Applications

VAST Foundation Stacks

Advanced

NVIDIA AI Blueprints get you a working demo; the gap to production is the integration tax. VAST Foundation Stacks are open-source implementations that close it - running RAG, AI-Q deep research, and Video Search & Summarization as production pipelines on the VAST AI OS.

From Blueprint to production

An NVIDIA AI Blueprint is a reference workflow - a runnable starting point built from NIM microservices. Getting it into reliable, secure, scalable production is the hard part: data pipelines, vector storage, eventing, orchestration, security, ops. VAST Foundation Stacks pre-package that on one platform, so a Blueprint becomes something you git clone and deploy. They are open source at github.com/vast-data/cosmos-labs.

Blueprint → Foundation Stack

An NVIDIA Blueprint gets you a working demo. A Foundation Stack adds the production plumbing - so the gap from pilot to production closes to a git clone.

VAST Foundation Stack

production-ready, on the VAST AI OS

Serverless ingest pipeline

Event-driven DataEngine functions + triggers - data is processed the moment it lands.

One unified store

VastDB holds documents, vectors, metadata, and conversation state - no separate vector DB to wire up.

Governed & secure

Tenant-aware auth, permission-filtered retrieval, lifecycle management built in.

Open-source, deployable

Public on github.com/vast-data/cosmos-labs with deploy scripts - clone and run on cloud or on-prem.

One shape under every stack

However different the jobs look, all three stacks share the same anatomy: a serverless ingest pipeline on DataEngine fills a unified VastDB store, and a Kubernetes app serves users by calling NVIDIA NIMs. Learn the shape once and all three click into place.

Kubernetes app

the serve plane

A web UI plus a REST/agent API (FastAPI), deployed with the included scripts. This is what users actually touch.

DataEngine ingest

serverless, event-driven

Triggers + functions fire the instant data lands in S3 - no always-on service. Chunk, embed, segment, write.

VastDB

one unified store

Vectors, document metadata, and conversation state in a single database - no separate vector DB to operate.

NVIDIA NIM

the AI, on GPUs

Embedding, reranking, LLM, and vision-language microservices - run locally or against the NVIDIA API.

The three pipelines, running

Each launch stack maps to a different NVIDIA Blueprint. Step through the real ordered stages - the function and model names come straight from the VAST blueprint repos.

RAG

The foundational retrieve-augmented-generation pipeline: ingest and embed documents, then answer questions grounded on the most relevant chunks with citations.

Enterprise RAG Blueprint

AI-Q Research Assistant

An agentic deep-research assistant that plans sub-questions, retrieves repeatedly, reflects on gaps, and writes a cited report - RAG wrapped in an agent loop.

dataengine-research-assistant-blueprint

VSS — Video Search & Summarization

Ingests live or archived video, describes each segment with a vision-language model, and makes it searchable and summarizable in natural language.

dataengine-vss-blueprint

The three Foundation Stacks - one production shape

Pick a stack and watch its real pipeline run. Each one is a serverless ingest pipeline that fills a unified VastDB store, plus a serve plane calling NVIDIA NIMs. Same backbone, three very different jobs.

Based on the NVIDIA Enterprise RAG Blueprint.

Ingest pipeline (serverless)

Upload / Sync
Trigger
Extract + Chunk
Embed (NIM)
Store

Query (read path)

Embed query
Retrieve (ANN)
Rerank (NIM)
Generate + cite
DataStore (S3)Upload / Sync

Documents land in an S3 bucket, or are pulled from Confluence / GDrive / SharePoint by the Sync Engine.

DataStore (S3)DataEngine (serverless)VastDB (unified store)NVIDIA NIM / GPU

One-shot retrieve → generate. The foundation every other stack builds on.

RAG vs AI-Q: one-shot vs a research loop

The clearest way to understand AI-Q is to see it as RAG with an agent wrapped around it. Plain RAG retrieves once and generates one answer. AI-Q plans the question into sub-questions, calls retrieval several times with different angles (each call is a full RAG retrieve-and-rerank), reflects on what is missing, and only then writes a structured, cited report. The same ingest pipeline and the same VastDB store feed both - VastDB also persists the agent's conversation state, so context never goes stale.

RAG — one pass

Embed query → retrieve top-30 → rerank to 10 → generate one grounded answer with citations. Fast, predictable, great for direct questions over your docs.

AI-Q — a research loop

Plan → retrieve repeatedly (multi-angle) → reflect on gaps → synthesize a long-form report. Trades latency for depth - built for “research this for me,” not “answer this.”

Go deeper

Foundation Stacks tie together the building blocks taught across the site - the retrieval mechanics, the agent patterns, the serverless engine, and the real-time RAG platform they run on.