Adding definition validation functionality

docs/howtos/use-llms.rst

@@ -115,9 +115,51 @@ Suggesting Definitions
          finger toe \
          --style-hints "write definitions in formal genus-differentia form"
 
+Validating Definitions
+~~~~~~~~~~~~~~~~~~~~~~
+
+The LLM adapter currently interprets ``validate-definitions`` as comparing the specified definition
+against the abstracts of papers cited in the definition provenance, or by comparing the definition
+against the database objects that are cited as definition provenance.
+
+Here is an example of validating definitions for GO terms:
+
+.. code-block:: bash
+
+    runoak --stacktrace -i llm:sqlite:obo:go validate-definitions \
+      i^GO: -o out.jsonl -O jsonl
+
+The semsql version of GO has other ontologies merged in, so the ``i^GO:`` query only validates
+against actual GO terms.
+
+You can also pass in a configuration object.
+This should conform to the `Validation Data Model <https://w3id.org/oak/validation-datamodel>`_
+
+For example, this configuration yaml provides a specific prompt and also a URL for
+documentation aimed at ontology developers.
+
+.. code-block:: yaml
+
+    prompt_info: Please also use the following GO guidelines
+    documentation_objects:
+      - https://wiki.geneontology.org/Guidelines_for_GO_textual_definitions
+
+All specified URLs are downloaded and converted to text and included in the prompt.
+
+The configuration yaml is passed in as follows:
+
+.. code-block:: bash
+
+
+    runoak --stacktrace  -i llm:{claude-3-opus}:sqlite:obo:go validate-definitions \
+         -C src/oaklib/conf/go-definition-validation-llm-config.yaml i^GO: -O yaml
+
 Validating Mappings
 ~~~~~~~~~~~~~~~~~~~
 
+The LLM adapter validates mappings by looking up info on the mapped entity and
+comparing it with the main entity.
+
 .. code-block:: bash
 
     runoak --stacktrace -i llm:{gpt-4}:sqlite:obo:go validate-mappings \
@@ -165,8 +207,14 @@ as a developer, then you can do this:
 
 This will install the plugin in the same environment as OAK.
 
-TODO: instructions for non-developers.
+If you need to update this:
+
+.. code-block:: bash
 
+    cd ontology-access-kit
+    poetry run llm install -U llm-gemini
+
+TODO: instructions for non-developers.
 
 Mixtral via Ollama and LiteLLM
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

docs/packages/interfaces/validator.rst

@@ -3,13 +3,24 @@
 Validator Interface
 --------------------
 
-.. warning ::
+The Validator Interface provides access to a number of different validation operations over ontologies.
 
-   Currently the main validator methods are only implemented for :ref:`SqlDatabaseImplementation`
+The notion of validation in OAK is intentionally very flexible, and may encompass:
 
-The validate method is configured using a *metadata schema*. The default one used is:
+ * *Schema* validation, for example, checking definitions are strings and have 0..1 cardinality.
+ * *Logical* validation, using a reasoner.
+ * *Lexical* validation, for example, ensuring there are no spelling errors
+ * *Stylistic* validation, against a style guide
+ * *Content* validation, checking the content of the ontology against domain knowledge or other ontologies.
 
-- `Ontology Metadata <https://incatools.github.io/ontology-access-kit/datamodels/ontology-metadata/index.html>`_
+Different adapters may implement different portions of this.
+
+Schema Validation
+~~~~~~~~~~~~~~~~~
+
+The core validate method is configured using a *metadata schema*. The default one used is:
+
+- `Ontology Metadata <https://w3id.org/oak/ontology-metadata>`_
 
 This is specified using LinkML which provides an expressive way to state constraints on metadata elements,
 such as :ref:`AnnotationProperty` assertions in ontologies. For example, this schema states that definition
@@ -19,6 +30,15 @@ Different projects may wish to configure this - it is possible to pass in a diff
 
 For more details see `this howto guide <https://incatools.github.io/ontology-access-kit/howtos/validate-an-obo-ontology>`_
 
+.. warning::
+
+    Currently only implemented for :ref`sql_implementation`
+
+LLM-based validation
+~~~~~~~~~~~~~~~~~~~~
+
+See :ref:`use_llms`
+
 
 .. currentmodule:: oaklib.interfaces.validator_interface
 

docs/packages/utilities.rst

@@ -17,5 +17,6 @@ being turned into :ref:`interfaces`.
    lexical.lexical_indexer
    subsets.slimmer_utils
    apikey_manager
-   taxon/taxon_constraint_utils
+   taxon.taxon_constraint_utils
    table_filler
+

notebooks/Commands/ValidateMappings.ipynb

@@ -5,7 +5,7 @@
    "id": "0a28b88d-4deb-4d0a-a110-f27adf077e23",
    "metadata": {},
    "source": [
-    "# OAK apply command\n",
+    "# OAK validate-mappings command\n",
     "\n",
     "This notebook is intended as a supplement to the [main OAK CLI docs](https://incatools.github.io/ontology-access-kit/cli.html).\n",
     "\n",

notebooks/Commands/input/validate-definition-conf.yaml

@@ -0,0 +1 @@
+lookup_references: true