Provider Setup Guide — opencode_initializer v2.0¶

Overview¶

22 AI providers: 19 cloud + 3 local. All configured via src/lib/26-providers.sh and src/lib/18-opencode-json.sh.

Provider Endpoints (verified 2026-07-18)¶

Provider	Endpoint	Model	Status
DeepSeek	api.deepseek.com	deepseek-v4-pro	✅ Free tier
z.ai GLM	api.z.ai/api/paas/v4	glm-5.2	✅ Free tier
OpenRouter	openrouter.ai/api/v1	100+ models	✅ Aggregator
MiniMax	api.minimax.io	MiniMax-M3	🔧 Token Plan credits
xAI Grok	api.x.ai	grok-4.3	✅ Working
Alibaba Qwen	dashscope-intl.aliyuncs.com	qwen3.7-plus	✅ Free tier
DeepInfra	api.deepinfra.com	Llama-4-Maverick	✅ Free tier
Ollama	localhost:11434	llama3.2, qwen3	✅ Local

API Key Sources¶

Provider	Key Source	Env Variable
DeepSeek	platform.deepseek.com/api_keys	DEEPSEEK_API_KEY
MiniMax	platform.minimax.io (Token Plan or PayG)	MINIMAX_API_KEY
xAI	console.x.ai	XAI_API_KEY
z.ai	open.bigmodel.cn	ZAI_API_KEY
OpenRouter	openrouter.ai/keys	OPENROUTER_API_KEY

Air-Gapped (Isolated Circuit Mode)¶

dev isolated on           # Enable: only local providers
ollama pull qwen3:8b      # Pull coding model
ollama pull llama3.2      # Fast chat model
# CPU-only: use llama.cpp with GGUF models
# Qwen3-Coder-30B-A3B, GigaChat-20B

Provider Health Check¶

bash scripts/provider-check.sh   # Tests all configured providers

Model Routing Profiles¶

Profile	Model	Use Case
coding	deepseek-v4-pro	Primary implementation
agentic	deepseek-v4-pro	Tool use, agents
reasoning	claude-opus-4-8	Architecture, planning
fast	deepseek-v4-flash	Quick answers
budget	glm-5.2	Cost-sensitive
ru_cn	glm-5.2	RU/CN language