Smart Agent & Server

A high-performance, RAG-orchestrated LLM agent and OpenAI-compatible server with deep MCP integration.

Packages

Package	What it is
@mcp-abap-adt/llm-agent	Core interfaces, types, MissingProviderError, lightweight helpers (CircuitBreaker, FallbackRag, LLM call strategies, ToolCache, adapters, normalizers). Zero provider dependencies.
@mcp-abap-adt/llm-agent-mcp	MCPClientWrapper, McpClientAdapter, createDefaultMcpClient, and MCP connection strategies.
@mcp-abap-adt/llm-agent-rag	RAG/embedder composition — makeRag (async), resolveEmbedder (sync), prefetch helpers, backend factories.
@mcp-abap-adt/llm-agent-libs	Core composition runtime: SmartAgentBuilder, SmartAgent, pipeline, sessions, history, resilience, observability, plugins, skills, makeLlm/makeDefaultLlm.
@mcp-abap-adt/llm-agent-server	Binary only — CLI (llm-agent, llm-agent-check, claude-via-agent) + HTTP SmartServer. Not importable as a library.
@mcp-abap-adt/openai-llm	OpenAI LLM provider (OpenAIProvider).
@mcp-abap-adt/anthropic-llm	Anthropic LLM provider (AnthropicProvider).
@mcp-abap-adt/deepseek-llm	DeepSeek LLM provider (DeepSeekProvider, extends OpenAI-compatible).
@mcp-abap-adt/sap-aicore-llm	SAP AI Core LLM provider via @sap-ai-sdk/orchestration.
@mcp-abap-adt/openai-embedder	OpenAI embeddings (OpenAiEmbedder).
@mcp-abap-adt/ollama-embedder	Ollama embeddings + RAG (OllamaEmbedder, OllamaRag).
@mcp-abap-adt/sap-aicore-embedder	SAP AI Core embeddings (SapAiCoreEmbedder).
@mcp-abap-adt/qdrant-rag	Qdrant vector store RAG (QdrantRag, QdrantRagProvider).
@mcp-abap-adt/hana-vector-rag	SAP HANA Cloud Vector Engine RAG (HanaVectorRag, HanaVectorRagProvider). Optional peer.
@mcp-abap-adt/pg-vector-rag	PostgreSQL + pgvector RAG (PgVectorRag, PgVectorRagProvider). Optional peer.

Quick install

(a) Server-managed declarative (most common)

Install server + exactly the peers your smart-server.yaml references:

# Fully local — Ollama LLM + Ollama embeddings
npm install @mcp-abap-adt/llm-agent-server \
            @mcp-abap-adt/ollama-embedder

# DeepSeek LLM + Ollama embeddings
npm install @mcp-abap-adt/llm-agent-server \
            @mcp-abap-adt/deepseek-llm \
            @mcp-abap-adt/ollama-embedder

# SAP AI Core LLM + SAP AI Core embeddings + Qdrant RAG
npm install @mcp-abap-adt/llm-agent-server \
            @mcp-abap-adt/sap-aicore-llm \
            @mcp-abap-adt/sap-aicore-embedder \
            @mcp-abap-adt/qdrant-rag

A missing peer throws MissingProviderError at startup with an install hint.

(b) Programmatic composition (no YAML)

Install llm-agent-libs + the peers you need. Construct SmartAgent via SmartAgentBuilder in code. Import provider classes from their packages and pass instances to the builder's fluent setters.

npm install @mcp-abap-adt/llm-agent-libs \
            @mcp-abap-adt/llm-agent-mcp \
            @mcp-abap-adt/llm-agent-rag \
            @mcp-abap-adt/deepseek-llm \
            @mcp-abap-adt/ollama-embedder

(c) Core-only (no SmartAgent, no server)

npm install @mcp-abap-adt/llm-agent

Build your own agent against the interfaces exported by core. Supply your own ILlm and IEmbedder implementations.

See docs/MIGRATION-v11.md if you are upgrading from v10.

Documentation

• QUICK_START.md — end-to-end guide: install, config, connect IDE
• ARCHITECTURE.md — architecture reference: thin proxy layer + SmartAgent/SmartServer/pipeline
• INTEGRATION.md — custom interface implementation guide with code examples
• PERFORMANCE.md — RAG, BM25, model selection, token budget tuning
• CLIENT_SETUP.md — connection instructions for Claude CLI, Cline, and Goose
• SAP_AI_CORE.md — SAP AI Core operational guidance and troubleshooting
• DEPLOYMENT.md — production deployment patterns (Docker, systemd, serverless)

Development

# Build project
npm run build

# Run tests
npm run test:all

# Development with hot-reload
npm run dev

# Smart server production entrypoint
npm run start

# Legacy compatibility aliases
npm run start:smart
npm run dev:llm
npm run start:llm
npm run test
npm run test:llm

License

MIT