adding helper to bootstrap the generation of the meta.yaml file

docs/api/meta_yaml_generator.md

@@ -0,0 +1,42 @@
+# Meta YAML Generator
+
+## Overview
+
+The Meta YAML Generator is a tool designed to automatically create a `meta.yaml` file for chemical datasets using Large Language Models (LLMs). It analyzes the structure of a given DataFrame and generates a comprehensive metadata file, including advanced sampling methods and template formats.
+
+The model used by default is `gpt4o`. For using it, you need to expose the `OPENAI_API_KEY` environment variable.
+
+## `generate_meta_yaml`
+
+::: chemnlp.data.meta_yaml_generator.generate_meta_yaml
+handler: python
+options:
+show_root_heading: true
+show_source: false
+
+## Usage Example
+
+```python
+import pandas as pd
+from chemnlp.data.meta_yaml_generator import generate_meta_yaml
+
+# Load your dataset
+df = pd.read_csv("your_dataset.csv")
+
+# Generate meta.yaml
+meta_yaml = generate_meta_yaml(
+    df,
+    dataset_name="Polymer Properties Dataset",
+    description="A dataset of polymer properties including glass transition temperatures and densities",
+    output_path="path/to/save/meta.yaml"
+)
+
+# The meta_yaml variable now contains the dictionary representation of the meta.yaml
+# If an output_path was provided, the meta.yaml file has been saved to that location
+```
+
+You can also use it as a command-line tool:
+
+```bash
+python -m chemnlp.data.meta_yaml_generator path/to/your_dataset.csv --dataset_name "Polymer Properties Dataset" --description "A dataset of polymer properties including glass transition temperatures and densities" --output_path "path/to/save/meta.yaml"
+```

src/chemnlp/data/helper.py

@@ -2,6 +2,7 @@
 import yaml
 from typing import Dict, Any
 from litellm import completion
+import fire
 
 CONSTANT_PROMPT = """
 
@@ -100,7 +101,10 @@
 4. Random Choices:
 - Use {#option1|option2|option3!} for random selection of text.
 
-Generate a similar meta.yaml structure for the given dataset, including appropriate targets, identifiers, and templates based on the column names and example data provided. Include at least one multiple choice template and one benchmarking template."""
+Generate a similar meta.yaml structure for the given dataset, including appropriate targets, identifiers, and templates based on the column names and example data provided. Include at least one multiple choice template and one benchmarking template.
+
+Just return raw YAML string, no need to wrap it into backticks or anything else.
+"""
 
 
 def generate_meta_yaml(
@@ -140,9 +144,11 @@ def generate_meta_yaml(
 
     # Call the LLM with the prompt
     llm_response = completion(
-        model=model, messages=[{"role": "user", "content": prompt}]
+        model=model, messages=[{"role": "user", "content": prompt}], temperature=0
     )
 
+    llm_response = llm_response.choices[0].message.content
+
     # Parse the LLM's response and convert it to a dictionary
     try:
         meta_yaml = yaml.safe_load(llm_response)
@@ -153,24 +159,41 @@ def generate_meta_yaml(
     return meta_yaml
 
 
-# Example usage
-if __name__ == "__main__":
-    # Load your DataFrame
-    df = pd.read_csv("your_dataset.csv")
-
-    # Generate meta.yaml
-    meta_yaml = generate_meta_yaml(
-        df,
-        dataset_name="Your Dataset Name",
-        description="A brief description of your dataset",
-    )
+def cli(
+    data_path: str,
+    dataset_name: str,
+    description: str,
+    model: str = "gpt-4o",
+    output_path: str = None,
+):
+    """
+    Generate a meta.yaml structure for a dataset using an LLM based on a CSV file.
 
+    Args:
+        data_path (str): The path to the CSV file containing the dataset.
+        dataset_name (str): The name of the dataset.
+        description (str): A brief description of the dataset.
+        model (str, optional): The LLM model to use. Defaults to 'gpt-4o'.
+        output_path (str, optional): The path to save the generated meta.yaml. Defaults to None.
+    """
+    # Load the dataset from the CSV file
+    df = pd.read_csv(data_path)
+
+    # Generate the meta.yaml structure
+    meta_yaml = generate_meta_yaml(df, dataset_name, description, model)
+
+    output_path = output_path or "meta.yaml"
     # Print or save the generated meta.yaml
     if meta_yaml:
         print(yaml.dump(meta_yaml, default_flow_style=False))
 
         # Optionally, save to a file
-        with open("meta.yaml", "w") as f:
+        with open(output_path, "w") as f:
             yaml.dump(meta_yaml, f, default_flow_style=False)
     else:
         print("Failed to generate meta.yaml")
+
+
+# Example usage
+if __name__ == "__main__":
+    fire.Fire(cli)

src/chemnlp/data/meta.yaml

@@ -0,0 +1,31 @@
+bibtex:
+  - "@article{martins2023,\nauthor = {Martins, John and Doe, Jane and Smith, Alice},\ntitle = {Study on Blood-Brain Barrier Penetration of Various Drugs},\njournal = {Journal of Pharmacology},\nvolume = {12},\nnumber = {3},\npages = {123-134},\nyear = {2023},\ndoi = {10.1234/jpharm.2023.56789}}"
+description: Describing the ability of different drugs to penetrate the blood-brain barrier.
+identifiers:
+  - description: Simplified Molecular Input Line Entry System
+    id: SMILES
+    type: SMILES
+  - description: Name of the compound
+    id: compound_name
+    names:
+      - noun: compound name
+    type: Other
+license: CC BY 4.0
+links:
+  - description: corresponding publication
+    url: https://example.com/publication
+  - description: data source
+    url: https://example.com/data_source
+name: blood_brain_barrier_martins_et_al
+num_points: 2030
+targets:
+  - description: Indicates whether the compound can penetrate the blood-brain barrier (1 for yes, 0 for no)
+    id: penetrate_BBB
+    names:
+      - noun: blood-brain barrier penetration
+    type: integer
+templates:
+  - The compound {compound_name__names__noun} with SMILES {SMILES#} can {#penetrate|not penetrate!} the blood-brain barrier.
+  - The compound {compound_name__names__noun} with SMILES {SMILES#} is in the {split#} set.
+  - "Question: Which of the following compounds can penetrate the blood-brain barrier?\nOptions: {%multiple_choice_enum%4%aA1}\n{compound_name%}\nAnswer: {%multiple_choice_result}"
+  - The compound with SMILES {SMILES#} can penetrate the blood-brain barrier:<EOI>{penetrate_BBB#}