Skip to content

Latest commit

 

History

History
 
 

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 
 
 

README.md

Sample Data

The SDK samples reference pre-built trace files, evaluation datasets, and industry-specific test data. This directory provides all the data needed to run the samples without connecting to a live AI provider. Use these files to test SDK operations locally, seed your LayerLens workspace with representative data, or build automated test suites for your evaluation pipelines.

Prerequisites

No additional dependencies are required. The data files are consumed by samples in other directories via relative path references.

Quick Start

Upload a trace file to your LayerLens workspace:

export LAYERLENS_STRATIX_API_KEY=your-api-key
python -c "from layerlens import Stratix; Stratix().traces.upload('samples/data/traces/simple_llm_trace.json')"

Expected output: the trace ID of the newly uploaded trace record.

Traces

File Description
traces/simple_llm_trace.json A single-agent OpenAI support-assistant trace with one LLM call and a cost record. The simplest structured trace for getting started.
traces/rag_pipeline_trace.json A LangChain retrieval-augmented generation trace with retriever, reranker, and answer-synthesis agents handed off across multiple spans.
traces/multi_agent_trace.json A CrewAI multi-agent trace where a researcher, fact-verifier, and analyst collaborate on an executive brief with peer review and one corrected error.
traces/error_trace.json A failed LangChain trace covering context-length errors, rate-limit retries with exponential backoff, fallback agent handoff, and final failure with diagnostic guidance.
traces/example_traces.jsonl A collection of example traces in JSONL format for batch processing samples.
traces/batch_traces.jsonl Fifty structured traces across multiple frameworks, models, and statuses. Designed for batch ingestion testing.

Datasets

File Description
datasets/golden_test_set.jsonl Ten curated question-answer pairs for evaluation and regression testing. Each entry includes an expected answer for judge validation.
datasets/generic_qa.jsonl A larger QA dataset spanning factual, reasoning, analytical, and creative categories. Suitable for benchmark runs and model comparison.

Industry Data

Domain-specific evaluation datasets with expected outcomes for judge testing. Each file is referenced by the corresponding sample in samples/industry/.

File Domain
industry/education_essays.jsonl Education -- student essays with grading rubrics
industry/healthcare_patient_cases.jsonl Healthcare -- patient cases with expected diagnoses
industry/healthcare_triage.jsonl Healthcare -- emergency triage with acuity levels
industry/financial_loans.jsonl Finance -- loan applications with risk ratings
industry/financial_transactions.jsonl Finance -- transactions with fraud indicators
industry/legal_contracts.jsonl Legal -- contract clauses with risk assessments
industry/legal_research.jsonl Legal -- research documents with analysis
industry/insurance_claims.jsonl Insurance -- claims processing data
industry/government_eligibility.jsonl Government -- eligibility determination cases
industry/retail_products.jsonl Retail -- product recommendations with user profiles
industry/energy_grid.jsonl Energy -- grid performance and diagnostics
industry/manufacturing_equipment.jsonl Manufacturing -- predictive maintenance data
industry/media_moderation.jsonl Media -- content moderation decisions
industry/real_estate_listings.jsonl Real estate -- property listings with valuations
industry/telecom_interactions.jsonl Telecom -- customer service interactions
industry/travel_bookings.jsonl Travel -- booking transactions with preferences