research_tools

Internal Python package with tooling and functionalities designed to improve research productivity.

Installation

Clone this repository from GitHub, then install it as a locally editable package with uv. Full steps—including using it from another project—are in SETUP.md.

git clone https://github.com/your-org/research_tools.git
cd research_tools
uv sync --extra test

Quick start

from pydantic import BaseModel
from research_tools import LLMService

class LabelResponse(BaseModel):
    label: int

service = LLMService()
result = service.structured_completion(
    messages=[{"role": "user", "content": "Classify this text."}],
    response_model=LabelResponse,
    model="gpt-5.4-nano",
)
print(result.label)

Set provider API keys in your environment (for example OPENAI_API_KEY, ANTHROPIC_API_KEY, OPENROUTER_API_KEY). Optionally pass the same keys per call via env={...} on structured completion APIs; see SETUP.md for details. Bedrock models use AWS credentials instead of an API key.

Documentation

See SETUP.md for:

• Cloning or downloading the repository
• Installing as an editable local package (in this repo or from another project)
• Optional install from GitHub without a local clone
• Environment variables, usage examples, and model config overrides