Tools System¶

Tools are organized as Toolkit classes that inherit from Toolkit (Agno framework).

Location: src/cs_copilot/tools/

Directory Structure¶

tools/
├── databases/          Database integrations
│   ├── base.py        BaseDatabaseToolkit (abstract)
│   ├── chembl.py      ChemblToolkit (REST API + MySQL backends)
│   ├── chembl_fetcher.py  RestChemblFetcher / SqlChemblFetcher strategies
│   └── types.py       Query types and configurations
│
├── chemography/       Dimensionality reduction
│   ├── gtm.py         GTMToolkit (high-level interface)
│   └── gtm_operations.py  Core GTM implementations
│
├── chemistry/         Molecular operations
│   ├── similarity_toolkit.py      Similarity calculations
│   ├── autoencoder_toolkit.py     LSTM autoencoder operations
│   └── descriptors.py             Molecular descriptors
│
├── io/                I/O and formatting
│   ├── pointer_pandas_tools.py   DataFrame ops + S3 integration
│   └── formatting.py              SMILES → images, markdown
│
└── constants.py       Configuration constants

Each toolkit registers methods as tools via self.register(method). Agents call these tools via the Agno tool-calling mechanism.

ChEMBL Backends¶

The ChemblToolkit supports two pluggable data backends via a strategy pattern (chembl_fetcher.py):

Backend	Trigger	Dependency	Use case
REST API	Default (no config needed)	`chembl_webresource_client` (included)	Quick setup, always-on access
MySQL	Set `CHEMBL_MYSQL_HOST` env var	`pymysql` (`uv sync --extra mysql`)	Faster queries, offline use, full SQL

Backend is auto-detected: MySQL when CHEMBL_MYSQL_HOST is present, REST otherwise. The REST API is always reported as available regardless of active backend.

Download the MySQL dump from the ChEMBL downloads page or the EBI FTP.

Adding a New Tool¶

Create a toolkit in src/cs_copilot/tools/:

from agno import Toolkit

class MyNewToolkit(Toolkit):
    def __init__(self):
        super().__init__(name="my_new_toolkit")
        self.register(self.my_tool_function)

    def my_tool_function(self, param: str) -> str:
        """Tool description for LLM."""
        return f"Result: {param}"

Import and pass to the agent factory's tools parameter