Quantum

Declarative web apps in XML, with AI and RAG built into the language. No build chain, no JavaScript, no frontend framework.

Quantum is a full-stack framework whose language is markup. State, database queries, forms, LLM calls and retrieval-augmented generation are all tags — not libraries you wire together. It takes its philosophy from ColdFusion and Adobe Flex: the markup is the app.

1.0, and honest about it. See Stability — the table there reflects what has actually been executed end-to-end, not what is aspirational.

The part that isn't like the others

Retrieval-augmented generation, as a language construct — this page answers a question from the Markdown files in knowledge/ and lists the ones it used:

<q:component name="Ask">
  <!-- The documents in knowledge/, split into chunks and embedded when the
       page runs. persist="false" keeps the index in memory; without it, it is
       stored in ./.quantum/knowledge and reused until a document changes. -->
  <q:knowledge name="docs" persist="false" chunkSize="300" chunkOverlap="30">
    <q:source type="directory" path="knowledge" pattern="*.md" />
  </q:knowledge>

  <q:set name="question" value="{query.q}" default="" />

  <q:if condition="question">
    <!-- The question retrieves the closest chunks; they reach the model
         numbered, with the instruction to answer only from them and cite
         them like [1]. -->
    <q:llm name="answer" knowledge="docs" top="2">
      <q:message role="user">{question}</q:message>
    </q:llm>
  </q:if>

  <ui:window title="Ask the store">
    <ui:form>
      <ui:input bind="q" value="{question}" placeholder="Your question" />
      <ui:button variant="primary">Ask</ui:button>
    </ui:form>
    <q:if condition="question">
      <ui:text>{answer}</ui:text>
      <q:if condition="answer_result.grounded">
        <ui:text>Sources:</ui:text>
        <q:loop items="{answer_result.sources}" var="s">
          <ui:text>[{s.n}] {s.name}</ui:text>
        </q:loop>
        <q:else>
          <ui:alert variant="warning">This answer cites none of the documents.</ui:alert>
        </q:else>
      </q:if>
    </q:if>
  </ui:window>
</q:component>

That splits the documents into chunks, embeds them, retrieves the closest ones and asks the model to answer only from them, citing each one — in the markup. answer_result.grounded says whether the answer cites any of them. There is no Python file behind it. (Recipe: Answers with their sources.)

An agent whose tools you write in Quantum, same idea:

<q:component name="Assistant">
  <!-- The model never writes SQL: it picks a tool and its arguments. The
       tool is a read-only query you wrote; its q:param says the argument's
       type, and the model's value is converted to it before the query runs. -->
  <q:agent name="stock" maxIterations="4" timeout="60000" onerror="continue">
    <q:instruction>You help a shop owner. Use the tools to look at the data,
      then answer in one sentence.</q:instruction>

    <q:tool name="low_stock" description="Products with fewer units in stock than `below`">
      <q:param name="below" type="integer" default="5" />
      <q:function name="lowStock">
        <q:query name="rows" datasource="db">
          SELECT name, stock FROM products WHERE stock &lt; :below ORDER BY stock
          <q:param name="below" value="{below}" type="integer" />
        </q:query>
        <q:return value="{rows}" />
      </q:function>
    </q:tool>

    <q:execute task="Which products are running out of stock?" />
  </q:agent>

  <ui:window title="Stock assistant">
    <q:if condition="stock_result.success">
      <ui:text>{stock}</ui:text>
      <q:else>
        <ui:alert variant="warning">The assistant did not finish: {stock_result.error.message}</ui:alert>
      </q:else>
    </q:if>
    <!-- Every tool call the agent made, written out. -->
    <q:loop items="{stock_result.actions}" var="a">
      <ui:text>Called {a.call}</ui:text>
    </q:loop>
  </ui:window>
</q:component>

The model never writes SQL: it picks a tool and its arguments, and the argument is converted to the q:param's type before the query runs. The reasoning loop, the tool calls and the failure contract (onerror, stock_result) are the runtime's job. (Recipe: An agent over your database.)

The rest of the language

<q:component name="Products">
  <!-- ?name=mouse from the URL; empty when it is not there. -->
  <q:set name="term" value="{query.name}" default="" />

  <!-- :pattern is bound to the q:param: the value is sent to the database
       apart from the SQL, so it can never change what the SQL does. -->
  <q:query name="products" datasource="db">
    SELECT name, price FROM products WHERE name LIKE :pattern ORDER BY price
    <q:param name="pattern" value="%{term}%" type="string" />
  </q:query>

  <ui:window title="Products">
    <ui:text>{products_result.recordCount} products</ui:text>
    <ui:table source="{products}">
      <ui:column key="name" label="Name" />
      <ui:column key="price" label="Price" />
    </ui:table>
  </ui:window>
</q:component>

Saved as components/index.q, with the database declared in quantum.config.yaml, quantum start serves it at http://localhost:8080/, and /?name=mouse filters it. (Recipe: A query with parameters.)

q:query refuses to run SQL with an undeclared :param — parameterised queries are enforced by the parser, not by discipline.

Also core: q:set with session. / application. / request. scopes, q:if, q:function, q:action for form handling, q:data for CSV/JSON/XML import, q:import / q:slot for composition.

The examples above are files of Cookbook recipes, byte for byte, and the recipes run in CI; the quick start below is run as shown (tests/docs/test_readme.py checks both).

Quick start

Requirements: Python 3.12+ and pip.

pip install quantum-framework

Create hello.q:

<q:component name="HelloWorld" xmlns:q="https://quantum.lang/ns">
  <q:return value="Hello World!" />
</q:component>

quantum run hello.q

[EXEC] Executing component: HelloWorld
[SUCCESS] Result: Hello World!

For a web app, put .q files in components/ and run quantum start (components/index.q is served at /). quantum stop stops it.

For the AI examples you also need a model server — Ollama at http://localhost:11434 unless QUANTUM_LLM_BASE_URL says otherwise — and the RAG extra. There is no built-in model name: say which one in quantum.config.yaml (llm: model: phi3) or QUANTUM_LLM_DEFAULT_MODEL.

pip install "quantum-framework[rag]"
ollama pull phi3 && ollama pull nomic-embed-text

Declare datasources in quantum.config.yaml (next to components/) and q:query works with nothing else running — SQLite needs no extra; PostgreSQL and MySQL drivers come with pip install "quantum-framework[db]":

datasources:
  db:
    driver: sqlite
    database: ./data/app.db

CLI

From source

To work on Quantum itself:

git clone https://github.com/danielgregorio/quantum.git
cd quantum
pip install -e ".[dev]"
quantum run examples/hello.q

See CONTRIBUTING.md for the test suite and the architecture.

Documentation

Full docs at quantumframework.net:

• Start: Installation · Quick Start · Tutorial: build the tasks app
• Learn: Guide · Cookbook — short tested recipes · Showcase — the complete apps
• Look up: Reference — every tag, function, command and SPEC rule, generated from the code
• Plan: Stability — what 1.0 promises · Roadmap · Changelog

The site is also in Português, Español and 中文.

Stability

Support levels are defined in SUPPORT_TIERS.md. Short version:

A functional audit in 2026-09 found that several of these surfaces had never been run end-to-end despite being documented as complete. They were fixed or re-labelled, and feature status is now verified by execution rather than asserted by hand.

From 1.0, Core and AI follow semantic versioning: a 1.x release does not break a program that uses only them — their meaning is fixed by the rules in SPEC.md, and a break waits for 2.0. Experimental and Laboratory surfaces carry no such promise and may change in any release.

Project layout

quantum/
├── quantum/
│   ├── core/        # Parser & AST (registry-based, modular)
│   ├── runtime/     # Execution engine, web server, renderer
│   └── cli/         # Command-line entry point
├── examples/        # runnable .q examples
├── tests/           # pytest suite; conformance/ cites SPEC.md
├── scripts/         # dev tools
└── docs/            # VitePress documentation

Adding a tag is one parser + one executor + a registry entry — see CONTRIBUTING.md.

Contributing

Read CONTRIBUTING.md for dev setup and how the modular parser/executor architecture works. By participating you agree to the Code of Conduct.

Found a security issue? Follow SECURITY.md — do not open a public issue.

License

MIT — see LICENSE.