Your knowledge.Your AI.Your servers.
Enterprise-grade generative AI within reach of your SMB. Deploy a RAG assistant over your own documents on a single machine, no per-user licences and not a single byte leaving your company.
The problem
Many SMBs want to use AI, but the reality is that generic solutions don't understand your internal context, don't work for critical processes, and force you to upload confidential documents to servers you don't control. The result: wasted time, legal exposure, and competitive risk.
Loss of Control
Confidential documents in the hands of external providers with no guarantees on how they are stored, used, or retained.
Legal Exposure
GDPR, LOPD, and sector regulations violated the moment sensitive data leaves your network.
Competitive Risk
Trade secrets, strategies, and client data accessible to third-party AI platforms.
One device. Your network. All your internal knowledge. An assistant that adapts to your company.
TENGWAR arrives as a complete system: hardware and software delivered as a single unit. We connect it to your network, ingest your document library, and your team starts receiving answers with precise citations — all without leaving the building.
Your Documents
NVIDIA DGX · 128 GB
Cited Answer
833 k€ of principal was amortized (visible in CashFlow > Loan Repayments).
| Entity | Long term | Short term | Total |
|---|---|---|---|
| Santander | 865 k€ | 485 k€ | 1,350 k€ |
| CaixaBank | 419 k€ | 148 k€ | 567 k€ |
| ICO | 200 k€ | 200 k€ | 400 k€ |
| Total | 1,484 k€ | 833 k€ | 2,317 k€ |
Every answer, auditable to the page.
TENGWAR doesn't make things up. Every claim is anchored to a specific section of the original document. A click on the citation opens the PDF, on the exact page, with the excerpt highlighted.
TENGWAR in action
A short promotional walkthrough of the application: chat, citations, documents, and the private workflow your team can use every day.
Who Is It For
Organizations that need private, secure, and instant access to their accumulated knowledge.
Ideal Sectors
Construction & Engineering
Contracts, project specs, blueprints, safety regulations, compliance documentation
Legal Firms
Case law, client files, contracts, regulatory frameworks, due diligence archives
Healthcare
Clinical protocols, patient records, pharmaceutical documentation, research papers
Manufacturing & Industry
Technical manuals, quality procedures, maintenance logs, supplier specifications
Professional Services
SOPs, internal knowledge bases, SLAs, project documentation, training materials
Energy & Utilities
Inspection reports, regulatory compliance, asset management, safety procedures
Use Cases
Internal Knowledge Base
An employee asks a question about company policy. TENGWAR finds the answer in seconds and cites the exact document, section, and page number.
"What is our policy on remote work for contractors?" → Answer from HR Manual v3.2, page 14
Contract Intelligence
Your legal team needs to find every contract expiring this quarter with automatic renewal clauses. TENGWAR searches across hundreds of documents instantly.
"Which contracts expire before July with auto-renewal?" → 12 contracts identified with clause references
Automated Report Generation
A project manager needs a weekly executive summary based on multiple progress reports. TENGWAR generates a formatted Word document with citations.
"Generate an executive summary from this month's site reports" → Downloadable DOCX with source references
Multilingual Document Access
Your team works with documentation in multiple languages. Ask questions in Spanish, get answers from English, French, or German source documents — automatically.
"¿Cuáles son los requisitos de seguridad del cliente?" → Answer from English safety requirements document
Operations & Compliance Support
A field technician needs the approved procedure for a specific situation. TENGWAR provides the exact steps from the official SOP and cites the source.
"Emergency shutdown procedure for Unit 4" → SOP-2024-087, Section 3.2, Steps 1-8
Powerful Features
Everything your organization needs to build a secure, intelligent knowledge base — from document ingestion to AI-generated reports.
Smart Document Ingestion
Upload any document. Our 18-phase pipeline handles structure detection, OCR, language identification, translation, and semantic chunking — automatically.
Hybrid Semantic Search
Finds answers by meaning, not just keywords. Combines vector search with traditional text matching for the most accurate results across your entire document library.
Cross-Language Search
Ask questions in any language and find answers regardless of the document's original language. Automatic detection and translation across 100+ languages.
Agentic AI Capabilities
Goes beyond simple Q&A. TENGWAR autonomously decides when to search documents, browse the web, generate files, or combine multiple sources to answer complex questions.
Desktop & Web Access
Native Windows and macOS desktop apps with mDNS auto-discovery on your LAN, hash-based folder sync, TLS pinning for sniff-resistant traffic, and signed auto-updates. Plus a full web client for browser-based access from any device on your network.
Live Streaming Responses
See answers as they are generated in real time — no waiting for the full response. Works across all client surfaces.
On-Demand Document Generation
Generate Word, Excel, and PDF documents directly from AI responses. Create reports, summaries, and exports without leaving the chat.
Connectors for Your Existing Tools
A growing library of ready-made connectors for the systems your business already runs on — ERPs, public registries, e-invoicing, mail and calendar. Bring real-time business data into every answer, no copy-paste required.
Enterprise-Grade Security
Built for organizations where data confidentiality is not optional.
Authorization Levels
Public
General access for all authenticated users
Confidential
Internal company information — restricted distribution
Secret
Sensitive data: contracts, strategy, financials
Top Secret
Maximum confidentiality — named-individual access only
Department Isolation
Users only access documents from their assigned departments. Combined with clearance levels for precise, auditable access control.
Flexible Authentication
Local login with BCrypt hashing, or SSO via Microsoft Entra ID and Google OAuth2. RSA asymmetric JWT signing. Per-user credential encryption with AES-256-GCM.
Full GDPR/LOPD Compliance
All data processed locally on your hardware. Zero transfers to third parties. Complete lifecycle control — from ingestion to deletion.
Governance — woven through every step
Not a separate layer — it surrounds every interaction with your data.
- Auth gate
- Bounded execution
- Tool authorization (default-deny)
- SSRF guard
- Audit log
- Per-document ACL re-check
- Credential encryption
Adapts to You — Privately
TENGWAR adapts to each user without sending anything to a third party. The implicit (memory) and the explicit (prompt) — both feeding the same private workspace.
Two ways to adapt
TENGWAR learns from how you work, and you can also tell it directly with a personal prompt. Both stay inside your private workspace.
Memory with categories
TENGWAR organises what it remembers — preferences, facts, relationships, skills — and weighs each item by how relevant it is to your work.
Smart deduplication
When something new is similar to what's already there, TENGWAR merges instead of cluttering. The most relevant memories surface first.
You stay in control
Every memory comes with a 24-hour undo window and a full audit trail. Review, edit, or take back anything at any time.
Your prompt is yours
The instructions you give TENGWAR about yourself are never logged in telemetry — redacted at the source.
Strict isolation
Your memory is yours alone. By design, no other user's session can see it, search it, or leak into your context.
Integrations
Plug TENGWAR into the systems you already use — without ever leaving your network.
ERP connectors
Read your operational system of record — customers, suppliers, invoices, orders, accounting, inventory.
ERP connectors
Read your operational system of record — customers, suppliers, invoices, orders, accounting, inventory.
Read customers, suppliers, invoices, sales / purchase orders, accounting balances and live inventory directly from Odoo (XML-RPC, read-only).
Same canonical surface against the open-source SME ERP, plus a no-effect "draft invoice" preview helper.
Mid-market on-prem ERP integration over OData v4 with Entra ID / Azure AD authentication.
REST Service Layer adapter with the same canonical surface.
The single cloud-touching ERP connector — SuiteTalk REST + SuiteQL, with an explicit threat-model exception.
Direct ODBC against the customer's MS SQL Server (no Sage API needed); read-only by hard guarantee at startup.
Spanish public data & legal intelligence
Search and monitor the regulator's gazette and registries that drive Spanish business decisions.
Spanish public data & legal intelligence
Search and monitor the regulator's gazette and registries that drive Spanish business decisions.
Search and monitor Spanish state regulations by date, section, department, free text. Pull full XML, set term watches, follow regulatory cross-references. SQLite + FTS5 cache absorbs upstream rate limits.
Province-scoped commercial-registry bulletin discovery for any Spanish company; daily summaries, province watches, structured PDF references for the ingester.
Property search by reference or address, retrieval of non-protected cadastral data, and map context — for construction, lender underwriting, due diligence.
Spanish compliance & e-invoicing
Validate, sign and submit invoices and tax artefacts to AEAT, FACe and the Basque haciendas.
Spanish compliance & e-invoicing
Validate, sign and submit invoices and tax artefacts to AEAT, FACe and the Basque haciendas.
Embedded Spanish compliance toolkit: NIF/CIF/IBAN/postal/phone validation, IVA & IRPF tables, fiscal calendar, Modelo lookup (303, 100, …), CSB43/Norma 43 bank-statement parsing, XAdES-EPES signing with your configured certificates. Pure compute, no external calls.
Compose, sign and submit invoice records to AEAT under the RD 1007/2023 anti-fraud regime; track lifecycle.
Build, validate, sign and submit Facturae XML to FACe / FACeB2B for public-sector invoicing.
Compose, chain and submit anti-fraud invoices to the three Basque haciendas forales.
Communication & productivity
Mail, calendar and electronic signature — protocol-based and provider-neutral.
Communication & productivity
Mail, calendar and electronic signature — protocol-based and provider-neutral.
One protocol-based connector for any standards-compliant provider: search, thread reading, deadline & commitment extraction, draft preparation, send-with-confirmation, event listing and event creation with confirmation. Replaces today's in-backend Email & Calendar tools.
Prepare a signature request, dispatch with confirmation, poll status, download signed evidence; provider-neutral with Signaturit, Viafirma and Docuten as adapters.
Vertical Spain
Domain-specific Spanish public sources — meteorology, pharmaceutical safety, grants, procurement, jurisprudence.
Vertical Spain
Domain-specific Spanish public sources — meteorology, pharmaceutical safety, grants, procurement, jurisprudence.
Spanish meteorological data — forecasts, historicals, weather alerts; valuable for construction, insurance, energy, agriculture verticals.
Spanish pharmaceutical safety database — drug lookups, contraindications, batches; healthcare-vertical workhorse.
Spanish national grants database — search public subsidies and funding programmes by region, sector, deadline.
Spanish public-procurement platform — discover bids, contracts and tender deadlines.
Spanish court decisions and case law — search jurisprudence, retrieve full sentencias.
Public administration
Notifications and court filings from the Spanish public administration — read-only by hard contract.
Public administration
Notifications and court filings from the Spanish public administration — read-only by hard contract.
Read-only metadata + deadline extraction from the Spanish public-admin notifications system. Requires customer X.509 certificate.
Read-only metadata from the Spanish court filing system. Requires customer X.509 certificate.
Admin & Operations
- Turnkey doesn't mean opaque. Here's what the operator runs.
- Every agent action, every tool call, every login — auditable.
- Tune the agents. Schedule the ingestion. Watch the GPUs.
From delivery to production in weeks, not months.
TENGWAR arrives as a complete system. No complex integration projects, no cloud setup.
Discovery
We assess your documents, security requirements, and team structure.
1 weekDeployment
Hardware delivered and configured on-site. Initial ingestion and permissions.
1-2 weeksValidation
Your team tests with real documents. We fine-tune search and permissions.
1 weekProduction
Full rollout across departments with user training included.
OngoingIncludes remote support, software updates, and model upgrades. On-site support available in Spain.
From Document to Knowledge
Every step traceable. Every chunk tied back to its source. Section-aware. Multilingual. Audit-ready.
Parsing
Layout, text, assets and fusion — the document is read and structured.
- 1Structure detectionDocling
- 2Text extractionDocling + OCR model
- 3Asset extractionOCR model + MinIO
- 4Content fusion—
Structure & Translation
Section hierarchy, cross-references and language alignment.
- 5Context enrichmentLLM
- 6Structure extractionDocling
- 7TranslationTranslation model
- 8Asset description translationTranslation model
Chunking & Vectorization
Globally optimal boundaries, hybrid index, security-stamped storage.
- 9ChunkingSemantic chunking
- 10Contextual retrievalLLM
- 11Asset linking—
- 12Structure linking—
- 13TokenizationTokenizer model
- 14EmbeddingEmbedding model
- 15StoringQdrant
- 16Summary generationLLM + Qdrant
Restart-safe
Heartbeat from the ingester to the backend lets the document workflow state machine detect stuck documents and recover them.
Security baked into the index
Clearance level and department IDs stamped onto every Qdrant point at storage time. The query layer participates in the search, not on top of it.
Structure detection
Structural analysis with Docling: section hierarchy, tables, and cross-references are preserved across all 18 supported formats — native or scanned.
Semantic OCR
For scanned documents we use a state-of-the-art vision-language model instead of classic OCR, capturing meaning and context — not just characters.
Figures and tables
Every figure and table is extracted as an image and described with type-specific prompts, so visual content becomes retrievable through its description.
Structural translation
Per-element translation preserves the original document's structure; content stays aligned across languages without collapsing the hierarchy.
Semantic chunking
Instead of splitting by paragraphs, fixed length, or sliding windows, we cluster content by semantic similarity so each fragment represents a coherent idea; section-aware, so fragments never cross headings.
Contextual retrieval
Each fragment can be enriched with LLM-generated context before indexing, improving relevance when retrieved standalone.
Hybrid index
Each fragment is indexed in parallel with a state-of-the-art embedding model for meaning and with BM25 for exact lexical matching. Searches combine both signals.
Hierarchical retrieval
Searches can filter at the document level first, then section, then fragment — using optional summaries at each level to speed up and focus queries.
Inside the Engine
How a question becomes an answer — every step traceable, every claim cited.
Retrieval and answering flow
- User queryWhat the user types in the chat.
- Smart routingThe system reads the question and picks the right path — most questions never need an LLM.
- Question preparationThe system clarifies pronouns, translates if needed, identifies the intent and rewrites the query into useful variants.
- Multi-channel searchThree searches run in parallel against your documents — every angle covered at the same time.
- Smart re-rankingA re-ranker model re-orders the candidates by how well each one really answers the original question.
- Self-correctionThree independent checks catch low-confidence or empty answers and re-run the search with a corrected query before responding.
- Context preparationPicks the highest-value fragments, places them where the LLM reads them best, and attaches the citations.
- Specialist agentsResearch, web, document-generation and synthesis agents each take a slice of the work — in parallel when they can.
- Agent responseStreamed back token-by-token, with every citation linked to the exact source document and page.
Two-path routing
Simple, unambiguous questions take the fast deterministic path with zero LLM cost. Only the genuinely complex ones spin up the full reasoning pipeline.
Hybrid search
Combines semantic search ("what does it mean?") with classic keyword search ("these exact words"), and weights them based on what kind of question you asked.
Three-layer self-correction
If the system can't find a confident answer, it re-runs retrieval with a corrected query before responding. It catches its own bad answers before you do.
Specialist agents
Different experts handle different parts of a question — research, web access, drafting documents, final synthesis — and run in parallel when they can.
Multi-part questions
Ask several things at once and the system splits them, answers each independently, and tells the LLM how confident it is in each piece.
Live activity & citations
You see which specialist is working, what tool is being called, and which documents are being cited — in real time, while the answer is still forming.
Security inside the search
Permission checks happen inside the search query itself, not on top of it. Users only see fragments their clearance and department allow.
Full audit trail
Every tool call, every search, every answer — logged with user, department, status and timing. Filterable for compliance reviews and incident response.
Experience the Interface
A modern, intuitive interface designed for productivity.
Technology Stack
Built with enterprise-grade technologies for reliability and performance.
Agent Framework
Qwen3.6-35B-A3BQwen3-Embedding-8BQwen3-Reranker-8B
GLM-OCR 0.9B
DoclingFrequently Asked Questions
Does any data leave my network?
No. TENGWAR runs entirely on hardware installed at your premises. All document processing, AI inference, and search happens locally. There are zero external API calls, zero cloud dependencies, and zero data transfers to third parties.
What hardware is required?
TENGWAR is delivered pre-installed on an NVIDIA DGX Spark with 128GB of unified memory, a Grace ARM64 CPU, and a Blackwell GB10 GPU. This single device runs the entire platform — no server room required. It connects to your network like any other appliance.
How many documents can it handle?
TENGWAR is designed for enterprise-scale document libraries. The system scales with your needs — from hundreds to tens of thousands of documents across multiple departments and security levels.
Does it work with documents in multiple languages?
Yes. TENGWAR ingests documents in their original languages and translates them at indexing time into your deployment's configured working language. You can ask questions in any supported language — answers arrive in your team's working language with citations back to the original document. This makes cross-language search seamless without per-user language switching.
How accurate are the answers?
TENGWAR uses a hybrid search approach combining semantic understanding with traditional keyword matching, followed by a cross-encoder reranking stage. Every answer includes citations pointing to the exact source document, page, and passage — so your team can always verify. For low-confidence results, TENGWAR re-runs retrieval with corrected queries (Self-RAG / Corrective RAG) and combines multi-query expansions (RAG-Fusion) — significantly reducing hallucinated or incomplete answers.
Is it GDPR compliant?
Yes. Because all data stays on your premises, there are no cross-border data transfers and no third-party processing. You maintain complete control over the data lifecycle — from ingestion to deletion. TENGWAR is designed for GDPR and Spanish LOPD compliance from the ground up.
Can different departments have different access levels?
Yes. TENGWAR includes a four-level security clearance system (Public, Confidential, Secret, Top Secret) combined with department-based isolation. Users only see documents they are authorized to access based on both their clearance level and department assignment.
What file formats are supported?
18 formats including PDF, DOCX, XLSX, PPTX, images (with OCR), CSV, JSON, XML, HTML, Markdown, and plain text. Documents are processed through an 18-phase intelligent pipeline that handles structure detection, text extraction, and semantic indexing automatically.
How long does deployment take?
Typical deployment takes 2-4 weeks from initial discovery to full production rollout. TENGWAR arrives as a pre-configured system — there is no complex integration project or cloud setup required.
What ongoing support is included?
TENGWAR includes remote technical support, software updates, and AI model upgrades. On-site support is available for organizations in Spain.
Does TENGWAR plug into our ERP?
Yes, via dedicated MCP connectors. Available today: Odoo, FacturaScripts. Coming soon: Microsoft Dynamics 365 BC, SAP Business One, Oracle NetSuite, Sage 50cloud Spain. Connectors are isolated Docker sidecars in your network — adding an integration does not redeploy your stack.
Can different users get different answers from the same documents?
Yes. Per-user memory, per-user personal prompts, per-user clearance level and department membership all influence retrieval and synthesis. Personal-prompt content is never logged in telemetry.
Where do email and calendar live?
In a single Mail & Calendar MCP that talks IMAP / SMTP for mail and CalDAV for calendar — so it works with whichever provider IT runs (self-hosted, Gmail, Microsoft 365, Fastmail, ProtonMail Bridge, …). Sends are gated by explicit confirmation. Status: coming soon.
Do you support Spanish public-sector data?
Yes: BOE and BORME today; Catastro, AEMET, AEMPS CIMA, BDNS, PLACSP, CENDOJ, DEHú and LexNET coming soon — all as isolated MCP connectors with their own configuration.
Who watches the agent?
Default-deny tool registry; SSRF guard on web tools; per-document ACL re-check inside the Qdrant query at retrieval time; full audit log filterable by user, department, tool, date and status. Personal-prompt content and sensitive tool args are redacted in audit records.
Enterprise AI Without Compromise
See how TENGWAR works with your documents. We'll set up a private demo with your real use case.