ChatRAG
A conversational assistant that answers questions about a company's internal documentation (resumes, candidate history, training spend, expenses) with verifiable citations for strategic decisions. Official landing page with live demo.
ChatRAG
A conversational assistant that answers questions about a company's internal documentation (resumes, candidate history, training spend, expenses) with verifiable citations for strategic decisions. Official landing page with live demo.
What needed to be solved
A company's critical information lives scattered across PDFs, spreadsheets and internal documents. Finding a reliable answer takes minutes or hours, and a generic AI assistant doesn't know those documents or where to cite each fact from.
How we tackled it
We designed a local-first conversational assistant built on retrieval-augmented generation (RAG): it indexes the company's real documentation and answers by citing the exact fragment each claim comes from, instead of generating plausible text with no backing. Instead of a single generic bot, ChatRAG runs on 8 department-specialized agents (Supervisor, Management, HR, Finance, Logistics, Marketing, Legal and Tech), each with access limited to its own area's documentation.
System components
Indexing and vectorization of internal documents (resumes, candidate history, training spend, expenses)
Retrieval-augmented (RAG) engine with verifiable citations per answer
8 department-specialized agents with role-based segmented access (RBAC)
A Supervisor agent that orchestrates and routes to the right specialist
Per-agent tools: financial calculator, contract and certificate PDF generation, dispatch guides, report export
Safety switches to disable internet access or code execution in deployments with sensitive data
Conversational interface for executive queries
Official landing page with interactive demo
Talk to your company's files without them ever leaving your company.
ChatRAG is new and almost nobody knows it. It is a local-first platform for AI, RAG and agents per department: internal documents (contracts, reports, SOPs, policies) are vectorized and queried on your own infrastructure, with one agent per area and an auditor that verifies every answer before delivering it. It was born on an initial technical base by Johann Pérez Giraldo, and José Santamaría adapted it at jsadsAI into an enterprise solution by department.
It has no closed price list: adoption goes in stages (diagnosis, private MVP, Enterprise) and each is quoted by scope. The scale is jsadsAI's: a private deployment that covers the work and a small margin, not a per-seat license like corporate AI platforms.
Visit ChatRAG ↗Where AI understands your company's internal knowledge
Companies have sensitive files scattered around and generic chatbots know nothing about internal context; uploading documents to public platforms is a risk. ChatRAG turns those documents into queryable business intelligence, with multi-user isolation and quality auditing.
Each agent is scoped to its domain: Supervisor, Management, HR, Finance, Marketing, Legal, Logistics and Tech. Ask about a balance sheet and the finance agent searches only its sources; ask about a contract and the legal agent takes over. A QA node checks consistency before answering.
It runs on local models (Qwen, Ollama) over a private vector database (ChromaDB) and an on-premise multi-agent router: a closed flow from the internal document to the audited answer, without passing through public clouds.
- Companies with sensitive documentation that cannot leave their environment.
- Teams that lose hours searching for information in scattered files.
- Legal, finance, HR and operations areas that need sourced answers.
- Companies that want AI without a mandatory dependency on public APIs.
What it includes
As published on chatrag.jsadsai.com.
Agents per department
Supervisor, Management, HR, Finance, Marketing, Legal, Logistics and Tech, each with its own sources.
Private local-first deployment
The platform runs on your infrastructure; internal knowledge stays in your environment.
Access control (RBAC)
Roles and permissions to define who queries what within each department.
Per-user isolation
Each user operates in their own context, with no cross-leakage of information.
Local models
Qwen and Ollama running locally, with no mandatory dependency on public APIs.
Private vector database
Your documents' embeddings live in ChromaDB under your control.
QA auditing
An auditor in the orchestrator verifies every answer and its sources before delivery.
Public demo, no sign-up
Query a contract, a Q3 report or an SOP with fictitious sources and watch the full flow.
Open ChatRAG Lab ↗From the internal document to the audited answer
Internal files
Your company's documents are vectorized in your environment.
Private vector database
ChromaDB stores the embeddings under your control.
Multi-agent router
The query goes to the agent of the right department.
Secure answer
The QA auditor checks consistency and sources before delivery.
A staged adoption path
As published on chatrag.jsadsai.com: start with a diagnosis and scale toward a full private deployment. Each stage is quoted by scope.
Enterprise AI Diagnosis
Quoted by scope
Process and data audit: mapping of sources and documents, use cases per area and an adoption roadmap.
Request a demo ↗Private MVP · most requested
Quoted by scope
Local-first base installation, RAG over your documentation and the main agents per department.
Request a demo ↗Enterprise
Quoted by scope
Custom connectors and integrations, advanced RBAC and the full department architecture.
Request a demo ↗No public prices: each stage is quoted by scope. The public demo uses fictitious sources.
Privacy, context and traceability
- Data never goes to OpenAI or Anthropic. Local-first architecture: documents are processed on your infrastructure or on private dedicated servers.
- Sourced answers. Every answer is verified against the documents before delivery.
- One agent per domain. Each area queries only the sources that belong to it.
- Acknowledged origin. Initial technical base by Johann Pérez Giraldo; enterprise adaptation by José Santamaría at jsadsAI.
- The public demo uses fictitious sources and does not process real documents.
- It is not a generic cloud chatbot: it requires deployment on your infrastructure or on dedicated servers.
- No public prices: each stage is quoted by scope.
All of ChatRAG, link by link
Every section lives on the project's official site. From here you go straight to it.
What people ask about ChatRAG
ChatRAG is a local-first platform for private AI, RAG and agents per department created by jsadsAI | José Santamaría on an initial technical base by Johann Pérez Giraldo: it turns a company's internal documents into auditable answers, with private deployment, local models and access control. It is live at chatrag.jsadsai.com.
It has no public prices. Adoption goes in stages (Enterprise AI Diagnosis, Private MVP and Enterprise) and each is quoted by scope.
No. ChatRAG runs on a local-first architecture: documents are vectorized and processed on your own infrastructure or on private dedicated servers.
The platform routes queries: a balance sheet is answered by the finance agent with its sources; a contract, by the legal agent. Each agent is scoped to its domain.
Every answer goes through a QA node (auditor) in the orchestrator that checks technical consistency and sources before delivering it.
Yes. ChatRAG Lab is a public demo with no sign-up that uses fictitious sources and shows the real flow: intent, agent, document retrieval, evidence-backed answer and quality control.
Official site and interactive demo live at chatrag.jsadsai.com; long project page at /en/proyectos/chatrag.
Services this project demonstrates
These are not related topics: they are the systems actually built in this project, available as services.
Category: AI Agent · RAG over enterprise documentation. To learn the full scope and how to apply a similar approach to your business, write to us — no strings attached.
Have a similar idea?
Let's design the architecture to turn your idea into a platform, automation or profitable digital asset.
A real human replies. No spam. Free initial diagnostic.