Installation¶
PlugBO is not on PyPI. Install it from source (Python 3.10+).
Editable install¶
That pulls torch, gpytorch, botorch, and the LLM clients (anthropic,
openai). Copy .env.example to .env if you will call an LLM:
# Anthropic (Claude)
ANTHROPIC_API_KEY=
# OpenAI (or any OpenAI-compatible endpoint: vLLM, Together, Groq, LM Studio, ...)
OPENAI_API_KEY=
OPENAI_BASE_URL=
# Ollama (local models). Defaults to http://localhost:11434/v1 if unset.
OLLAMA_BASE_URL=
Dev install¶
Adds test dependencies:
LoRA HPO extra¶
Needed only for the BoLT benchmark (LoRA hyperparameter optimization emulator):
Console scripts¶
The editable install registers these entry points:
| Command | Package | Purpose |
|---|---|---|
lenz |
lenz.cli:main |
BoTorch backend CLI |
sara |
sara.cli:main |
Agent-in-the-loop CLI |
sara-blind-test |
benchmarks.run_blind_test:main |
Blind synthetic-benchmark runner |
plugbo-viz |
viz.server:main |
Run viewer |
plugbo-viz-merged |
viz.merged_server:main |
Merged run viewer |
Next: the quickstart.