Intelligence (Developer)

This chapter explains the Intelligence System used by Tuff Core-App, and how Nexus / plugins interact with it.

What you get

• Multi-provider routing (OpenAI / Anthropic / DeepSeek / SiliconFlow / Local / Custom)
• Capability-based APIs (text.chat, vision.ocr, embedding.generate, agent.run, ...)
• Strategies (provider selection + fallback chain)
• LangChain unified model access layer (provider execution no longer does direct HTTP calls)
• LangGraph agent state machine (session.start -> plan -> execute -> reflect -> finalize)
• Prompt Registry admin surface (/api/admin/intelligence-agent/prompts* + /prompt-bindings*)
• Governance
  • Provider/model compatibility validation
  • API key fast-fail routing
  • Embedding input chunking/truncation
  • OCR payload + storage governance

Where the code lives

• Core runtime
  • apps/core-app/src/main/modules/ai/intelligence-sdk.ts
  • apps/core-app/src/main/modules/ai/intelligence-config.ts
  • apps/core-app/src/main/modules/ai/intelligence-strategy-manager.ts
  • apps/core-app/src/main/modules/ai/runtime/provider-manager.ts
• Providers
  • apps/core-app/src/main/modules/ai/providers/*.ts

See also

• API reference: /docs/dev/api/intelligence
• Architecture: /docs/dev/architecture/intelligence-system
• Configuration: /docs/dev/intelligence/configuration
• LangChain agent orchestration: /docs/dev/intelligence/langchain-agent
• Unified schema migration: /docs/dev/intelligence/schema-migration
• Capabilities & examples: /docs/dev/intelligence/capabilities
• Troubleshooting: /docs/dev/intelligence/troubleshooting