Learn AutoResearch

English

简体中文

English

简体中文

Appearance

L10 Code — 12-Dimension Scenario Exploration ​

Goal: Score and rank the 12 scenario dimensions for your domain — so you know which dimension to explore first before you start writing scenarios.

Run it:

sh
cd docs/en/lectures/lecture-10-twelve-dimension-scenarios/code
python scenario_priority.py api
python scenario_priority.py ml
python scenario_priority.py security

Step 1: Define the 12 dimensions ​

python
import sys

DIMENSIONS = [
    {"id": "D01", "name": "Happy Path",
     "desc": "Normal usage with valid inputs and expected conditions.",
     "tags": ["all"]},
    {"id": "D02", "name": "Edge Cases",
     "desc": "Boundary values, empty inputs, maximum sizes.",
     "tags": ["all"]},
    {"id": "D03", "name": "Adversarial Input",
     "desc": "Malformed, malicious, or unexpected data from untrusted sources.",
     "tags": ["api", "web", "security"]},
    {"id": "D04", "name": "Concurrency / Race Conditions",
     "desc": "Multiple simultaneous users or threads accessing shared state.",
     "tags": ["api", "database", "distributed"]},
    # D05-D12 follow same structure...
]

Key line: "tags": ["all"] — dimensions tagged "all" apply to every domain. Dimensions tagged with specific domains (like "api" or "ml") get a bonus score only for those domains.

Step 2: Define domain-specific boosts ​

python
DOMAIN_BOOSTS: dict[str, dict[str, int]] = {
    "api":        {"D03": 3, "D04": 2, "D05": 2, "D06": 2, "D10": 1},
    "ml":         {"D07": 4, "D08": 3, "D05": 2, "D10": 2},
    "security":   {"D03": 4, "D11": 3, "D04": 2},
    "web":        {"D12": 3, "D03": 2, "D05": 2},
    "database":   {"D04": 3, "D05": 3, "D06": 3},
    "distributed":{"D04": 4, "D06": 4, "D08": 2},
}

BASE_SCORE    = 5
TAG_MATCH_BONUS = 2

Key line: "ml": {"D07": 4, ...} — ML gets a +4 boost for D07 (Data Distribution Shift) because production data differing from training data is the most common and most dangerous ML failure. Domain expertise is encoded as numbers.

Step 3: Score and sort dimensions ​

python
def score_dimensions(domain: str) -> list[tuple[int, dict]]:
    domain  = domain.lower().strip()
    boosts  = DOMAIN_BOOSTS.get(domain, {})
    scored  = []
    for dim in DIMENSIONS:
        score = BASE_SCORE
        if domain in dim["tags"] or "all" in dim["tags"]:
            score += TAG_MATCH_BONUS
        score += boosts.get(dim["id"], 0)
        scored.append((score, dim))
    scored.sort(key=lambda x: -x[0])
    return scored

Key line: boosts.get(dim["id"], 0) — dimensions not in the boost table get 0 added. Only explicitly prioritized dimensions get the bonus. Unknown dimensions still get the base score.

Step 4: Print ranked output with recommendations ​

python
def main():
    domain  = sys.argv[1] if len(sys.argv) > 1 else "api"
    results = score_dimensions(domain)

    print(f"\nScenario Priority for domain: '{domain}'")
    for rank, (score, dim) in enumerate(results, 1):
        print(f"{rank:<10} {score:<7} {dim['id']:<5} {dim['name']:<35} {dim['desc']}")

    top3 = [d["name"] for _, d in results[:3]]
    print(f"\nRecommended first three: {', '.join(top3)}")
    print("Run these scenarios before moving to lower-priority dimensions.")

Key line: The recommendation is to start with the top 3 and work down — not to generate all 12 at once. One scenario per dimension per iteration keeps the loop systematic.

Try Changing It ​

  1. Compare api and security outputs — which dimensions appear in both Top 3s? Which appear in only one?
  2. Change the "ml" boost for D07 from 4 to 1 — does Data Distribution Shift drop out of the Top 3?
  3. Add a new domain "mobile" to DOMAIN_BOOSTS — which 3-4 dimensions matter most for mobile apps? Write the dict.
  4. Pick a feature you built recently, choose its domain, and write one concrete scenario for each of the Top 3 dimensions.