API Reference

Validates agent output against compliance gates. Runs G1–G8 based on domain, tier, and regulatory flags. Returns findings, violations, recommendations, and a tamper-evident evidence package.

Retrieve a past validation result by its UUID. Returns the full result including gate outcomes and evidence chain.

List recent validations. Paginated and filterable by status.

Generate a full compliance audit package for a regulation. Packages all evidence with a top-level SHA-256 attestation hash suitable for regulatory submission.

List all supported regulations with jurisdiction, effective date, and control counts.

List all 8 quality gates (G1–G8) with descriptions, check IDs, and applicability rules.

Global and per-key usage statistics: total validations, pass rate, average duration, and per-gate breakdowns.

Liveness probe. No authentication required. Use for uptime monitoring and pre-flight checks.

Check whether the authenticated API key has an active subscription and which plan it is on.

Current-period usage and projected cost for the authenticated key.

Full payment history for the authenticated API key with invoice IDs, amounts, and statuses.

Scan up to 100 texts in a single request. Accepts the same redact and mode options as /v1/scan.

Register a custom PII detection pattern for your workspace. Returns 409 if the pattern ID already exists.

Append an entry to the tamper-evident audit chain. Returns the entry ID, hash, and chain position.

Retrieve a specific audit log entry by its UUID, including hash and chain position.

List audit entries, newest first. Filterable by entity, action, actor, tags, and time range.

Verify hash chain integrity over the full chain or a specific range. Returns the first broken entry ID if tampering is detected.

Export audit entries as JSON or CSV with a top-level SHA-256 attestation hash for tamper-evident packaging.

Classify a single task. Returns domain assignment, tier (T1–T4), regulatory flag, AI flag, confidence score, and recommended agent routing chain.

Classify up to 50 tasks in one request. Returns an array of classification results plus batch-level timing statistics.

List all 12 supported domains with descriptions and applicable regulations. No authentication required.

List all agents grouped by hierarchy level (orchestrator → chief → VP → director → manager → team-lead → specialist → analyst → junior). No authentication required.

Record a single learning event (state, action, reward) into your tenant's learning namespace.

Record up to 100 learning events in a single call for bulk offline training scenarios.

Get the best action for a given state using Thompson Sampling with Q-Table fallback from your tenant's learned model.

Retrieve per-agent performance statistics from your tenant's learning namespace. Returns reward history, win rates, and confidence intervals.

Store a memory with content, tier, domain, tags, and importance score. Returns a UUID and the initial strength value.

Hybrid search combining semantic similarity (HNSW cosine), BM25 keyword matching, and recency scoring. Returns ranked results with relevance scores.

Retrieve a memory by ID. Also updates its last-accessed timestamp and decay state.

Permanently delete a memory from the tenant's vector store. Returns 204 No Content on success.

Run Ebbinghaus memory decay consolidation. Decays low-importance memories not accessed recently. Returns counts of strengthened and decayed memories.

Run a full bias audit on a prediction dataset. Computes DIR, SPD, EOD, and demographic parity. Returns violations and remediation recommendations.

Generate a formatted fairness report from a prior audit result. Includes executive summary, metric breakdowns, regulatory citations, and remediation recommendations.

List all fairness regulations supported by BiasGuard with thresholds, jurisdictions, and applicability rules. No authentication required.