Integrations

Integrations

OpenRunner supports 30 framework integrations out of the box.

Training Frameworks

Framework	Import	Auto-logs
PyTorch	openrunner.watch(model)	Gradients, parameters
PyTorch Lightning	OpenRunnerLogger	Metrics, hyperparams, checkpoints
HuggingFace Transformers	report_to="openrunner"	Loss, eval metrics, model config
Keras / TensorFlow	OpenRunnerCallback	Epoch metrics, model summary
XGBoost	OpenRunnerCallback	Boosting metrics per round
CatBoost	OpenRunnerCallback	Train/eval metrics
LightGBM	OpenRunnerCallback	Metrics + model summary
scikit-learn	log_model() / log_summary()	Params, metrics, feature importance
FastAI	OpenRunnerCallback	Metrics per epoch
JAX/Flax	log_params()	Parameter norms, gradients
Optuna	OpenRunnerCallback	Trial params, objectives
Ultralytics (YOLO)	OpenRunnerCallback	mAP, loss, images
Stable Baselines 3	OpenRunnerCallback	Rewards, episode lengths
HF Accelerate	OpenRunnerTracker	Distributed training metrics
Diffusers	autolog()	Prompts, images, pipeline config
PyTorch Ignite	OpenRunnerLogger	Event-based metrics
TRL (RLHF)	OpenRunnerCallback	Reward, KL divergence
Gymnasium	monitor(env)	Episode rewards, video

LLM & Tracing

Framework	Import	Captures
LangChain	OpenRunnerTracer	Full chain execution, tool calls
OpenAI	trace_openai(client)	Tokens, cost, latency, messages
Anthropic	trace_anthropic(client)	Tokens, cost, latency
LlamaIndex	OpenRunnerHandler	Retrieval, generation spans
OpenAI Fine-tuning	sync_finetune()	Training metrics, model artifacts
OpenTelemetry	POST /api/v1/otel/v1/traces	Any OTLP-instrumented app

Audio & Voice AI

Framework	Import	Captures
Whisper	OpenRunnerCallback	WER, transcriptions, audio
Gladia	log_gladia_result() / GladiaTracer	Transcripts, speakers, confidence
TTS	OpenRunnerCallback	Synthesis metrics, audio samples
Voice Agent	VoiceAgentTracer	Turn latency, ASR/TTS spans
Forced Alignment	log_alignment()	Token timestamps, confidence

Usage Example

# PyTorch Lightning
from openrunner.integration.lightning import OpenRunnerLogger
logger = OpenRunnerLogger(project="my-org/my-project")
trainer = pl.Trainer(logger=logger, max_epochs=10)

# LangChain
from openrunner.integration.langchain import OpenRunnerTracer
tracer = OpenRunnerTracer(project="my-org/llm-app")
chain.invoke(input, config={"callbacks": [tracer]})

# OpenAI
from openrunner.integration.openai import trace_openai
client = trace_openai(openai.OpenAI())
response = client.chat.completions.create(...)