examples : add pydantic models to GBNF grammar generator (#4883)

* Create pydantic-models-to-grammar.py

* Added some comments for usage

* Refactored Grammar Generator

Added example and usage instruction.

* Update pydantic_models_to_grammar.py

* Update pydantic-models-to-grammar-examples.py

* Renamed module and imported it.

* Update pydantic-models-to-grammar.py

* Renamed file and fixed grammar generator issue.

1 files changed, 136 insertions, 0 deletions

diff --git a/examples/pydantic-models-to-grammar-examples.py b/examples/pydantic-models-to-grammar-examples.py
new file mode 100644
index 00000000..a8a4919c
--- /dev/null
+++ b/examples/pydantic-models-to-grammar-examples.py
@@ -0,0 +1,136 @@
+# Function calling example using pydantic models.
+
+import json
+from enum import Enum
+from typing import Union, Optional
+
+import requests
+from pydantic import BaseModel, Field
+
+import importlib
+from pydantic_models_to_grammar import generate_gbnf_grammar_and_documentation
+
+# Function to get completion on the llama.cpp server with grammar.
+def create_completion(prompt, grammar):
+    headers = {"Content-Type": "application/json"}
+    data = {"prompt": prompt, "grammar": grammar}
+
+    response = requests.post("http://127.0.0.1:8080/completion", headers=headers, json=data)
+    data = response.json()
+
+    print(data["content"])
+    return data["content"]
+
+
+# A function for the agent to send a message to the user.
+class SendMessageToUser(BaseModel):
+    """
+    Send a message to the User.
+    """
+    chain_of_thought: str = Field(..., description="Your chain of thought while sending the message.")
+    message: str = Field(..., description="Message you want to send to the user.")
+
+    def run(self):
+        print(self.message)
+
+
+# Enum for the calculator function.
+class MathOperation(Enum):
+    ADD = "add"
+    SUBTRACT = "subtract"
+    MULTIPLY = "multiply"
+    DIVIDE = "divide"
+
+
+# Very simple calculator tool for the agent.
+class Calculator(BaseModel):
+    """
+    Perform a math operation on two numbers.
+    """
+    number_one: Union[int, float] = Field(..., description="First number.")
+    operation: MathOperation = Field(..., description="Math operation to perform.")
+    number_two: Union[int, float] = Field(..., description="Second number.")
+
+    def run(self):
+        if self.operation == MathOperation.ADD:
+            return self.number_one + self.number_two
+        elif self.operation == MathOperation.SUBTRACT:
+            return self.number_one - self.number_two
+        elif self.operation == MathOperation.MULTIPLY:
+            return self.number_one * self.number_two
+        elif self.operation == MathOperation.DIVIDE:
+            return self.number_one / self.number_two
+        else:
+            raise ValueError("Unknown operation.")
+
+
+# Here the grammar gets generated by passing the available function models to generate_gbnf_grammar_and_documentation function. This also generates a documentation usable by the LLM.
+# pydantic_model_list is the list of pydanitc models
+# outer_object_name is an optional name for an outer object around the actual model object. Like a "function" object with "function_parameters" which contains the actual model object. If None, no outer object will be generated
+# outer_object_content is the name of outer object content.
+# model_prefix is the optional prefix for models in the documentation. (Default="Output Model")
+# fields_prefix is the prefix for the model fields in the documentation. (Default="Output Fields")
+gbnf_grammar, documentation = generate_gbnf_grammar_and_documentation(
+    pydantic_model_list=[SendMessageToUser, Calculator], outer_object_name="function",
+    outer_object_content="function_parameters", model_prefix="Function", fields_prefix="Parameters")
+
+print(gbnf_grammar)
+print(documentation)
+
+system_message = "You are an advanced AI, tasked to assist the user by calling functions in JSON format. The following are the available functions and their parameters and types:\n\n" + documentation
+
+user_message = "What is 42 * 42?"
+prompt = f"<|im_start|>system\n{system_message}<|im_end|>\n<|im_start|>user\n{user_message}<|im_end|>\n<|im_start|>assistant"
+
+text = create_completion(prompt=prompt, grammar=gbnf_grammar)
+# This should output something like this:
+# {
+#     "function": "calculator",
+#     "function_parameters": {
+#         "number_one": 42,
+#         "operation": "multiply",
+#         "number_two": 42
+#     }
+# }
+function_dictionary = json.loads(text)
+if function_dictionary["function"] == "calculator":
+    function_parameters = {**function_dictionary["function_parameters"]}
+
+    print(Calculator(**function_parameters).run())
+    # This should output: 1764
+
+
+# A example structured output based on pydantic models. The LLM will create an entry for a Book database out of an unstructured text.
+class Category(Enum):
+    """
+    The category of the book.
+    """
+    Fiction = "Fiction"
+    NonFiction = "Non-Fiction"
+
+
+class Book(BaseModel):
+    """
+    Represents an entry about a book.
+    """
+    title: str = Field(..., description="Title of the book.")
+    author: str = Field(..., description="Author of the book.")
+    published_year: Optional[int] = Field(..., description="Publishing year of the book.")
+    keywords: list[str] = Field(..., description="A list of keywords.")
+    category: Category = Field(..., description="Category of the book.")
+    summary: str = Field(..., description="Summary of the book.")
+
+
+# We need no additional parameters other than our list of pydantic models.
+gbnf_grammar, documentation = generate_gbnf_grammar_and_documentation([Book])
+
+system_message = "You are an advanced AI, tasked to create a dataset entry in JSON for a Book. The following is the expected output model:\n\n" + documentation
+
+text = """The Feynman Lectures on Physics is a physics textbook based on some lectures by Richard Feynman, a Nobel laureate who has sometimes been called "The Great Explainer". The lectures were presented before undergraduate students at the California Institute of Technology (Caltech), during 1961–1963. The book's co-authors are Feynman, Robert B. Leighton, and Matthew Sands."""
+prompt = f"<|im_start|>system\n{system_message}<|im_end|>\n<|im_start|>user\n{text}<|im_end|>\n<|im_start|>assistant"
+
+text = create_completion(prompt=prompt, grammar=gbnf_grammar)
+
+json_data = json.loads(text)
+
+print(Book(**json_data))