Add OpenAI compatible chat backend

.env.example

@@ -103,6 +103,10 @@ OLLAMA_MAX_CONCURRENT_REQUESTS=1
 # OpenAI
 OPENAI_API_KEY=
 OPENAI_BASE_URL=
+# Backend mode:
+# official = OpenAI responses API
+# compatible = OpenAI-compatible chat.completions servers like llama.cpp
+OPENAI_BACKEND=official
 OPENAI_MODEL=gpt-4.1-nano
 OPENAI_IMAGE_MODEL=gpt-image-1-mini
 OPENAI_TRANSCRIPTION_MODEL=gpt-4o-mini-transcribe
@@ -133,6 +137,7 @@ MCP_SERVERS=
 # OLLAMA_ADDRESS or OLLAMA_BASE_URL.
 # Capability aliases are also supported: IMAGE, THINK, RAG, EMBEDDING,
 # TRANSCRIPTION, STT, TTS.
+# Backend override: OPENAI_BACKEND=official|compatible.
 #
 # Examples:
 # OPENAI_SPEECH_TO_TEXT_MODEL=gpt-4o-mini-transcribe