NNEXUSRAG INTELLIGENCE

Answers your organization can trust.

NEXUS RAG turns your documents into a governed knowledge base with a live knowledge graph, and serves grounded, cited answers through an OpenAI-compatible API — for people, apps and AI agents.

Create your workspace →See the API
9file formats
100%cited answers
/v1OpenAI-compatible

Everything between your documents and a trustworthy answer.

⌁

OpenAI-compatible API

Point any OpenAI SDK, agent framework or tool at /v1 — chat completions with streaming, embeddings and models. Only base_url, api_key and model change.

▤

Every format, with provenance

PDF, DOCX, PPTX, XLSX, CSV, Markdown, HTML, JSON and text. Page, slide and sheet are preserved on every chunk, so every answer points to its exact source.

◎

Hybrid retrieval + reranking

Dense vectors (pgvector HNSW) fused with BM25 lexical search, metadata filters, graph expansion and a cross-encoder reranker.

⌘

Knowledge graph, without hallucinated edges

Entities and typed relationships are extracted from your documents — each relationship stores the exact quote that proves it, plus a confidence score.

⌖

Semantic & geographic maps

See your knowledge as a landscape of semantic regions, or on a world map when documents mention places.

⇄

Retrieval traces

Inspect every stage of every answer: embedding, vector hits, BM25, graph expansion, reranker scores, context and model latency.

From upload to grounded answer in three steps.

01

Drop knowledge

Upload files or whole folders. Ingestion runs asynchronously: parse → chunk → embed → index → connect.

02

Ask or integrate

Use the Query Lab, or call /v1/chat/completions from your apps and agents with a scoped API key.

03

Trust the answer

Inline citations link to the exact chunk. Invalid citations are detected and discarded automatically.

Drop-in for the OpenAI SDK.

Use rag-default for all knowledge, rag-<collection> for one knowledge base, or rag/<model> to choose the upstream LLM. Streaming responses, scoped API keys, tenant isolation and per-document access control are built in. Retrieval metadata arrives under rag.

  • POST /v1/chat/completions — grounded chat, streaming
  • POST /v1/rag/search — ranked chunks, scores, graph neighbours
  • POST /v1/documents — upload & ingest
  • GET /v1/graph · /v1/traces/{id} · /v1/embeddings
from openai import OpenAI

client = OpenAI(base_url="https://rag.nexuslayer.eu/v1",
                api_key="rag_live_…")

r = client.chat.completions.create(
    model="rag-default",
    messages=[{"role": "user",
               "content": "How does our DR failover work?"}])

print(r.choices[0].message.content)   # grounded answer [1][2]
print(r.model_extra["rag"]["sources"]) # citations + quotes

Bring your knowledge online today.

Private workspace, free to start. Your documents are never used to train models.

Create your workspace →