All Classes and Interfaces
Class
Description
Backpressure-aware
EventSamplingPolicy (issue #82).Externally-supplied utilisation signal.
Health indicator for TnsAI agents.
OpenTelemetry instrumentation for TnsAI agents.
Structured logger for agent operations with automatic trace context propagation.
Response from a single agent in parallel validation.
History entry for one
ReviewAgentSpi invocation against a single
FileRecord (TNS-291).Top-level trace for an agent interaction.
Strategy for aggregating results from multiple LLM agents.
Captures every event at full fidelity (issue #82).
Redactor decorator that emits a RedactionAuditEvent
to a RedactionAuditListener every time the wrapped redactor
matches at least one pattern (issue #80).Thrown by
CostBudgetEnforcer when an LLM call would exceed
a configured CostBudget under BudgetPolicy.HARD_STOP
(issue #82).What the framework should do when a
CostBudget is exhausted
(issue #82).Snapshot of a
CostBudget's consumption (issue #82).Coarse health bucket.
Factory for the framework's built-in CWE matchers (TNS-291).
Multi-window burn-rate alerting thresholds (issue #92), per the SRE
workbook recipe — a single burn-rate signal is too noisy on its own,
so the framework requires a fast-window AND a slow-window signal to
fire before treating an SLO as actively burning.
Burn-rate sample for one
SLOState (issue #92): how fast is
the error budget being consumed in the short and long windows?Top-level orchestrator for the agent-driven code-review pipeline (TNS-291).
Builder for
CodeReviewPipeline.One executable step inside a
Sandbox (TNS-296).Pipes input through a chain of
Redactors in order (issue #80).Represents a conditional transition rule like "DEAD if neighbors invalid input: '<' 2".
Sandbox backend that runs commands inside a fresh
container per ContainerSandbox.execute(com.tnsai.quality.sandbox.Command) call (TNS-296).Container-runtime-backed
SandboxFactory (TNS-296).Interface for content moderation.
Tag identifying which kind of project-context file a scan operated on.
Declarative budget — "this scope can spend at most X USD per period;
here's what to do when exhausted" (issue #82).
Hook that enforces a CostBudget on every pre-LLM
action (issue #82).Resolves a
CostBudget's current BudgetState by
querying its CostBudgetStore for cumulative spend (issue #82).SPI for tracking cumulative LLM spend per
CostScope per
rolling window (issue #82).Per-spend scope key — a (tenantId, agentId, capabilityId) triple
used by
CostBudgetStore to bucket recorded spend (issue #82).Outcome of an
EventRateLimiter probe (issue #82).Default factory that creates
EvalHookManager instances.Default factory that creates
InvariantChecker instances.In-memory permission registry implementing
PermissionProvider.Default factory that creates
SecurityEnforcer instances.What an SLO-aware enforcement layer should do when the budget is
exhausted (issue #92).
Disable a specific feature flag for the duration of the exhaustion.
Just alert; let traffic through unchanged.
Reject new work outright (with a structured reason) until the
budget recovers.
Switch the agent to a cheaper / simpler LLM target identified by
its provider qualifier (e.g.
Drop the configured fraction of low-priority requests.
Content-level
ToolCallFilter that scans tool arguments for
destructive shell / git / filesystem commands hidden inside generic
tools like terminal_execute("rm -rf /") (issue #36).A single named pattern with its severity classification.
Severity tier of a destructive-command match.
Configuration for AES-256-GCM encryption used by SecurityEnforcer.
Builder for creating EncryptionConfig with custom settings.
The
enrich stage (TNS-291).Sampling policy that always captures problems and head-samples
everything else at a single rate (issue #82).
Manager for coordinating multiple evaluation hooks.
SPI for per-tenant event rate limiting (issue #82).
SPI for deciding whether to capture an event (issue #82).
Per-attempt scope description fed to
SLOScope.matches(EventScope).Output formats supported by the
export stage (TNS-291).The
export stage (TNS-291).File-based persistent audit store using JSON.
Per-file canonical state in the code-review pipeline (TNS-291).
Disk-backed
PipelineStateStore (TNS-291 phase 2).Filesystem mount point for a sandbox (TNS-296).
SPI for content guardrails that evaluate input and/or output.
Registry for pre-response and post-response guardrails.
Result of a guardrail evaluation.
Coin-flip-per-event sampling at a fixed rate (issue #82).
Result of a health check operation.
Interface for components that can report their health status.
Central registry for health indicators with aggregation support.
Represents the health status of a component.
Result of prompt injection detection.
Builder for InjectionDetectionResult.
Confidence level of detection.
Type of injection detected.
In-memory audit store for testing and development.
Process-local
CostBudgetStore (issue #82).Process-local
SLOEvaluator that holds attempt records in an
in-memory rolling buffer (issue #92).One observed attempt feeding the SLO calculation.
Runtime invariant checker for TnsAI roles.
Evaluates
InvariantExpression against a target object.Exception thrown when evaluation fails.
Represents a parsed invariant expression.
Logical AND (invalid input: '&'invalid input: '&').
Binary comparison (==, !
Comparison operators.
Field reference (e.g., "state", "numberOfLiveNeighbors").
Literal value (number, string, boolean, enum).
Method call (e.g., "isEmpty()").
Logical NOT (!).
Logical OR (||).
Parses invariant strings into
InvariantExpression AST.Exception thrown when parsing fails.
Bounds the number of distinct values a Prometheus label can take.
Builder.
Health indicator for LLM providers.
Builder for LLMHealthIndicator.
OpenTelemetry instrumentation for LLM calls following semantic conventions.
Builder for creating LLM spans with full configuration.
Types of log events that can be themed.
SLF4J-backed
RedactionAuditListener — emits one structured
log line per audit event (issue #80).Log levels supported by the listener — mirrors
Logger's
standard levels minus ERROR (audit success isn't an error).Interface for themed log message formatting.
Built-in log themes for fun developer experience.
SPI for static vulnerability matchers — regex / AST / dependency-version
checks that pre-filter files for the agent stage (TNS-291).
One location matched by a
MatcherSpi.Simple metrics collection for agent operations.
Result of content moderation check.
Action to take based on moderation.
Builder for ModerationResult.
Content categories for moderation.
Network access policy applied to a sandbox (TNS-296).
Allow only the listed hosts.
No network access.
Sandbox inherits the host's network.
A single observation within a trace.
Types of observations within a trace.
Decorator wrapping a
Sandbox so every ObservedSandbox.execute(com.tnsai.quality.sandbox.Command)
call also produces a SandboxExecutionEvent on the
configured SandboxExecutionListener (TNS-296).Bridge between TnsAI's Metrics and OpenTelemetry metrics.
Bridge between TnsAI's TraceContext and OpenTelemetry tracing.
A scope wrapper for OpenTelemetry spans that implements AutoCloseable.
LLMCallPublisher that exports each captured call as an
OpenTelemetry span plus a triplet of GenAI metrics
(token-usage counter, cost histogram, latency histogram).Builder for
OtlpLLMCallExporter.Executes the same task across multiple agents for validation and comparison.
Pattern-based content moderator using regex matching.
Builder for PatternBasedModerator.
Regex-driven
Redactor (issue #80).Three-tier permission level for tool execution.
Basic regex-based PII detection guardrail.
Read-only context passed to every
PipelineStage (TNS-291).Pipeline stage SPI (TNS-291).
Persistence layer for the code-review pipeline (TNS-291).
Audit-trail entry for one stage execution.
Level-aware sampling policy (issue #82).
Portable
Sandbox backend using JVM ProcessBuilder
(TNS-296).Default
SandboxFactory that backs sandboxes with the JVM's
ProcessBuilder (TNS-296).The
process stage (TNS-291).Prometheus metrics exporter for TnsAI.
Builder for PrometheusMetricsExporter.
Detects prompt injection attacks in LLM inputs.
Builder for PromptInjectionDetector.
AgentEventPublisher decorator that consults an
EventRateLimiter before delegating each publish call
(issue #82).In-memory
SandboxExecutionListener that records every
event for later inspection (TNS-296).AgentEventPublisher decorator that scrubs PII / secrets out of
every event before delegating (issue #80, sink integration #2 of 6).LLMCallPublisher decorator that scrubs PII / secrets out of
a captured LLMCallLog before delegating (TNS-417, follow-up
to TNS-374).MemoryStore decorator that scrubs PII / secrets out of every
write before delegating to the underlying store (issue #80, sink
integration #1 of 6).Audit record for one redaction call (issue #80).
Sink for
RedactionAuditEvents emitted by AuditingRedactor
(issue #80).Per-call context handed to a
Redactor (issue #80).One redacted match inside a string (issue #80).
One named pattern in a redactor's catalog (issue #80).
Default catalog of
RedactionPatterns shipped with the
framework (issue #80).Output of a
Redactor.scrubString(java.lang.String, com.tnsai.quality.redaction.RedactionContext) call (issue #80): the
redacted text plus the RedactionFindings the redactor
matched.Where the redacted content originated (issue #80).
SPI for scrubbing PII / secrets out of strings or structured values
before they leave the framework's internal boundaries — log lines,
trace attributes, memory writes, captured LLM prompts (issue #80).
Internal reflection utilities shared across tnsai-quality subpackages.
Generic regex-based
MatcherSpi (TNS-291).Token-bucket
EventRateLimiter backed by Resilience4j
(issue #82).Health indicator for system resources.
Per-sandbox resource caps (TNS-296).
Post-execution resource accounting from a sandbox (TNS-296).
The
revalidate stage (TNS-291).Default
ReviewAgentSpi backed by an LLMClient (TNS-291).SPI for an agent backend that analyses a
FileRecord and returns
structured ReviewFindings (TNS-291).One vulnerability discovered by the code-review pipeline (TNS-291).
Rule-driven
EventSamplingPolicy (issue #82).One rule entry.
Runtime engine for evaluating conditional transition rules.
Represents a rule violation.
Exception thrown when rules are violated.
Parses conditional rule strings into
ConditionalRule objects.Exception thrown when rule parsing fails.
AgentEventPublisher decorator that consults an
EventSamplingPolicy before delegating each publish call
(issue #82).Outcome of an
EventSamplingPolicy.sample(com.tnsai.quality.sampling.SamplingInput) call (issue #82).Capture this event at full fidelity.
Capture but flag as downsampled.
Drop this event entirely — no publication.
Frozen view of an event's metadata used by an
EventSamplingPolicy to decide whether to capture it (issue #82).LLMCallPublisher decorator that consults an
EventSamplingPolicy before delegating each captured call
(TNS-417, follow-up to TNS-374).Convenience factories for the bundled
EventSamplingPolicy
implementations (issue #82).A live sandbox instance (TNS-296).
Thrown when the sandbox infrastructure layer fails (TNS-296).
Per-execute observability event for the sandbox layer (TNS-296).
Sink for
SandboxExecutionEvents (TNS-296).SPI for creating
Sandbox instances (TNS-296).Reuses
Sandbox instances across multiple borrowers to
skip the cold-start cost (TNS-296).Lease wrapping a borrowed sandbox.
Outcome of a single
Sandbox.execute(com.tnsai.quality.sandbox.Command) call (TNS-296).Declarative specification of a sandbox instance (TNS-296).
The
scan stage (TNS-291).A score attached to a trace or observation.
Source that produced a score.
Types of score values.
Pre-LLM secret scrubbing filter.
Builder for configuring a SecretFilter.
Result of a scrub operation with metadata.
Defines a single secret detection pattern.
Enforces security policies defined via @Security annotation.
Sensitivity tier of a
RedactionFinding (issue #80).Severity classification for a
ReviewFinding (TNS-291).Declarative definition of one Service Level Objective (issue #92).
Computes the current
SLOState for a given SLODefinition
(issue #92).In-memory store of
SLODefinitions, looked up by id (issue #92).What an
SLODefinition measures (issue #92): the framework
as a whole, a single tenant, a specific agent inside a tenant, or
every implementation of a Capability interface.Single-agent-within-tenant scope.
Every implementation of a capability interface — useful for
cross-tenant SLOs like "all chat agents must hit p95 invalid input: '<' 3s".
Single-tenant scope.
Whole-framework scope — every event counts.
Point-in-time snapshot of an SLO's health, produced by
SLOEvaluator.evaluate(SLODefinition, Instant) (issue #92).Coarse-grained health bucket for an
SLOState (issue #92).What kind of metric an SLO measures (issue #92).
Canonical span-name and metric-name builder for TnsAI OpenTelemetry data.
Outcome of one
PipelineStage execution (TNS-291).Internal string utilities shared across tnsai-quality subpackages.
Numeric target an SLO must hold (issue #92).
Direction of comparison between observed value and target.
Redactor that dispatches each call to a per-tenant
Redactor based on the active RedactionContext (issue #80).EventSamplingPolicy that dispatches each call to a per-tenant
policy based on EventContext.tenantId() (issue #82).Logger wrapper that applies themed formatting to messages.
Central configuration for OpenTelemetry in TnsAI framework.
Builder for TnsAITelemetry configuration.
Glob-based tool filtering with allow/deny patterns.
Builder for ToolFilter.
Manages per-tool permission levels using the 3-tier model.
Distributed tracing context for tracking operations across agents.
Thread-local context for automatic parent-child nesting of observations.
AutoCloseable scope for try-with-resources pattern.
Span status values.
Metrics from parallel validation execution.
Result of input validation.
Result of parallel validation across multiple agents.
Builder for ValidationResult.
Represents a point where agents diverge.
Comprehensive input validation service.
Builder for ValidationService.
WebAssembly + WASI
SandboxFactory entry-point (TNS-296).