stormlog.infer

Inference profiling helpers for OpenAI-compatible serving endpoints.

Modules

analysis	Analysis helpers for inference profiling artifacts.
arrival_report	Per-case arrival accounting for inference reports.
arrivals	Arrival schedules for inference profiling workloads.
cache_state	Requested and verified prefix-cache state for each workload case.
cli	CLI for Stormlog inference profiling.
config	Configuration models for inference profiling.
correlation_accounting	Resolve inference evidence before computing overlap-aware GPU time.
correlation_capture	Optional server evidence capture using existing inference and run artifacts.
correlation_events	Versioned, engine-neutral inference execution records.
diagnosis_loop	Stalls of the vLLM engine loop, from the execution hook's raw records.
diagnosis_signals	Cheap signals that decide whether a window of scrapes looks like an incident.
diagnosis_thresholds	The versioned threshold table shared by online triggers and the diagnoser.
diagnosis_vocabulary	The closed vocabulary of inference diagnosis findings.
errors	Errors that tell the inference CLI which contract exit code a failure gets.
events	Inference profiling event records and JSONL persistence.
host_clock	Identify one host boot when deciding if wall timestamps share a clock.
latency_report	Per-case latency quantiles with sufficiency, and chunk-level streaming.
open_loop	Send requests at scheduled times, bounded by an in-flight limit.
openai_client	OpenAI-compatible Chat Completions client used by inference profiling.
otlp_wire	Count an OTLP/HTTP protobuf trace export's messages before it is parsed.
populations	Who a case's numbers are about, and the interval its rates divide by.
profile	Active inference profiling runner.
prompts	Deterministic prompts with controlled prefix sharing.
quantiles	Latency quantiles, how much data they need, and what failures do to them.
report_stats	Small statistics helpers shared by the inference reports.
samplers	Best-effort system telemetry samplers for inference profiling.
scrape_window	Aggregate one window of vLLM /metrics scrapes for one engine.
server_clock	Place server telemetry timestamps on the client artifact's clock.
server_collector	Optional on-host process and NVML memory collection for inference runs.
server_group	Declared groups of server processes, such as tensor-parallel GPU workers.
slo	SLO policies: latency criteria declared at an explicit boundary.
telemetry	Engine-neutral, scoped samples from an inference host.
tokens	Token counting helpers for inference profiling.
trace_capture	Bounded vLLM profiler windows during stormlog infer profile.
trace_import	Import profiler traces (Kineto, Nsight Systems) into an inference artifact.
trace_kineto	Import PyTorch/Kineto profiler traces as inference activity references.
trace_nsys	Import Nsight Systems SQLite exports as inference activity references.
trace_ranges	Name iteration ranges so a trace importer can link GPU work to iterations.
trace_torch	Bounded in-process PyTorch profiler capture for import with import-trace.
vllm_analysis	Explain a run's queueing, cache and token signals from vLLM's own telemetry.
vllm_execution	Reduce the vLLM execution hook's raw log into canonical correlation records.
vllm_execution_devices	Map a profiler trace to the GPU it ran on, from the vLLM hook's worker hellos.
vllm_execution_import	Import a vLLM execution-hook log into an inference artifact.
vllm_execution_log	Read the vLLM execution hook's raw log (docs/vllm_execution.md).
vllm_execution_report	Coverage of an imported vLLM execution log, as separate dimensions.
vllm_hook	Install the vLLM execution hook's patches; see docs/vllm_execution.md.
vllm_metrics	Prometheus text parsing and the vLLM metric catalog.
vllm_scraper	Scrape vLLM's /metrics during a profile and keep each response whole.
vllm_spans	Ingest vLLM's OpenTelemetry request spans.
vllm_telemetry	Records for vLLM native telemetry inside an inference artifact.
workload	The workload record: what traffic a run sent, so it can be repeated.
workload_report	Per-case prompt and length summaries for inference reports.