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