Automated Prompt Regression Testing with Promptfoo Integration

You updated the prompt to support a new response format, and a week later the model started hallucinating on simple queries — accuracy dropped by 30%. Each prompt change can silently break scenarios that worked for years. In one project, changing a single word in a prompt reduced accuracy by 30% and required a week of manual testing to catch the regression. We use Promptfoo to build automatic regression testing in a couple of days and forget about such surprises forever. Promptfoo lets you run testing 5x faster than manual and provides objective metrics like LLM-rubric, ROUGE-N, and contains-any instead of subjective evaluation. It is an open-source CLI application with dozens of assertion types that integrates into any CI/CD pipeline.

Our engineers have executed 40+ LLM integration projects, and in all cases Promptfoo reduced regression detection time from a week to a few minutes. We configure tests for your specific scenarios: check for required phrases, absence of forbidden words, compare model responses, and measure similarity to a reference via embeddings. As a result, you get a transparent process where every PR is checked automatically, and prompt quality is controlled by metrics.

What problems we solve

• Regressions after prompt changes: a new condition can break other cases. Promptfoo runs hundreds of tests in seconds.
• Model instability: the provider updated the model, and it started producing a different format. Tests catch this before deployment.
• Subjective evaluation: instead of "seems to answer well" — objective metrics (contains, LLM-rubric, ROUGE).
• Lack of CI/CD for prompts: development without quality control — like committing without unit tests.
• RAG scenarios: verifying correctness of retrieval and relevance of answers based on vector search.
• QA for AI: ensuring the model does not produce incorrect or unsafe responses in production.

Why prompt regression testing counts

LLMs are probabilistic systems. One word in a prompt can change the entire semantics of the response. Without automated tests, you release changes blindly. Promptfoo provides guarantee that old scenarios still work and new ones don't break existing logic. This is especially important in production systems with thousands of requests per day. According to the official documentation, Promptfoo is an open-source tool for evaluating and testing LLM prompts, used in projects handling up to 10,000 daily requests.

How we configure Promptfoo

We start with an audit of current prompts: collect all used templates, highlight key scenarios, and write test cases with assertions. A typical config:

providers:
  - openai:gpt-4o
  - openai:gpt-4o-mini
  - anthropic:claude-3-5-sonnet-20241022

prompts:
  - "Summarize the following text in 3 sentences: {{text}}"
  - "Create a concise 3-sentence summary of: {{text}}"

tests:
  - vars:
      text: "Long article about machine learning..."
    assert:
      - type: contains
        value: "machine learning"
      - type: llm-rubric
        value: "The summary is accurate and covers the main points"
      - type: javascript
        value: "output.split('.').length >= 3"

  - vars:
      text: "Another test document..."
    assert:
      - type: not-contains
        value: "I cannot"
      - type: rouge-n
        value: reference_summary
        threshold: 0.5

We don't use all assertions indiscriminately, only those actually needed. We add custom checks via JavaScript: for example, that the response contains exactly 3 points in markdown. This yields 100% confidence in output format.

Integrating Promptfoo into CI/CD

Add one command to your workflow:

promptfoo eval --ci

GitHub Actions example:

- name: Run prompt tests
  run: npx promptfoo eval --ci

If a test fails, CI exits with an error — PR not merged. We set up automatic report viewing via promptfoo view so the team can see diffs between prompt versions.

What's included in our service

• Audit of current prompts and identification of critical scenarios.
• Writing 10–50 test cases with assertions tailored to your metrics.
• Promptfoo configuration and provider setup.
• CI/CD integration (GitHub Actions, GitLab CI) with notifications.
• Complete documentation and team training on adding new tests and interpreting results.
• Access to our private dashboard with test history and pass rates.
• Support for 30 days after deployment, including prompt troubleshooting.

Process

1. Prompt analysis — collect current templates, identify critical cases (up to 2 hours).
2. Test design — write 10–50 test cases with assertions for your metrics.
3. Implementation — create Promptfoo configuration, set up providers and variables.
4. Testing — run eval, fix failing tests, clarify requirements.
5. Deploy to CI/CD — add step to pipeline, configure notifications.

Timeline and cost

Work takes from 3 to 7 days depending on the number of prompts and complexity of assertions. Basic setup starts at $500; complex projects may reach $3,000. Cost is calculated individually after the audit. Get a free consultation — we'll evaluate your project in one day. Typical savings: automating regression testing reduces manual QA costs by 80% and cuts time-to-deploy by 75%.

Checklist: what to check in prompt tests

Promptfoo vs manual testing

import promptfoo

results = promptfoo.evaluate({
    "prompts": ["Classify sentiment: {{text}}"],
    "providers": ["openai:gpt-4o-mini"],
    "tests": [
        {
            "vars": {"text": "Great product!"},
            "assert": [{"type": "contains", "value": "positive"}]
        }
    ]
})

print(f"Pass rate: {results.stats.successes}/{results.stats.total}")

Our engineers with 5+ years of experience in AI/ML have executed over 40 LLM integration projects. We guarantee test coverage will be maintained and expanded. Promptfoo is the most flexible tool for prompt regression testing. Request a consultation — we'll show how Promptfoo fits into your workflow and estimate scope in one day. Turnkey approach: from test cases to CI/CD, with documentation and training. Get a free prompt audit — contact us!

MLOps: Infrastructure for Training, Deploying, and Monitoring ML Models

The model is trained, metrics — F1 0.94 on validation. Three months later in production, quality drops by 12%. No one knows when — there is no monitoring. It's impossible to retrain quickly — the training script is in a Jupyter notebook of a data scientist who has already left. Data for retraining is collected manually from three disparate systems. About half of the projects come to us with this pain. We build a turnkey MLOps platform: from experiment tracking to automatic deployment and data drift monitoring. We will assess your infrastructure in 1–2 weeks, and in 4–6 weeks you will get a basic MLOps core running in production. Our team has 10+ years of experience in ML infrastructure, over 50 implementations.

How does MLOps infrastructure benefit your ML projects?

Experiment Tracking and Reproducibility

Without tracking, an ML project turns into chaos: it's unclear which checkpoint is better, which hyperparameters were used, which dataset. Reproducing a result a month later is a quest.

Why is experiment tracking the foundation of reproducibility?

MLflow is an open source standard for tracking. It logs parameters, metrics, artifacts (models, graphs), and code. MLflow Model Registry is a centralized model storage with versioning and lifecycle stages (Staging → Production → Archived). Deployment via MLflow Serving or integration with external systems.

Typical initialization in code:

import mlflow

mlflow.set_experiment("fraud-detection-v2")
with mlflow.start_run():
    mlflow.log_params({"learning_rate": 3e-4, "batch_size": 64, "epochs": 10})
    mlflow.log_metric("val_f1", val_f1, step=epoch)
    mlflow.pytorch.log_model(model, "model")

This is the minimum. In production, we add logging of system metrics (GPU utilization, memory), dataset (hash, version), code (git commit hash). Weights & Biases — richer UI, collaboration features, sweep for hyperparameter optimization. MLflow — for on-premise deployment without external dependencies.

DVC (Data Version Control) — versioning of data and models on top of git. Data is stored in S3/GCS/Azure Blob, only metadata (hashes) in git. dvc repro reproduces the entire pipeline from raw data to metrics.

To ensure reproducibility of training, fix random seeds (torch.manual_seed, numpy.random.seed, random.seed) and record them in experiment metadata. Without this, debugging irregular results is painful. Log the dataset version (DVC hash) and git commit — then any experiment can be reproduced down to the byte.

Pipeline Orchestration: Kubeflow, Airflow, Prefect

A pipeline orchestrator becomes necessary when: A 100-line training script in cron is fine for simple tasks. But as soon as you have a multi-step pipeline (data loading → preprocessing → feature engineering → training → validation → deployment if quality above threshold), you need an orchestrator with retry logic, visualization, and alerts.

Kubeflow — Kubernetes-native orchestrator for ML (see Kubeflow). Each step is a Docker container. Supports parallel steps, conditional branches, artifacts between steps. Integrates with Katib (AutoML), KServe (serving), Feast (feature store).

Apache Airflow — more general DAG orchestrator. Wide ecosystem of operators (S3, Spark, DBT, Kubernetes). Easier to deploy if Airflow already exists in the company.

Prefect / Metaflow — less boilerplate. Prefect 2.x with @flow and @task decorators — quick start for small teams.

Typical training pipeline architecture on Kubeflow:

1. Data ingestion component — fetches data from S3/DB, validates schema via Great Expectations
2. Preprocessing component — transformations, normalization, train/val/test split
3. Training component — training on GPU, logging to MLflow
4. Evaluation component — metric calculation, comparison with baseline in Model Registry
5. Conditional deployment — deploy only if new model is better than current by >2% F1

Each component is a separate Docker image. Pipeline is versioned in git. Scheduled run (retraining once a week on new data) or manual.

Model Registry and Lifecycle Management

Model Registry is not just a checkpoint store. It is a centralized system that knows:

• Which model is currently in production (and with what metrics)
• History of all versions with training parameters
• Metadata: dataset, git commit, validation results
• Lifecycle stage: None → Staging → Production → Archived

MLflow Model Registry — standard. For enterprise — Vertex AI Model Registry (GCP), SageMaker Model Registry (AWS), Azure ML Model Registry.

Model promotion through stages: automatically move model to Staging after successful eval, then manual or automatic (during A/B test) promotion to Production. Rollback — switch to previous Production version in seconds.

Serving: From FastAPI to Triton Inference Server

Simple case. FastAPI + PyTorch/ONNX on one server — 80% of production ML deployments are exactly that. Sufficient for most tasks with load up to 100 req/s.

from fastapi import FastAPI
import onnxruntime as ort

app = FastAPI()
session = ort.InferenceSession("model.onnx", providers=["CUDAExecutionProvider"])

@app.post("/predict")
async def predict(request: PredictRequest):
    inputs = preprocess(request.text)
    outputs = session.run(None, {"input_ids": inputs})
    return {"label": postprocess(outputs)}

Triton Inference Server — production standard for high loads (500+ req/s). Dynamic batching, concurrent model execution, model ensemble. Supports TensorRT, ONNX, PyTorch TorchScript, TensorFlow SavedModel.

KServe — Kubernetes-native ML serving with autoscaling, canary deployments, A/B testing out of the box. Scale-to-zero for inactive models — savings on infrastructure up to 40% annually for a project with 10 models.

Monitoring: Data Drift, Model Drift, Infrastructure Metrics

Monitoring — what is usually done last and regretted first. Three levels.

Infrastructure monitoring. Latency (P50/P95/P99), throughput (req/s), error rate (4xx, 5xx), GPU/CPU utilization. Prometheus + Grafana — standard. Alert when P99 latency > threshold or error rate > 1%.

Data drift monitoring. Distribution of input data changes over time. Detect via PSI (Population Stability Index) for numerical features: PSI > 0.2 — strong drift. Chi-squared test for categorical, Kolmogorov-Smirnov test for continuous. Evidently AI — open source library with ready-made drift tests.

Model drift monitoring. If ground truth is delayed (e.g., we know conversion after a week) — monitor real metrics. If not — surrogate metrics: distribution of prediction scores, proportion of confident predictions.

Alerting. Three levels: INFO (minor drift, log it), WARNING (significant, notify team), CRITICAL (quality dropped below threshold — automatic switch to fallback model).

Why is data drift monitoring important?

Without it, you learn about model degradation only from user complaints or ringing SLA. A drift alert allows you to retrain the model in advance, before errors start causing losses. In one of our projects, PSI monitoring detected drift 2 days after a data source change — this saved the campaign.

Feature Store

Feature Store solves the training-serving skew problem. If preprocessing during training and inference is implemented in two different places — divergence is inevitable.

A Feature Store is needed when:

• Several models use the same features
• Features are computed from streaming data (real-time)
• Large team with different people on feature engineering and model training

Feast — open source Feature Store. Offline store (S3 + Parquet) for training, online store (Redis, DynamoDB) for low-latency inference. Feature definitions as code, materialization job syncs offline → online.

Tecton (commercial), Vertex AI Feature Store (GCP), SageMaker Feature Store (AWS) — managed options with less ops overhead.

CI/CD for ML

ML CI/CD is regular CI/CD plus specific ML steps.

ML-specific checks in CI:

• Reproducibility check: run training with a fixed seed, result must match
• Data validation: Great Expectations or Pandera on schema/distribution checks
• Model performance check: automatic eval on holdout, block merge if degradation > threshold
• Latency regression test: inference must meet SLA

GitOps for deployment. Merge to main → CI triggers training → eval → if passes → automatic deployment to Staging → smoke tests → manual promotion to Production or automatic upon successful canary.

Tools: GitHub Actions / GitLab CI for CI, ArgoCD for GitOps deployment on Kubernetes.

What's Included in MLOps Platform Development

We provide a full cycle of work, documentation, and team training.

Total time from audit to full MLOps platform: 3–5 months. Also possible phased launch: basic level (tracking + serving) in 4–6 weeks.

Cost is calculated individually based on data volume, number of models, and infrastructure requirements. Order an MLOps infrastructure audit — get a roadmap in 1–2 weeks. Contact us for a project assessment — we will send a preliminary estimate within 2 business days.

Note: warranty on architectural solutions — 12 months. We provide integration certificates with major cloud providers (AWS, GCP, Azure). During our work, we have not lost a single client after the first implementation — the experience of 50+ successful MLOps projects speaks for itself. Get a consultation on building an MLOps platform today.