Packaged AI product · 60 days
Semantic search over your private knowledge. On-premise.
A dedicated RAG instance over your knowledge base (Notion, SharePoint, Drive, wiki, contracts, tickets). Deployed in your cloud or on-premise for strict compliance, with role-based access control, answers that cite their source and auditable logs. Built for legal, consulting and industrial firms with dense documentation.
In brief
What is enterprise private RAG?
Private RAG (Retrieval-Augmented Generation) is a semantic search instance over your internal knowledge (Notion, SharePoint, Drive, wiki, contracts, tickets) with embeddings in pgvector or Weaviate. Answers cite their source, with role-based access control, auditable logs and automatic updates whenever documents change. The infra lives in your cloud or on-premise for strict compliance.
cronuts.digital delivers it for 5,600 €, one-time fee, in 60 days: infra deployed in your cloud or datacenter, ingestion of your knowledge base, a search interface, access control, auditable logs and team training. Optional maintenance from 300 €/month. Built for legal, consulting and industrial firms with dense documentation.
They have worked with cronuts.digital
Your knowledge is spread across wikis, folders, contracts and tickets. Private RAG makes it searchable, with the source cited and each person's permissions respected.
Hire private RAG · 5,600 € →What it does · private RAG capabilities
Six capabilities. Everything included in the fixed price.
It is not a keyword search box or a generic chat. It is a retrieval layer over your documents that answers only with what it finds in them, citing where each answer comes from and respecting who can see what.
Semantic search
Answers with the source cited
Your team asks in natural language. The system retrieves the relevant passages from your documents and the LLM answers using only those, citing the document each answer comes from.
Ingestion and indexing
Your knowledge base, indexed
We connect your sources (Confluence, SharePoint, Drive, Notion, tickets, contracts) and index them in a vector database. PDF, Word, Excel, PowerPoint, HTML and images with OCR.
Access control
Everyone sees only what is theirs
Permissions are replicated from the source. Every query filters results by who is asking, so nobody gets answers built from documents they had no access to.
Continuous updates
Up to date when documents change
The index is updated when the sources change, and the right to be forgotten is applied by selectively re-indexing the affected document.
Private deployment
Your infra, your perimeter
On-premise, in your VPC or in a private cloud in the EU. If your compliance requires it, the LLM is self-hosted too and your data does not leave your perimeter.
Audit trail
Every query, auditable
Every query and every answer is logged, along with the documents used to build it. To review, to correct and to answer to compliance.
Tech stack · sources
On top of your current sources. No lock-in.
The model is chosen to fit your compliance: OpenAI, Anthropic or a self-hosted open-source model. Embeddings on-premise or in a private cloud, a pgvector or Weaviate vector database and connectors to the sources where your documentation already lives.
- LLM: OpenAI · Anthropic · self-hosted open-source (Llama, Mistral) depending on your compliance
- Embeddings: on-premise (Ollama) or private cloud (AWS Bedrock)
- Vector database: pgvector (PostgreSQL) or Weaviate · full control of your data
- Sources: Confluence · SharePoint · Google Drive · Notion · Jira · Zendesk · Salesforce Knowledge
- Formats: PDF · Word · Excel · PowerPoint · Markdown · HTML · images with OCR
- Infra: your datacenter, your VPC or a private cloud in the EU
Common connectors
- 01Wiki
Confluence · Notion
- 02Documents
SharePoint · Google Drive
- 03Tickets
Jira · Zendesk
- 04CRM
Salesforce Knowledge
- 05Data
S3 · Snowflake · BigQuery · Redshift
- 06Custom
Custom connectors via API
Fixed price · 5,600 €
Implementation in 60 days, one-time fee. Operational handoff with SOPs. No mandatory retainer. Optional maintenance from 300 €/month.
How it's deployed
60 days. RAG in production. Your team running it.
Three stages with clear milestones, from discovery with IT and compliance to a progressive go-live. SOPs and team training included.
Weeks 1-2 · Discovery
Sources + permissions + compliance
Sessions with IT and compliance. We list the sources to be indexed and decide where it is deployed (on-premise, VPC or private cloud in the EU) and which LLM is used. DPA signed before kickoff.
Weeks 3-6 · Build
Connectors + ACL + indexing
Connection to the sources and a permission map replicated from the source. Initial indexing in the vector database and retriever tuning so it brings back the right passages.
Weeks 7-8 · Validation + go-live
Real cases + production
Validation with real questions from your team and prompt engineering. Progressive go-live with daily answer sampling. Team training and documented SOPs to run it.
Truths about enterprise RAG
Six things worth knowing before you connect your documents to an AI.
An LLM does not know your internal documents. Without retrieval over your knowledge base, it answers with what it knows in general.
An answer without a source cannot be checked. That is why every answer cites the document it comes from.
If search ignores permissions, it shows people what they should not see. Permissions replicated from the source.
Indexing once is not enough. If the index does not follow the sources, it answers with outdated documents.
When there is not enough grounding, the right answer is «I could not find it», not an improvised one.
Open-ended AI projects with no delivery = black hole. We productise: 60 days, fixed price.
Verified Google reviews
What people who have worked with cronuts.digital say.
Frequently asked questions
What IT, compliance and management ask us before signing.
What is enterprise private RAG and how does it differ from public ChatGPT?
What exactly is included for 5,600 €?
How long does it take to get it into production?
Which sources and formats can it index?
How are each person's permissions respected?
Which LLM is used and where does my data stay?
How do you stop the system from inventing answers?
How is GDPR handled?
Why private RAG and not a generic AI assistant?
Is maintenance mandatory?
Which companies is it right for?
Hire · 5,600 € · implementation in 60 days
Private RAG in production in 60 days.
Fixed price, one-time fee. Scope written into the contract. Deployed in your datacenter, your VPC or a private cloud in the EU. Operational handoff with SOPs and training for your team.