, and semantic/prompt cache settings APIs
• 60f2808 feat: token cost pricing engine and /api/pricing CRUD endpoints
• b917280 feat: token health badge aggregator and quota groups API
• bd259bc feat: tool parameter sanitizer, session affinity eviction, MITM CONNECT tunnel, and cost ledger
• ede7a9c feat: upstream proxy registry and /api/settings/proxies endpoints
• db4facc feat: web-cookie session harvester and format normalizer
• f5889bd fix(antigravity): accurately count tool_result bytes in token estimation and sanitize message sizes
• 02dcfa4 fix(claude): expose gemini as claude-gemini-3.1-pro in gateway discovery and default fable model
• 1b26cca fix(claude-code): register 7th model gemini-3.1-pro with apinexbond provider and multi-route compatibility
• 91011ba fix(providers): remove fixed 120s client timeout on streaming LLM connections
• be6b4dd fix(proxy): eliminate artificial limits and truncation, ensure 100% full-fidelity passthrough
• cb331b7 fix(proxy): expand guardrail context limit to 800k chars and preserve history turns in kiro pruner
• 8dee13d fix(router): auto-prune context exceeding 1M tokens and prioritize active Antigravity in combos
• a6e1ced fix(router): prevent copilot token fail loop, fix model alias fallback and kiro turn pairing
• 72accd6 fix(schema): sanitize empty enum items in Gemini tool schemas to prevent 400 during /compact with large MCP toolsets
• dad5b73 perf(antigravity): optimize context pruning to O(N) single-pass integer calculation
• f6d895f refactor: eliminate all legacy naming references across code, comments, and web dashboard
• 6b2c417 test(e2e): 121/121 live tests passing, expose standard model aliases and Anthropic event stream tests
• 763eb65 test(probe): add advanced deep probe suite covering Ollama, VS Code, budgets, chaos, and 120-turn transcripts