{"passport":{"unfragile":{"@version":"1.0","version":"2026-05","artifact":{"id":"tool_nijta","slug":"nijta","name":"Nijta","type":"product","url":"https://nijta.com","page_url":"https://unfragile.ai/nijta","categories":["voice-audio"],"tags":[],"pricing":{"model":"paid","free":false,"starting_price":null},"status":"active","verified":false},"capabilities":[{"id":"tool_nijta__cap_0","uri":"capability://safety.moderation.real.time.voice.anonymization.with.pii.stripping","name":"real-time voice anonymization with pii stripping","description":"Processes live audio streams during call recording to identify and remove personally identifiable information (names, account numbers, SSNs, credit card numbers) while preserving speech intelligibility and call context. Uses speaker diarization combined with entity recognition models trained on contact center lexicons to detect PII patterns in real-time, applying audio masking or synthetic voice replacement techniques to strip sensitive data without requiring post-processing delays.","intents":["I need to record customer support calls for quality assurance without storing personally identifiable information","I want to comply with GDPR and CCPA regulations that require PII removal from stored voice data","I need to enable agent training on real customer calls while protecting customer privacy","I want to reduce liability exposure from storing unmasked voice recordings containing sensitive financial or health data"],"best_for":["Contact center operations teams in regulated industries (healthcare, finance, insurance, telecommunications)","Compliance officers managing data retention policies for voice recordings","Customer support teams needing to balance quality assurance with privacy obligations"],"limitations":["Audio quality degradation occurs when masking PII — synthetic voice replacement may sound unnatural or introduce artifacts that impact agent training effectiveness","Real-time processing latency (typically 100-500ms) may cause slight delays in call recording startup or introduce buffering in live monitoring scenarios","Entity recognition accuracy depends on audio clarity and accent diversity — heavily accented speech or poor audio quality reduces PII detection rates below 95%","Cannot retroactively anonymize existing call recordings without re-processing, requiring architectural changes to legacy call recording systems","Specialized PII patterns (medical codes, insurance policy numbers) require custom model training per organization, increasing deployment complexity"],"requires":["Existing call recording infrastructure (VoIP system, PBX, or contact center platform with audio stream access)","Network bandwidth for real-time audio processing (minimum 128 kbps per concurrent call)","API integration capability or middleware layer to intercept call audio before storage","Compliance framework documentation (GDPR, CCPA, HIPAA, PCI-DSS) to define which data types require anonymization"],"input_types":["raw audio streams (PCM, G.711, G.729 codecs)","live call feeds from VoIP/PBX systems","pre-recorded audio files (WAV, MP3, OGG formats)"],"output_types":["anonymized audio streams (same codec as input)","metadata logs (PII detection timestamps, confidence scores, masking applied)","compliance audit reports (anonymization coverage %, PII types detected)"],"categories":["safety-moderation","audio-processing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_nijta__cap_1","uri":"capability://data.processing.analysis.speaker.diarization.and.voice.identity.separation","name":"speaker diarization and voice identity separation","description":"Automatically identifies and segments different speakers in a multi-party call recording, assigning unique speaker labels to each participant (agent, customer, supervisor). Uses neural speaker embedding models (typically x-vector or speaker verification networks) to distinguish voices based on acoustic characteristics, enabling selective anonymization of only customer voices while preserving agent identification for quality assurance purposes.","intents":["I need to identify which parts of a call contain customer speech vs agent speech to apply different anonymization rules","I want to preserve agent voice identity for training and performance evaluation while anonymizing customer data","I need to detect when a supervisor joins a call and apply appropriate privacy rules to their participation","I want to generate speaker-labeled transcripts that show who said what without revealing customer identity"],"best_for":["Contact center quality assurance teams needing to audit agent performance on real calls","Compliance teams requiring granular control over which participants get anonymized","Training departments building agent coaching materials from real customer interactions"],"limitations":["Speaker diarization accuracy degrades with more than 4-5 concurrent speakers, limiting use in conference call scenarios","Background noise, overlapping speech, or heavy accents reduce speaker separation accuracy to 85-90% in real-world contact center environments","Cannot reliably distinguish between similar voices (e.g., family members, colleagues with similar acoustic profiles) without explicit speaker enrollment","Requires minimum 10-15 seconds of speech per speaker to establish reliable speaker embeddings, causing delays in short call segments"],"requires":["Audio input with clear speaker turns (minimum SNR of 20dB for reliable diarization)","Pre-trained speaker embedding model (x-vector, ECAPA-TDNN, or similar) compatible with target languages","Computational resources for real-time neural inference (GPU recommended for sub-100ms latency per call)"],"input_types":["multi-channel audio (separate agent/customer channels preferred)","mono audio with multiple speakers","audio files or live streams with known speaker count"],"output_types":["speaker-labeled audio segments with timestamps","speaker turn metadata (speaker ID, start time, end time, confidence score)","speaker-attributed transcripts (who said what)"],"categories":["data-processing-analysis","audio-processing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_nijta__cap_2","uri":"capability://data.processing.analysis.entity.recognition.and.pii.pattern.detection.in.speech","name":"entity recognition and pii pattern detection in speech","description":"Identifies personally identifiable information patterns in real-time speech using acoustic-to-text conversion combined with named entity recognition (NER) models trained on financial, healthcare, and insurance lexicons. Detects sequences like credit card numbers (Luhn algorithm validation), social security numbers, medical codes, account numbers, and names by analyzing both the transcribed text and acoustic patterns (e.g., digit-by-digit spelling patterns), enabling high-confidence PII detection even in noisy audio.","intents":["I need to automatically detect when a customer provides sensitive information (credit card, SSN, account number) during a call","I want to mask only the specific PII mentioned in a call, not the entire conversation","I need to generate compliance reports showing what types of PII were detected and anonymized","I want to flag calls where PII was mentioned but not properly anonymized for manual review"],"best_for":["Financial services contact centers handling payment information and account access","Healthcare customer support teams managing patient data and medical record access","Insurance companies processing claims and policy information over phone","Compliance teams auditing PII handling practices across recorded calls"],"limitations":["Entity recognition accuracy varies by PII type — credit card detection is ~98% accurate but medical code detection drops to 85-90% due to domain-specific terminology","Requires accurate speech-to-text conversion as a prerequisite; ASR errors (especially with accents or technical terms) cascade into missed PII detection","Cannot distinguish between legitimate business context (agent reading back a number) and actual PII disclosure without additional context rules","False positives occur when numbers or names mentioned in non-sensitive contexts (e.g., 'call back at 555-1234' or 'ask for John') are flagged as PII","Requires custom model training per organization to handle industry-specific terminology and naming conventions"],"requires":["Speech-to-text engine (ASR) with minimum 90% word error rate for contact center audio","Named entity recognition model trained on target industry (financial, healthcare, insurance)","Validation rules for PII patterns (Luhn algorithm for credit cards, checksum validation for SSNs)","Custom entity dictionaries for organization-specific account numbers, product codes, or internal identifiers"],"input_types":["real-time audio streams converted to text via ASR","pre-transcribed call transcripts","acoustic features (pitch, duration patterns for digit-by-digit speech)"],"output_types":["PII detection results with confidence scores and timestamps","masked/redacted text and audio segments","PII audit logs (type, value, detection confidence, action taken)","compliance reports (PII coverage %, false positive rate, detection latency)"],"categories":["data-processing-analysis","safety-moderation"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_nijta__cap_3","uri":"capability://data.processing.analysis.audio.masking.and.synthetic.voice.replacement","name":"audio masking and synthetic voice replacement","description":"Applies selective audio anonymization techniques to detected PII segments using either spectral masking (replacing frequency bands with noise) or synthetic voice replacement (generating natural-sounding speech to replace PII utterances). Uses voice synthesis models (TTS) to generate replacement audio that matches the original speaker's acoustic characteristics (pitch, speaking rate, accent) to maintain call naturalness while removing identifying information.","intents":["I want to mask sensitive information in recorded calls while keeping the rest of the conversation intelligible","I need the anonymized audio to sound natural for agent training purposes, not obviously redacted","I want to replace customer-spoken PII with synthetic speech that matches the original speaker's voice characteristics","I need to generate multiple anonymization versions of the same call for different compliance requirements"],"best_for":["Contact centers using recorded calls for agent training and quality assurance","Organizations requiring high-fidelity anonymization that doesn't degrade training effectiveness","Compliance teams needing to share call recordings with third parties (auditors, regulators) without exposing PII"],"limitations":["Synthetic voice replacement introduces 50-200ms latency per PII segment, making real-time processing challenging for calls with frequent PII mentions","Audio quality of synthetic replacement varies — accent matching and prosody preservation are imperfect, potentially sounding unnatural to trained listeners","Spectral masking (noise-based approach) is faster but degrades intelligibility of surrounding speech, making it unsuitable for training materials","Requires speaker voice enrollment (5-10 seconds of clean speech) to generate convincing synthetic replacements, adding setup complexity","Cannot perfectly preserve emotional tone or emphasis in replaced speech, potentially altering call context for training purposes"],"requires":["Text-to-speech (TTS) engine with speaker adaptation capability (e.g., Tacotron 2, FastPitch with speaker embeddings)","Speaker voice samples for enrollment (minimum 5-10 seconds per speaker)","Audio processing pipeline for seamless splicing of synthetic audio into original call","Computational resources for real-time TTS inference (GPU recommended for sub-200ms latency)"],"input_types":["audio segments containing PII (with timestamps and speaker identity)","text transcription of PII to be replaced","speaker voice samples for voice adaptation"],"output_types":["anonymized audio with replaced PII segments","quality metrics (naturalness score, spectral similarity to original, intelligibility rating)","metadata (masking technique used, replacement confidence, audio quality impact)"],"categories":["data-processing-analysis","audio-processing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_nijta__cap_4","uri":"capability://safety.moderation.compliance.audit.logging.and.reporting","name":"compliance audit logging and reporting","description":"Automatically generates detailed audit logs of all anonymization operations, including what PII was detected, when it was detected, what anonymization technique was applied, and confidence scores for each decision. Produces compliance reports mapping anonymization coverage to regulatory requirements (GDPR Article 32, CCPA Section 1798.100, HIPAA 45 CFR 164.512), enabling organizations to demonstrate data protection practices to auditors and regulators.","intents":["I need to prove to auditors that we're complying with GDPR/CCPA requirements for voice data anonymization","I want to generate compliance reports showing what percentage of calls were anonymized and what PII types were detected","I need to track which calls were anonymized, when, and by what method for regulatory inquiries","I want to identify gaps in anonymization coverage (calls with PII that weren't masked) for remediation"],"best_for":["Compliance and legal teams managing regulatory obligations for voice data","Internal audit teams conducting data protection assessments","Organizations preparing for regulatory inspections or audits","Risk management teams quantifying data protection effectiveness"],"limitations":["Audit logs can grow very large (100+ MB per 1000 calls) requiring external storage and archival solutions","Reports are generated post-hoc and cannot retroactively fix anonymization failures in already-stored calls","Compliance mapping is generic — organization-specific regulatory requirements may not be fully captured by standard report templates","Audit logs themselves contain metadata that could be sensitive (call timestamps, speaker counts, PII types detected), requiring additional protection","No real-time alerting for anonymization failures — organizations must actively review logs to identify compliance gaps"],"requires":["Secure audit log storage (encrypted database or immutable log store)","Compliance framework definitions (GDPR, CCPA, HIPAA, PCI-DSS requirements mapped to anonymization rules)","Report generation infrastructure (templating engine, data aggregation pipeline)","Access control for audit logs (role-based access to compliance reports)"],"input_types":["anonymization operation logs (PII detected, technique applied, confidence scores)","call metadata (duration, participants, timestamp, recording ID)","compliance framework definitions"],"output_types":["detailed audit logs (JSON or CSV format with full operation history)","compliance reports (PDF or HTML with coverage metrics, regulatory mapping, risk assessment)","executive summaries (anonymization coverage %, PII types detected, false positive rates)","remediation reports (calls with anonymization failures, recommended actions)"],"categories":["safety-moderation","data-processing-analysis"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_nijta__cap_5","uri":"capability://tool.use.integration.integration.with.contact.center.platforms.and.call.recording.systems","name":"integration with contact center platforms and call recording systems","description":"Provides native integrations or middleware adapters for major contact center platforms (Genesys, Avaya, Five9, NICE) and call recording systems (Verint, Calabrio, Aspect), enabling real-time anonymization without requiring custom development. Uses standard APIs (CTI, media stream APIs) to intercept call audio, apply anonymization, and return processed audio to the recording system, maintaining compatibility with existing call workflows and quality assurance tools.","intents":["I want to add voice anonymization to our existing contact center without replacing our current platform","I need anonymization to work transparently in our call recording workflow without manual intervention","I want to ensure anonymized calls are properly labeled and tracked in our QA system","I need to integrate anonymization with our existing compliance and audit workflows"],"best_for":["Contact center operations teams with existing platform investments (Genesys, Avaya, Five9)","Organizations with complex call recording infrastructure requiring minimal disruption","Teams lacking in-house development resources to build custom integrations"],"limitations":["Integration complexity varies significantly by platform — some platforms have limited media stream access, requiring workarounds or custom middleware","Real-time processing adds latency to call recording startup (typically 100-500ms), which may be noticeable in some platform configurations","Integrations are platform-specific — organizations using multiple contact center platforms require separate integrations for each","API rate limits or bandwidth constraints in some platforms may limit concurrent anonymization capacity","Updates to contact center platform APIs may break integrations, requiring maintenance and re-certification"],"requires":["Compatible contact center platform (Genesys, Avaya, Five9, NICE, or similar with media stream API access)","Compatible call recording system (Verint, Calabrio, Aspect, or similar with integration points)","Network connectivity between anonymization service and contact center platform","API credentials and permissions for contact center platform","Deployment environment (cloud or on-premises) matching organization's infrastructure"],"input_types":["call audio streams from contact center platform","call metadata (caller ID, call duration, agent ID, call type)","configuration parameters (anonymization rules, PII types to detect)"],"output_types":["anonymized audio streams returned to call recording system","anonymization metadata (PII detected, masking applied, confidence scores)","integration status logs (connection health, processing latency, error rates)"],"categories":["tool-use-integration","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_nijta__cap_6","uri":"capability://data.processing.analysis.multi.language.and.accent.adaptive.speech.processing","name":"multi-language and accent-adaptive speech processing","description":"Processes voice data across multiple languages and accents using language-agnostic acoustic models and multilingual speech-to-text engines, adapting PII detection patterns and voice synthesis to match target language phonetics and prosody. Automatically detects language and accent from call audio, selecting appropriate ASR models and entity recognition rules to maintain anonymization accuracy across diverse speaker populations.","intents":["I need to anonymize calls in multiple languages (English, Spanish, French, etc.) with consistent accuracy","I want to handle diverse accents and regional dialects without degrading anonymization quality","I need to detect PII patterns that vary by language (e.g., different credit card formats, ID number structures)","I want to generate synthetic replacement speech that sounds natural in the original speaker's accent and language"],"best_for":["Global contact centers serving multilingual customer bases","Organizations with diverse workforce and customer populations","International companies with regulatory obligations in multiple jurisdictions"],"limitations":["Multilingual ASR accuracy varies significantly by language — English and Spanish are ~95% accurate, but less common languages (Vietnamese, Tagalog) drop to 80-85%","Accent adaptation requires training data for target accents; generic models perform poorly on heavily accented speech (e.g., thick regional accents)","PII pattern detection is language-specific — credit card validation, SSN formats, and medical codes vary by country, requiring custom rules per region","Voice synthesis quality degrades for less-resourced languages; accent-adaptive TTS is only available for major languages (English, Spanish, French, Mandarin)","Language detection errors (e.g., confusing Spanish with Portuguese) cascade into incorrect entity recognition and poor anonymization"],"requires":["Multilingual speech-to-text engine (e.g., Whisper, Google Cloud Speech-to-Text, Azure Speech Services) supporting target languages","Language-specific entity recognition models for PII detection","Multilingual voice synthesis engine with accent adaptation capability","Language detection model (automatic language identification from audio)","Regional PII validation rules (credit card formats, ID structures, medical codes per country)"],"input_types":["audio in multiple languages and accents","language hints or metadata (if available)","regional configuration (country/region for PII pattern selection)"],"output_types":["anonymized audio with language-appropriate masking","language detection results with confidence scores","PII detection results with language-specific patterns applied","accent-adaptive synthetic replacement audio"],"categories":["data-processing-analysis","audio-processing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_nijta__cap_7","uri":"capability://data.processing.analysis.quality.assurance.and.audio.fidelity.monitoring","name":"quality assurance and audio fidelity monitoring","description":"Continuously monitors anonymized audio quality using objective metrics (spectral similarity, speech intelligibility scores, signal-to-noise ratio) and subjective evaluation (MOS scores from human raters or automated speech quality models). Detects anonymization artifacts (clicks, pops, unnatural transitions) and flags calls where anonymization degraded audio quality below acceptable thresholds, enabling quality control and continuous improvement of anonymization algorithms.","intents":["I want to ensure anonymized calls maintain sufficient audio quality for agent training purposes","I need to detect when anonymization introduces artifacts or unnatural-sounding speech","I want to measure the impact of anonymization on call intelligibility and training effectiveness","I need to identify calls where anonymization failed or degraded quality for manual review"],"best_for":["Quality assurance teams using anonymized calls for agent training","Organizations concerned about training effectiveness impact from anonymization","Compliance teams needing to demonstrate that anonymization doesn't compromise call quality"],"limitations":["Objective audio quality metrics (spectral similarity, SNR) don't always correlate with subjective listening experience — a call may pass objective tests but sound unnatural to human listeners","MOS (Mean Opinion Score) evaluation requires human raters, making continuous quality monitoring expensive and time-consuming","Automated speech quality models are trained on specific audio conditions and may not generalize to contact center environments with background noise","Quality thresholds are subjective — what constitutes 'acceptable' audio quality varies by use case (training vs compliance archival)","Monitoring adds computational overhead (typically 10-20% additional latency per call)"],"requires":["Reference audio quality metrics (baseline intelligibility scores, acceptable artifact thresholds)","Speech quality assessment models (e.g., PESQ, STOI, or neural quality estimators)","Optional: human rater pool for MOS evaluation or subjective quality assessment","Quality threshold definitions (minimum intelligibility score, maximum artifact level)"],"input_types":["original audio (pre-anonymization)","anonymized audio (post-anonymization)","anonymization metadata (technique used, PII segments replaced)"],"output_types":["quality metrics (spectral similarity, intelligibility score, SNR, artifact detection)","quality flags (pass/fail against thresholds, artifact warnings)","quality reports (average scores across calls, trend analysis, problem areas)","remediation recommendations (alternative anonymization techniques, re-processing suggestions)"],"categories":["data-processing-analysis","safety-moderation"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_nijta__cap_8","uri":"capability://safety.moderation.selective.anonymization.with.role.based.rules","name":"selective anonymization with role-based rules","description":"Applies different anonymization rules based on speaker role (customer, agent, supervisor) and call context (payment processing, account access, general inquiry), enabling fine-grained control over what data gets anonymized. Uses speaker diarization and call classification to determine applicable rules, allowing organizations to preserve agent identity for training while anonymizing only customer PII, or to apply stricter rules during sensitive call types (payment processing) vs routine inquiries.","intents":["I want to anonymize only customer data while preserving agent identity for training and performance evaluation","I need stricter anonymization rules for payment processing calls vs general customer service calls","I want to apply different anonymization rules based on call classification (complaint, billing, technical support)","I need to preserve supervisor or manager identity while anonymizing customer and agent data for compliance"],"best_for":["Contact centers with granular compliance requirements (different rules for different call types)","Organizations using calls for agent training and performance evaluation","Compliance teams needing context-aware anonymization rules"],"limitations":["Call classification accuracy affects rule application — misclassified calls may receive incorrect anonymization rules","Role-based rules require manual configuration per organization, increasing deployment complexity","Conflicting rules (e.g., preserve agent identity but anonymize agent-spoken PII) require careful design to avoid compliance gaps","Dynamic rule changes (e.g., stricter rules during high-risk periods) require real-time rule updates and testing","Audit trails become more complex when tracking which rules were applied to which calls"],"requires":["Call classification model (payment processing, account access, complaint, technical support, etc.)","Speaker role identification (customer, agent, supervisor, third-party)","Rule engine for applying context-specific anonymization policies","Rule configuration interface for compliance teams to define and update rules"],"input_types":["call audio with speaker diarization results","call metadata (call type, duration, participants)","anonymization rule definitions (role-based, context-based)"],"output_types":["anonymized audio with role-specific rules applied","rule application logs (which rules were applied to which segments)","compliance reports (coverage by rule type, rule effectiveness)"],"categories":["safety-moderation","data-processing-analysis"],"confidence":0.5,"matches":0,"success_rate":0}],"trust":{"score":39,"verified":false,"data_access_risk":"high","permissions":["Existing call recording infrastructure (VoIP system, PBX, or contact center platform with audio stream access)","Network bandwidth for real-time audio processing (minimum 128 kbps per concurrent call)","API integration capability or middleware layer to intercept call audio before storage","Compliance framework documentation (GDPR, CCPA, HIPAA, PCI-DSS) to define which data types require anonymization","Audio input with clear speaker turns (minimum SNR of 20dB for reliable diarization)","Pre-trained speaker embedding model (x-vector, ECAPA-TDNN, or similar) compatible with target languages","Computational resources for real-time neural inference (GPU recommended for sub-100ms latency per call)","Speech-to-text engine (ASR) with minimum 90% word error rate for contact center audio","Named entity recognition model trained on target industry (financial, healthcare, insurance)","Validation rules for PII patterns (Luhn algorithm for credit cards, checksum validation for SSNs)"],"failure_modes":["Audio quality degradation occurs when masking PII — synthetic voice replacement may sound unnatural or introduce artifacts that impact agent training effectiveness","Real-time processing latency (typically 100-500ms) may cause slight delays in call recording startup or introduce buffering in live monitoring scenarios","Entity recognition accuracy depends on audio clarity and accent diversity — heavily accented speech or poor audio quality reduces PII detection rates below 95%","Cannot retroactively anonymize existing call recordings without re-processing, requiring architectural changes to legacy call recording systems","Specialized PII patterns (medical codes, insurance policy numbers) require custom model training per organization, increasing deployment complexity","Speaker diarization accuracy degrades with more than 4-5 concurrent speakers, limiting use in conference call scenarios","Background noise, overlapping speech, or heavy accents reduce speaker separation accuracy to 85-90% in real-world contact center environments","Cannot reliably distinguish between similar voices (e.g., family members, colleagues with similar acoustic profiles) without explicit speaker enrollment","Requires minimum 10-15 seconds of speech per speaker to establish reliable speaker embeddings, causing delays in short call segments","Entity recognition accuracy varies by PII type — credit card detection is ~98% accurate but medical code detection drops to 85-90% due to domain-specific terminology","builder identity is not verified yet","no observed match outcomes yet"],"rank_breakdown":{"adoption":0.31666666666666665,"quality":0.67,"ecosystem":0.15000000000000002,"match_graph":0.25,"freshness":0.75,"weights":{"adoption":0.25,"quality":0.25,"ecosystem":0.1,"match_graph":0.35,"freshness":0.05}},"observed_outcomes":{"matches":0,"success_rate":0,"avg_confidence":0,"top_intents":[],"last_matched_at":null},"maintenance":{"status":"active","updated_at":"2026-05-24T12:16:31.858Z","last_scraped_at":"2026-04-05T13:23:42.560Z","last_commit":null},"community":{"stars":null,"forks":null,"weekly_downloads":null,"model_downloads":null,"model_likes":null}},"distribution":{"claim_url":"https://unfragile.ai/submit?claim=nijta","compare_url":"https://unfragile.ai/compare?artifact=nijta"}},"signature":"z3fZkTRCyjxqaXrplBbIO5qlXVImygv+10gEkDE5UjiY+hnQUKmr2SFaGDgzcJ8SsC9TMzQelYlonn5a5yxmAg==","signedAt":"2026-06-21T15:06:46.805Z","signedBy":"unfragile.ai","version":1},"_links":{"self":"https://unfragile.ai/api/v1/passport/nijta","artifact":"https://unfragile.ai/nijta","verify":"https://unfragile.ai/api/v1/verify?slug=nijta","publicKey":"https://unfragile.ai/api/v1/trust-passport-public-key","spec":"https://unfragile.ai/trust","schema":"https://unfragile.ai/schema.json","docs":"https://unfragile.ai/docs"}}