We use OpenAI, Claude, DeepSeek, Qwen, Gemini, local models, private models, Codex and Claude Code based on task requirements. Strong reasoning models handle complex analysis, Chinese-optimized models handle local consultation, cost-efficient models handle high-volume FAQ, and code agents support development, testing and refactoring.
Engineering includes model routing, tool calling, RAG retrieval, permission policies, audit logs, scheduled jobs, fallback paths, human handoff and dashboards. Delivery formats include websites, H5, WeChat mini programs, Enterprise WeChat, admin consoles, dashboards, iOS/Android native apps and cross-platform apps. Backend stacks include Go, Node.js, Python, TypeScript, PostgreSQL, Redis, object storage, vector search, Docker, Nginx and Linux operations.