AI Companion – LLM Player Bot (Ollama, ChatGPT)

Config screen: endpoint and model are now dropdowns with the common providers and their models pre-filled (cheapest first), plus validation that catches a trailing / or /v1 in the endpoint URL. Suggestion lists are readable now, and tooltips no longer cover the form.

OpenAI endpoints now work — requests were failing with unsupported_parameter because OpenAI retired max_tokens in favour of max_completion_tokens. temperature is also omitted for gpt-5.x/o-series, which reject it.

⚠️ On OpenAI, prefer a non-reasoning model (gpt-4.1-nano, gpt-4o-mini, gpt-4.1). Reasoning models count hidden thinking against maxTokens and can return an empty reply at the default of 200 — give them 1000+.

Bundled engine updated to PlayerEngine 1.0.20.