@mcp-abap-adt/llm-agent-rag

RAG and embedder composition for the SmartAgent runtime.

Exports

• makeRag(cfg, options): Promise<IRag> — async, dynamic-imports the configured backend (e.g. OllamaRag).
• resolveEmbedder(cfg, options): IEmbedder — synchronous, requires prior prefetchEmbedderFactories(...).
• prefetchEmbedderFactories(names): Promise<void>, prefetchRagFactories(names): Promise<void> — warm-up helpers.
• resolvePrefetchedEmbedder(name, opts), resolveRag(name, opts) — synchronous resolvers from the prefetched cache.
• builtInEmbedderFactories — registry record of built-in embedder factories.
• Types: RagResolutionConfig, RagResolutionOptions, EmbedderResolutionConfig, EmbedderResolutionOptions, EmbedderFactoryOpts.

Two patterns

Common case (one-shot async resolution)

import { makeRag } from '@mcp-abap-adt/llm-agent-rag';
const rag = await makeRag(
  { type: 'ollama', model: 'llama3', /* ... */ },
  { embedder: yourEmbedder, breaker: yourBreaker },
);

Hot-path consumers (prefetch once, sync resolve)

import {
  prefetchEmbedderFactories,
  prefetchRagFactories,
  resolveRag,
} from '@mcp-abap-adt/llm-agent-rag';

await prefetchEmbedderFactories(['openai']);
await prefetchRagFactories(['qdrant']);

// Inside a hot loop:
const rag = resolveRag('qdrant', { embedder, breaker, /* ... */ });

Optional peer dependencies

Install only the backends you use:

| Backend | Package | |---|---| | OpenAI embeddings | @mcp-abap-adt/openai-embedder | | Ollama embeddings | @mcp-abap-adt/ollama-embedder | | SAP AI Core embeddings | @mcp-abap-adt/sap-aicore-embedder | | Qdrant vector store | @mcp-abap-adt/qdrant-rag | | HANA vector store | @mcp-abap-adt/hana-vector-rag | | Postgres+pgvector | @mcp-abap-adt/pg-vector-rag |

Missing backends throw MissingProviderError (from @mcp-abap-adt/llm-agent) at first use.

See docs/ARCHITECTURE.md for the full SmartAgent package layout.