An open vocabulary for representing knowledge — entities, relationships, facts and rules — that both people and AI systems can read, validate and exchange.
See the codeREADME
AODM
AODM (AI Optimized Data Markup) is a small, open vocabulary for representing knowledge — entities, relationships, facts, rules — in a form both humans and AI systems can read, validate, and exchange.
This is the AODM 1.2 specification and its reference implementations.
AODM 1.2 Specification (start here)
The current core: 4 knowledge primitives (entity, relationship,
fact, rule) and 3 annotations (source, confidence, value),
plus a data-quality layer — content hashing, polarity, derivation
tracking, temporal validity, and structured measurement.
Namespace: http://fucaspark.com/aodm/1.2
| File | Purpose |
|---|---|
AODM/AODM-1.2-SPECIFICATION.md | The authoritative specification |
AODM/aodm-core-1.2.xsd | XML Schema |
AODM/aodm-core-1.2.schema.json | JSON Schema |
AODM/VALIDATION-RULES.md | Semantic rules processors must enforce beyond schema validation |
AODM/example-core-1.2.xml / .json | Worked examples using every feature |
conformance-check.py | Verifies the schemas accept valid documents and reject invalid ones |
Running the conformance check
python3 -m venv .venv && source .venv/bin/activate && pip install lxml jsonschema && python3 conformance-check.py
It also documents which constraints the schemas enforce and which require
processor logic — notably referential integrity, since libxml2 does not
enforce xs:IDREF for XML Schema.
Content digests are processor logic of the same kind: every reference
implementation recomputes an item's sha256 hash and rejects a document
whose digest disagrees with its text, in both the XML and JSON
serialisations.
Reference implementations
| Directory | Contents |
|---|---|
parsers/ | Parsers for Python, JavaScript, Java and C#, each parsing AODM 1.2 XML into the published JSON model and applying the validation rules no schema language can express. parsers/conformance.py runs every parser against the same suite. |
graph/ | Graph compiler and forward-chaining inference engine |
converters/ | XBRL to AODM conversion |
llm/context.py | Builds LLM context windows from an AODM document |
reference/ | Browser scripts — a validator and the HTML profile extractor |
Each parser is standard library only, with no dependency to add.
Embedding AODM in web pages
Use the HTML Embedding Profile:
AODM/AODM-1.2-HTML-PROFILE.md. It marks
up content with data-aodm attributes on the elements that display it, so
the human-readable and machine-readable versions cannot drift apart.
<p data-aodm="fact" data-aodm-id="license" data-aodm-about="aodm"
data-aodm-confidence="1.0">
AODM is released under the Apache License 2.0.
</p>
Extract and validate embedded markup with
reference/aodm-html-extract.js, or
online at https://fucaspark.com/validator.php.
Status
The 1.2 core ontology and its formal definition are complete and tested, with reference parsers in 4 languages, a graph compiler, an inference engine, an LLM context builder and an XBRL converter, all covered by the conformance suite.
License
Contributors
1 commits
See what people are saying
Hacker News (1)
Similar repos
Languages
Python
39.7%
JavaScript
24.9%
Java
18.1%
C#
17.4%
An open vocabulary for representing knowledge — entities, relationships, facts and rules — that both people and AI systems can read, validate and exchange.
See the codeREADME
AODM
AODM (AI Optimized Data Markup) is a small, open vocabulary for representing knowledge — entities, relationships, facts, rules — in a form both humans and AI systems can read, validate, and exchange.
This is the AODM 1.2 specification and its reference implementations.
AODM 1.2 Specification (start here)
The current core: 4 knowledge primitives (entity, relationship,
fact, rule) and 3 annotations (source, confidence, value),
plus a data-quality layer — content hashing, polarity, derivation
tracking, temporal validity, and structured measurement.
Namespace: http://fucaspark.com/aodm/1.2
| File | Purpose |
|---|---|
AODM/AODM-1.2-SPECIFICATION.md | The authoritative specification |
AODM/aodm-core-1.2.xsd | XML Schema |
AODM/aodm-core-1.2.schema.json | JSON Schema |
AODM/VALIDATION-RULES.md | Semantic rules processors must enforce beyond schema validation |
AODM/example-core-1.2.xml / .json | Worked examples using every feature |
conformance-check.py | Verifies the schemas accept valid documents and reject invalid ones |
Running the conformance check
python3 -m venv .venv && source .venv/bin/activate && pip install lxml jsonschema && python3 conformance-check.py
It also documents which constraints the schemas enforce and which require
processor logic — notably referential integrity, since libxml2 does not
enforce xs:IDREF for XML Schema.
Content digests are processor logic of the same kind: every reference
implementation recomputes an item's sha256 hash and rejects a document
whose digest disagrees with its text, in both the XML and JSON
serialisations.
Reference implementations
| Directory | Contents |
|---|---|
parsers/ | Parsers for Python, JavaScript, Java and C#, each parsing AODM 1.2 XML into the published JSON model and applying the validation rules no schema language can express. parsers/conformance.py runs every parser against the same suite. |
graph/ | Graph compiler and forward-chaining inference engine |
converters/ | XBRL to AODM conversion |
llm/context.py | Builds LLM context windows from an AODM document |
reference/ | Browser scripts — a validator and the HTML profile extractor |
Each parser is standard library only, with no dependency to add.
Embedding AODM in web pages
Use the HTML Embedding Profile:
AODM/AODM-1.2-HTML-PROFILE.md. It marks
up content with data-aodm attributes on the elements that display it, so
the human-readable and machine-readable versions cannot drift apart.
<p data-aodm="fact" data-aodm-id="license" data-aodm-about="aodm"
data-aodm-confidence="1.0">
AODM is released under the Apache License 2.0.
</p>
Extract and validate embedded markup with
reference/aodm-html-extract.js, or
online at https://fucaspark.com/validator.php.
Status
The 1.2 core ontology and its formal definition are complete and tested, with reference parsers in 4 languages, a graph compiler, an inference engine, an LLM context builder and an XBRL converter, all covered by the conformance suite.
License
See what people are saying
Hacker News (1)
Similar repos
Contributors
1 commits
Languages
Python
39.7%
JavaScript
24.9%
Java
18.1%
C#
17.4%