— Understand · Formalize · Reason · Execute —

VAILS & VVL

All of AI in a single server.
A language built for the applications of tomorrow.
My VVL notebook Read the book Discover VVL
— Origin —
VDOM
the founding technology · 1998
VDOM (Visual Dynamic Object Model) is the first technology aimed at abstracting web applications.
begets
— Platform —
VAILS
The first AI server
Serveur VAILS — un seul binaire
A neuro-symbolic server: a REPL, an API, an LLM orchestrator, a Prolog engine and a CP-SAT solver — all under one roof.
speaks
— Language —
VVL
The Rosetta Stone
$dx = [@ "flu"] $dx.symptoms $dx.duration # neuro → symbolic ($case ~~ $dx) $bio ? [§ "age?"] [print $dx] VVL
A homoiconic Lisp syntax. 879 builtins. 20 semantic operators. A language where neural and symbolic cohabit as first-class citizens.
879
documented builtins
34
chapters + appendix
8
parts of the book
10
training cycles
~95%
full N-S fidelity
— Huit solutions, prêtes à l'emploi —

Des applications concrètes, en un clic

Huit démonstrations interactives, une par produit VAILS Agents. Cliquez une carte pour l'ouvrir dans le catalogue.

— One server, the whole neuro-symbolic stack —

To an intent, the right engine

VVL is not just a planner: the same grammar routes a natural-language intent to the right paradigm — semantics, logic, optimization, ontology, learning — and returns an executed, inspectable result.

Semantics

neuro
[§ "chat"] == [§ "félin"]
→ 1 · synonyms recognised, no prompt

Proven compute

§§
[§§ "12 % de 240 ?"]
→ 36 · plain language compiled, then executed

Logic

prolog
[prologValues "fam" "ancestor(jean, X)" X]
→ [pierre marie luc] · deductive recursion

Optimization

cp-sat
[optimize "au plus 2 objets, valeur max" "kp"]
→ OPTIMAL, objective 9 · proven by OR-Tools

Extraction

neuro-sym
[ontology:fromText $O "Person" "Alice, 30 ans, Paris"]
→ <Person name=Alice age=30 city=Paris> typed

Expert system

fuzzy
[Triage "client premium, fort impact"]
→ URGENT · classifier born from a description

Knowledge

rag
[ragAsk $kb "Qui pilote le symbolique ?"]
→ sourced answer from your documents

Learning

ilp
[ilp:learn $bias "grandparent" E+ E−]
→ grandparent(X,Z) :- parent(X,Y), parent(Y,Z) induced
❦ ❦ ❦
— Four pillars —

One language, four layers

Neural

νοῦς · intuition

Understand natural language with LLMs, extract intent, reify concepts and tap into an indexed RAG memory.

λ

Symbolic

λόγος · formalization

Lambdas, recursion, unified OO with mixins/MRO, Prolog as a first-class citizen, priority ontologies, possible worlds.

Algebraic

τέχνη · rigour

Algebra of the 8 Zalta operators, fuzzy logic with T-norms, global constraints compiled down to CP-SAT and OR-Tools.

Verifiable

ἀλήθεια · proof

Audited, re-runnable results, improvable without calling the LLM again. Traced assumptions, why-infeasible diagnostics, minimal delta computed.

— Book contents —

Nine parts, forty chapters

A journey from theoretical foundations to production pipelines and the language in depth: medical diagnosis, construction estimation, transactional fraud, agents, optimization, meta-programming.

I.

History and foundations

3 chapters · Frege, Zalta, the neuro-symbolic oscillation
 II.

The VVL language

3 chapters · syntax, functional, unified OO
 III.

LLM co-processor

4 chapters · 20 semantic operators, @, §§
 IV.

Symbolic bridge

3 chapters · Prolog, ontologies, Zalta operators
 V.

Neural bridge

4 chapters · RAG, extraction, Gherkin, induction
 VI.

Pipelines in practice

4 chapters · medical, construction, fraud
 VII.

Autonomy & advanced systems

6 chapters · agents, fuzzy expert, atoms, constraints, genetics, ILP
 VIII.

The language in depth

7 chapters · meta-programming, dialects, modules, typed scalars, tests, signed plugins
Browse all 34 chapters in detail →
❦
VVL is not just a language: it is a bridge between human understanding and the reliable execution of machines.— book preamble —
Open a public REPL See the server