Orchid Ranker¶
Orchid is an adaptive-practice engine for learning products.
It ranks an eligible set of exercises, observes the learner's result, and
adapts what it recommends next. The public interface is one class:
AdaptiveRanker.
Install¶
Complete example¶
import pandas as pd
from orchid_ranker import AdaptiveRanker
history = pd.DataFrame({
"user_id": ["a", "a", "a", "b", "b", "b"],
"item_id": [101, 102, 201, 101, 102, 201],
"outcome": [1, 1, 0, 1, 0, 0],
"timestamp": [1, 2, 3, 1, 2, 3],
})
ranker = AdaptiveRanker().fit(history)
ranked = ranker.recommend("a", [101, 102, 201], top_k=2)
ranker.observe(
user_id="a",
item_id=ranked[0].item_id,
outcome=1,
timestamp=4,
)
Orchid chooses the appropriate starting learner: a transparent empirical baseline for sparse pilots and knowledge tracing only after data-support checks pass. Users do not select from a model catalog.
Where it fits¶
Orchid works best when:
- learners complete multiple scored practice attempts;
- an existing curriculum supplies multiple pedagogically valid next exercises;
- the next exercise should react to prior attempts; and
- success can be measured later through retained mastery, a post-test, or a certification outcome.
It is not a generic content feed, product recommender, LMS, or curriculum authoring system. Your learning product owns the content, eligibility rules, and learner experience; Orchid supplies adaptive practice sequencing and the decision evidence for a controlled evaluation.
Choose your path¶
| If you want to… | Start here |
|---|---|
| Try the learner loop on historical attempts | Quickstart |
| Check whether your data and catalog are ready | Adaptive-practice data readiness |
| Add durable decisions and delayed outcomes to an existing product | Production serving |
| Run an evidence-oriented controlled pilot | Pilot workflow |
| Look up a class or method | API reference |
| Check compatibility guarantees | API support policy |
Start with the quickstart, then read how Orchid works, the small API reference, adaptive-practice data readiness, or how to run a learning-efficacy pilot. The pilot integration contract defines the handoff between Orchid and a learning platform. The end-to-end reference-pilot workflow turns that contract into a runnable sequence. The product roadmap describes the path from a single-course pilot to an evidence-backed integration. The design-partner council keeps that roadmap grounded in simulated customer review without claiming any real-company affiliation.