Spaces:
Sleeping
Sleeping
Commit History
-
feat: add Cloudflare Workers AI fallback for kimi models
OpenCode- Add api/lib/cloudflare.js with credential rotation - Try CF first for kimi-k2.7/kimi-k2.6, fallback to Kimchi on failure - Map kimi-k2.7 -> @cf/moonshotai/kimi-k2.7-code - Apply auto-continue to CF responses - Add provider tracking in stats and headers - Update health check with cloudflare status - Export requestUpstream/requestUpstreamStreaming from proxy.js3f8984a
-
feat: non-streaming auto-continue, higher timeout, default max_tokens
OpenCode- Increase FETCH_TIMEOUT_MS and AUTO_CONTINUE_TIMEOUT_MS to 120s - Increase AUTO_CONTINUE_MAX to 5 - Add DEFAULT_MAX_TOKENS=16384 when client omits max_tokens - Add non-streaming auto-continue when finish_reason=length - Merge continued responses and usage stats - Add X-Proxy-Continued and X-Proxy-Continue-Attempts headersd584275 -
feat: add PROXY_API_KEY validation
OpenCode- Add api/lib/auth.js helper - Protect /v1/chat/completions with PROXY_API_KEY - Protect /health with PROXY_API_KEY - /v1/models remains public - /dashboard remains protected by DASHBOARD_PASSWORD94703c2
-
fix: detect stream completion without [DONE] marker
dandia67c779 -
feat: auto-continue + keepalive for streaming
dandi0d4ab93
-
fix: move stats to completions handler for shared /tmp access
dandi0d0e178 -
fix: remove setKeyTotal import (read from env now)
dandibbe6bde -
feat: API key stats, error capture, key states tracking
dandi153b6dc -
feat: dashboard with stats, recent requests, console log, login
dandi8c87fee -
feat: proper SSE streaming via pipe for stream:true
dandi9816b39 -
refactor: read body in proxy, return text directly
dandi4e18ea2 -
fix: key rotation - zero delay on 402, 55 retries, handler timeout 55s
dandi1788deb -
perf: reduce timeouts to fit Vercel 60s limit
dandi4da59fb -
feat: key rotation on retry - rotate key per attempt
dandi8b83ebd -
fix: require paths to api/lib/
dandif611836 -
init: kimchi proxy with multi-key rotation
dandib65c987