Agentic Data-Driven Design and Analysis — the agentic layer over f3dasm (Framework for Data-Driven Design & Analysis of Structures & Materials), itself the framework for the 3dasm course.
You write one file describing an engineering design or data problem — the objective, the design space, what counts as valid. adda runs a team of LLM agents that decide what to try, build the code to evaluate it, run real experiments, review their own conclusions before accepting them, and hand you back a notebook that reproduces the result end to end.
It builds on f3dasm for the
data-driven primitives (ExperimentData, Domain, DataGenerator, the
Pipeline); adda is the agentic layer on top and carries no copy of f3dasm
core.
pip install "adda @ git+https://github.com/bessagroup/adda.git"You'll also need a model to drive the agents — by default, the Claude CLI:
npm install -g @anthropic-ai/claude-code
claude # first run prompts you to log inThe only required input is a PROBLEM_STATEMENT.md in the study directory.
from adda import AgenticRun
report = AgenticRun(
study_dir="studies/my_study",
model="claude-haiku-4-5-20251001",
).execute()
print(report)See the Quickstart for a worked example, start to finish.
https://bessagroup.github.io/adda/ — or run mkdocs serve locally.
BSD-3-Clause.