Results

client.results — fetch evaluation results, with auto-pagination helpers.

After an evaluation completes, retrieve per-prompt results via client.results.

All results (auto-paginated)

all_results = client.results.get_all(evaluation=evaluation)
# Returns List[Result] across all pages

# By ID
all_results = client.results.get_all_by_id(evaluation.id)

One page

response = client.results.get(evaluation=evaluation, page=1, page_size=100)
# By ID
response = client.results.get_by_id(evaluation_id, page=1, page_size=100)

# response is ResultsResponse:
#.results -> List[Result]
#.evaluation_id
#.pagination.total_count,.pagination.page_size,.pagination.total_pages

Result object

for r in all_results:
 print(r.subset) # benchmark subset (e.g., "math/algebra"); optional
 print(r.prompt) # the input prompt
 print(r.result) # model output
 print(r.truth) # expected answer
 print(r.score) # 0.0–1.0 typically
 print(r.duration) # Python timedelta
 print(r.metrics) # Dict[str, float], benchmark-specific

Manual pagination loop

page = 1
while True:
 response = client.results.get_by_id(evaluation_id, page=page, page_size=100)
 if not response or not response.results:
 break
 for r in response.results:
 process(r)
 if page >= response.pagination.total_pages:
 break
 page += 1

Concurrent across many evaluations

import asyncio
from layerlens import AsyncStratix

async def fetch_all(eval_ids):
 client = AsyncStratix()
 return await asyncio.gather(*[
 client.results.get_all_by_id(eid) for eid in eval_ids
 ])

On the Evaluation object

Evaluation instances have shortcut methods:

results_page = evaluation.get_results(page=1, page_size=50) # sync
all_results = evaluation.get_all_results() # sync
results_page = await evaluation.get_results_async(page=1) # requires AsyncStratix
all_results = await evaluation.get_all_results_async() # requires AsyncStratix

See also

• Evaluations
• SDK examples: retrieving-results
• SDK sample: paginated_results.py