Playground
The playground/ folder contains small applications that can be used as coding
targets while developing ProtoAgent. They are not the runtime engine.
Recipe Recommendation
Path:
playground/recipe-reco/
Shape:
| File | Role |
|---|---|
api.py | Flask API. |
app.py | App entrypoint. |
client.py | Example client. |
recommender.py | Recommendation loop. |
filters.py | Ingredient and tag filters. |
scoring.py | Intentionally simple scoring logic. |
recipes.py | Seed recipes. |
storage.py | In-memory recipe store. |
models.py | Data models. |
Good ProtoAgent tasks:
explain @playground/recipe-reco/recommender.py
refactor the recipe scoring logic and keep behavior understandable
add tests for ingredient matching edge cases
Taskflow
Path:
playground/taskflow/
Shape:
| File | Role |
|---|---|
server.py | Flask routes for task list, create, complete, search, stats. |
service.py | Task business logic. |
storage.py | In-memory task store. |
models.py | Task model. |
utils.py | Serialization helpers. |
client.py | Example client. |
app.py | App entrypoint. |
Good ProtoAgent tasks:
explain the taskflow route structure
add validation for empty task titles
propose tests for task completion and search
Why Keep Playground Apps Small
Small apps are useful because they let you validate:
- Context Loom ranking.
- Explorer read/search behavior.
- Coder diff previews.
- Approval flow.
- Session memory.
- Trace and timeline quality.
They should stay simple enough that a local model can reason about them without expensive context or many exploratory turns.