Initialize Flameox in the fixed project root after it has been verified.

Return workspace status; on first use, follow WORKSPACE_NOT_FOUND recovery.

Rebuild only the disposable catalog from immutable corpus evidence.

Inspect flameox.toml without writing or executing anything.

    Use this after workspace initialization to decide whether to call configure_workload
    or list_declared_workflows. Invalid configuration is reported without replacement.
    

Write one validated named workload without executing it.

    Writes only the project workload configuration, preserves existing workloads and
    experiments, and returns the next discovery step. It never executes the command.
    Use operation='replace' with the current configuration_id to update an existing workload.
    

Create or replace a local inference-server declaration without starting it.

Create a bounded generic benchmark scenario, not a provider rollout replay engine.

List declared inference servers and scenarios without probing or execution.

Qualify one benchmark launcher and construct a typed execution plan.

Execute one qualified generic benchmark against a managed or existing-local server.

Page through prompt-free requests with one typed outcome and no derived status fields.

Build a diagnostic-only profile plan for one managed vLLM server.

Run one diagnostic-only profile window against a managed vLLM server.

List capabilities and setup actions scoped to a selected capture adapter.

    Omit adapter for a complete read-only inventory. In that mode, the per-capability setup
    fields and available_setup_adapters are informational only; select an adapter and call
    this tool again before mutating the managed environment. Managed setup never executes a
    workload.
    

Start detached capability provisioning and return its durable operation ID.

    Use the same idempotency key to reconnect after a lost request. Poll
    get_capability_setup for named phases, item outcomes, and the terminal receipt;
    cancel_capability_setup requests cleanup of owned work.
    

Read durable capability setup state after the original request disappears.

Request cancellation and cleanup of a server-owned capability setup operation.

Approve one installed third-party adapter by exact installed package identity.

    This records agent-created provenance under the workspace lock; it does not install a
    package, import plugin code, or execute a workload. Call list_capabilities again.
    

Inspect declared workload Python distributions in its bound interpreter.

This operation never installs packages, mutates an environment, or executes the workload; in other words, it never executes user code. The result includes an active preflight and tells the agent whether to plan capture or fix the declared environment outside Flameox.

Discover declared workflows before planning; this never runs them.

    With no arguments, list workloads. Pass kind='experiment' or
    kind='fault_experiment' to list declared experiments.
    

Inspect allowed parameters and validation metadata, then call the matching plan tool.

Bind one current capture without running it.

    The default auto mode runs the declared workload directly in the local environment and
    records that no enforced descendant containment was used. Use managed only when the
    project policy explicitly requires containment, and use trusted_local to request the
    same direct local execution explicitly. This tool never runs a measurement, but an
    adapter may execute a bounded compatibility probe such as NVBench ``--version``.
    

Run one current plan with side effects; the token is single-use, then get_run.

Start one current plan once; reconnect by run_id without keeping this call open.

Reconnect to bounded progress and lifecycle status for one detached run.

Cancel only the exact detached task owned by this server; repeated calls are safe.

After workflow and capability discovery, bind a declared experiment; then run it.

Execute all current trials from one single-use plan, then inspect get_experiment.

Bind a declared loopback Toxiproxy experiment and its exact toxic scenarios.

    Toxiproxy must already be prepared by start_capability_setup; planning neither
    downloads nor starts it.
    

Run every baseline and declared treatment through its managed loopback proxy.

Read one immutable completed fault experiment result.

Bind immutable input and approved reducer/predicate identities before execution.

Execute one bound reducer lifecycle and independently revalidate its candidate.

Reconnect to one immutable terminal reduction result.

Reconstruct one bounded experiment outcome from durable evidence.

Return one bounded page of immutable trials for an experiment.

Import one project-local artifact and preserve producer identity.

    Chrome traces with Torch profiler markers are identified automatically. Use
    kind='execution_trace' for Chrome/Torch traces, then run extract_perfetto before
    analyze_pytorch. If detection is ambiguous, set producer='torch.profiler'.
    

Preserve one native SARIF 2.1.0 report and project bounded static candidates.

The native report remains an immutable analysis-result artifact. Static candidates are source-scoped analyzer output, not Findings or confirmation of runtime behavior.

Return one cursor-bounded page of source-scoped static candidates.

Return one cursor-bounded page of provider-reported Triton autotune selections.

Validate preserved native bytes and create a semantically qualified import run.

Attach validated immutable correctness evidence to its exact producing run.

Preserve a native Metal trace bundle and bounded xctrace TOC export.

Import native compiler evidence and register one lineage per dump group.

Import an NVBench JSON and its provider-declared sidecars as one atomic bundle.

    Parses the primary JSON first to discover which sidecar files the
    document references.  Only those files are imported — no arbitrary
    sibling files are accepted.  Each sidecar's expected byte length is
    bound to ``declared_size * 4`` (float32) and verified after import.
    

Discover a filtered run cohort; follow next_cursor without changing filters.

Hydrate a run selected by list_runs; use an analyze_* tool for bounded interpretation.

Return bounded metadata and an opaque resource URI, never a host path or bytes.

Page every reduction that produced an artifact selected by get_artifact.

Read bounded UTF-8 process or validation output without exposing a host path.

Bind a bounded ordered pipeline to existing immutable run artifacts.

Discover bounded immutable pipelines by run, name, producer, or artifact.

Read one pipeline and bounded compatible comparison candidates.

Compare compatible ordered stages without returning native artifact content.

Render one bounded canonical proof summary and its Markdown view.

List bounded artifact metadata from one pinned corpus snapshot.

Create a durable diagnostic question.

List bounded current investigation projections.

Return one current investigation projection.

Record or revise a falsifiable hypothesis.

Return the current hypothesis revision.

Record or revise an evidence-linked finding.

Return the current finding revision with its exact evidence edges.

List bounded current finding projections.

Freeze a bounded cohort against one corpus snapshot.

Preview compatible frozen cohorts without persistence; use record_comparison to save.

Persist a reviewed comparison; use compare_run_sets for read-only preview.

Compare exact correctness metrics from two immutable run cohorts.

Persist a correctness comparison and exact input provenance.

Persist a curated analysis; use analyze_* first for read-only preview.

Analyze sampled-profile runs or artifacts for bounded source-linked hotspots; use extract_pyperf/query_measurements for benchmark_samples instead.

Analyze memory-profile runs or artifacts for peak, retained-end, and allocation evidence.

Analyze persisted Nsight Compute rule facts without reopening the native .ncu-rep.

The bounded result includes target qualification, coverage, provider provenance, and a concrete recapture selection when more evidence is needed.

Inspect one bounded execution-evidence collection. When next_cursor is present, repeat the same query with that cursor; use record_analysis to preserve a result.

Summarize normalized Perfetto evidence from a torch.profiler run or artifact.

    This read-only tool never extracts implicitly. If normalized rows are absent, follow
    the typed recovery result and call extract_perfetto for the exact run.
    

Analyze observed runtime launches, graph launches, kernels, and idle gaps.

Summarize an existing experiment without collecting missing trials.

Analyze an explicit filtered failure cohort read-only after list_runs discovery.

Return bounded source-linked direct callers for a frame.

Return bounded source-linked direct callees for a frame.

Return bounded representative stacks containing a frame.

Return bounded trace slices overlapping a declared time window.

Normalize an explicitly registered OTLP file artifact into evidence tables.

Return normalized OTLP spans overlapping a bounded time range.

Return bounded parent/child transitions without interpreting causality.

Return repeated span signatures as bounded derived evidence.

Return explicit timestamp, identity, and missing-parent evidence.

Return bounded process observations with explicit visibility coverage.

Return, but never execute, the installed native viewer command.

Query normalized measurements through reviewed filters and cursors.

Resolve a known typed reference; pass ref_type and its ID separately after discovery.

Validate manifests and schemas; optionally hash every payload.

Extract public pyperf run, warmup, loop, and value evidence.

Extract raw accelerator benchmark samples with explicit timing semantics.

Extract bounded evidence from a Node/V8 CPU profile.

Extract bounded evidence from a Node/V8 sampling heap profile.

Extract bounded per-case metrics from registered kernel-validation evidence.

Extract bounded findings from an official Compute Sanitizer XML report.

Extract NVBench sample times and frequencies from a preserved bundle.

The run must contain exactly one primary NVBench JSON artifact and zero or more sidecar artifacts declared by the JSON document.

Extract bounded metrics through NVIDIA's installed ncu_report interface.

Extract bounded prompt-free Mooncake request schedule evidence.

Extract prompt-free AIPerf requests or vLLM aggregate measurements.

Extract repeated startup, peak RSS, and package-grouped import evidence.

Extract pytest phase, fixture, worker, outcome, and failure-latency evidence.

Extract bounded execution-path evidence through coverage.py's public API.

Start durable Memray extraction; poll get_extraction with the operation ID.

Read durable extraction progress, outcome, and recovery after disconnect.

Request cancellation of one server-owned extraction and return bounded cleanup state.

Run versioned curated queries through a configured local Trace Processor.

Extract curated evidence from an imported official Nsight Systems SQLite export.

Extract bounded semantic observations emitted through flameox.sdk.