Repository Issue Activity (beta)

#8680 docs: consolidated provider payload documentation (quality/limits/uptime ranking) + PR gate

#10067 feat(dashboard): interactive keyless-model playground chat

#11018 Memory scaling and 5.1 GB RAM plateau on large multi-provider setups (70+ connections, 2600+ models)

#11025 fix(backend): context guard estimate is ~14x the compression pass on the same body — every sub-225k model rejected for tool-heavy clients

#8075 feat(backend): pluggable external durable-state DB backend (PostgreSQL + MySQL) — unified design & demand thread

#10940 [BUG] `Missing key provider.omniroute.models.{model}.limit.output`

#11024 docs: scale-out large /v1/responses with N independent DATA_DIRs, not one V8 heap

#11023 feat(compression): run default RTK/Caveman off the HTTP isolate so large /v1/responses scale with RAM

#10296 fix(startup): v3.8.49 still ships the /v1/models event-loop hog; #10055 merged but no 3.8.50 image

#11017 docs: DEFAULT_RATE_LIMIT_PER_DAY unset is unlimited; ENVIRONMENT.md still says 1000/day

#11016 [BUG] OMNIROUTE_DISABLE_CREDENTIAL_HEALTH_CHECK still logs scheduler started

#11015 [BUG] SIGTERM never drains in-flight /v1 SSE; Recreate surfaces as 502 Unknown error

#11014 [BUG] event: codex.rate_limits 502s OpenAI SDK; drop flag defaults off

#9985 🔴 Release branch not green: release/v3.8.50

#10852 [Feature] LLM Interlang for token saving

#11006 [Feature] Carry the origin on terminal writes — probe vs production, transient vs terminal

#10950 [BUG] POST /v1/chat/completions defaults to SSE when stream is omitted and Accept: */*, breaking n8n / AI SDK

#9282 🔴 main branch not green

#9506 fix(backend): Claude-native path hoists mid-conversation system messages into system[] and duplicates them

#11005 [BUG] Multiple issues found. I had GPT Luna debug and fix. Here is the summary of the issues.run

#9653 fix(backend): Completed SSE streams recorded as 499 client-disconnected with 0 tokens (finalization race)

#11003 feat(providers): structured free-tier cadence field - one-time vs forever vs daily/monthly/weekly recurring

#11002 fix(providers): dify key check always fails - validator probes /v1/chat/completions but Dify only serves /v1/chat-messages

#11001 docs(providers): bluesminds signup is waitlist-gated manual approval, not instant self-serve key

#11000 fix(providers): hailuo-web token validation fails - hailuo.ai/models now 301s to chat.minimax.io