Runnable guides for proto_tools. Each guide is a notebook that runs top to
bottom; open one and work through it.
| Guide | What it covers | Hardware |
|---|---|---|
tool_environments.ipynb |
How a tool's isolated environment is built on first call and cached afterward, shown live with a small ESM2 model | CPU |
tool_persistence.ipynb |
Keeping a model warm across calls (persist() / persist_tool() / get()) |
1 GPU |
device_management.ipynb |
GPU selection, LRU eviction, CPU offload, and packing models per GPU | 1+ GPU |
parallel_execution.ipynb |
Fanning work out across GPUs with ToolPool |
2+ GPU |