PraisonAI · Issues· 88 open

Gateway horizontal-scale story is asymmetric: turn-lock has a Redis backend but inbound idempotency and the session store do not, so multiple replicas double-process the same message

Gateway does not shed live sessions under memory pressure: the cgroup-aware eviction planner is used only for the bot warm-agent cache, not for WebSocketGateway sessions

Gateway reply delivery is not durable end-to-end: inbound is marked complete before the reply is sent, and effectively-once delivery is opt-in (Slack-only)

Gateway approvals have no per-operator visibility or spawn-lineage audience — every APPROVALS-scoped operator sees (and is notified of) every agent's pending requests

src/praisonai wrapper: multi-tenant chat_history bleed, per-loop httpx pool leak, and a framework-availability probe that pins on the first transient miss

praisonaiagents core: default-path chat_history rollback still corrupts concurrent turns, tool guardrails block the event loop, and hooks leak across unrelated Agent instances

Core SDK: `param_resolver` duplicates the `canonical_preset_key` normalisation inline instead of calling the single source of truth

Gateway has no shared HTTP ingress: each webhook channel binds its own port, so running multiple webhook bots needs multiple ports/public URLs (and collides on 8080)

Gateway hot-reload has no candidate validation or rollback — a runtime-invalid config can take down a running gateway

CLI: auto session titles are built but not wired into the default session store — `session list` shows opaque IDs instead of generated titles

praisonaiagents core: MCP server-registry never shrinks (stale capability gating), async goal loop blocks the event loop on every judge call, and FileLock's stale-lock recovery busts a live lock

IRC bot crashes on inbound message due to wrong `BotMessage` field names

Prepaid Tavily vs TAVILY_API_KEY for Claw search?

Wrapper gaps: recipe rate-limiter bypassable+worker-unsafe, multiedit corrupts files on crash, __main__ Typer/legacy split billed to LLM

Core SDK: cross-tenant token-cost collisions, dropped litellm success-callbacks, and NoneType-masked import failures