{"id":41521,"date":"2026-05-05T15:33:14","date_gmt":"2026-05-05T07:33:14","guid":{"rendered":"https:\/\/algorithm-atelier.com\/?p=41521"},"modified":"2026-05-05T15:33:14","modified_gmt":"2026-05-05T07:33:14","slug":"ai-prototype-architecture-for-an-amanah-companion-system","status":"publish","type":"post","link":"https:\/\/mithaqpraxis.com\/articles\/ai-prototype-architecture-for-an-amanah-companion-system\/","title":{"rendered":"AI Prototype Architecture for an Amanah Companion System"},"content":{"rendered":"

Amanah Companions should not begin with a humanoid body.<\/p>\n

They should begin with an architecture.<\/p>\n

Before we ask whether an AI companion can stand in the room, we need to ask whether the system behind it can be trusted with care knowledge.<\/p>\n

Can it protect privacy?
\nCan it preserve dignity?
\nCan it keep human authority visible?
\nCan it distinguish care memory from surveillance?
\nCan it support communication without speaking over the person?
\nCan it escalate safety concerns without becoming the decision-maker?<\/p>\n

That means the first prototype should not be a robot.<\/p>\n

The first prototype should be a governed care continuity system.<\/p>\n

A body may come later.<\/p>\n

The spine comes first.<\/p>\n

Prototype Goal<\/h2>\n

The first Amanah Companion prototype should answer one question:<\/p>\n

Can a human-led AI system help preserve, organize, and responsibly use care knowledge for a vulnerable person without turning that person into a dataset?<\/strong><\/p>\n

This is not a medical device claim.<\/p>\n

It is not a replacement caregiver.<\/p>\n

It is not a diagnostic tool.<\/p>\n

It is a care continuity prototype: a system for Care Profiles, Communication Maps, Sensory Maps, Guardian Gates, Care Memory Ledgers, source trace, audit logs, and human-reviewed AI support.<\/p>\n

Why Start With Software, Not a Body?<\/h2>\n

Assistive technology already includes both physical and digital tools. WHO describes assistive products as tools that maintain or improve functioning related to cognition, communication, hearing, mobility, self-care, and vision, supporting health, well-being, inclusion, and participation. WHO also notes that many people need more than one assistive product, making integrated services important. (World Health Organization<\/a>)<\/p>\n

That supports the Amanah Companion path.<\/p>\n

The first useful layer may be:<\/p>\n

care dashboard
\nparent\/guardian app
\ncaregiver handoff system
\nAAC-adjacent communication map
\nsensory map
\nreview queue
\nAI summary assistant
\nsafety alert workflow
\nexportable care profile<\/p>\n

No humanoid body required.<\/p>\n

The body is the last-mile interface, not the foundation.<\/p>\n

Core Architecture<\/h2>\n

A v0.1 Amanah Companion system could have eight core layers:<\/p>\n

    \n
  1. Human Interface Layer<\/strong><\/li>\n
  2. Care Profile Layer<\/strong><\/li>\n
  3. Guardian Gate \/ Permissions Layer<\/strong><\/li>\n
  4. Care Memory Ledger<\/strong><\/li>\n
  5. AI Assistance Layer<\/strong><\/li>\n
  6. Integration Layer<\/strong><\/li>\n
  7. Audit \/ Source Trace Layer<\/strong><\/li>\n
  8. Safety and Privacy Layer<\/strong><\/li>\n<\/ol>\n

    Each layer has a different job.<\/p>\n

    The system becomes dangerous if these layers collapse into one vague \u201cmemory.\u201d<\/p>\n

    1. Human Interface Layer<\/h2>\n

    This is what people actually use.<\/p>\n

    Different users need different interfaces:<\/p>\n

    Parent \/ guardian dashboard
    \nCaregiver view
    \nTeacher \/ therapist contribution form
    \nEmergency summary view
    \nAI-assisted review queue
    \nCare profile editor
    \nCommunication Map editor
    \nSensory Map editor
    \nCare Memory Ledger
    \nExport panel
    \nPermission panel<\/p>\n

    The interface should not look like a chatbot first.<\/p>\n

    It should look like a care system.<\/p>\n

    The chatbot can exist, but it should not be the authority surface.<\/p>\n

    A caregiver should be able to see:<\/p>\n

    what is approved
    \nwhat is only a draft
    \nwhat needs review
    \nwhat is sensitive
    \nwhat has expired
    \nwho added it
    \nwhat the AI suggested
    \nwhat a human confirmed<\/p>\n

    Ethics must be visible in the interface.<\/p>\n

    2. Care Profile Layer<\/h2>\n

    The Care Profile is the central care object.<\/p>\n

    It stores the governed map of the person\u2019s support needs:<\/p>\n

    identity and dignity notes
    \nguardian authority
    \ncommunication map
    \nsensory map
    \nroutine and transition map
    \nregulation plan
    \nsafety map
    \ncare circle
    \nsetting differences
    \nreview status<\/p>\n

    This layer must be structured.<\/p>\n

    Not one giant note.<\/p>\n

    Not a pile of chat logs.<\/p>\n

    Not \u201cmemory\u201d in the vague companion-app sense.<\/p>\n

    A care profile needs fields, sources, review status, sensitivity levels, and expiry\/review dates.<\/p>\n

    3. Guardian Gate \/ Permissions Layer<\/h2>\n

    The Guardian Gate decides who can do what.<\/p>\n

    It should define:<\/p>\n

    who can view
    \nwho can edit
    \nwho can approve
    \nwho can export
    \nwho can invite others
    \nwho can connect AI tools
    \nwho can see sensitive records
    \nwho can approve a care-rule change
    \nwho can delete or archive entries<\/p>\n

    This is where the system preserves human authority.<\/p>\n

    UNICEF\u2019s child-centred AI guidance emphasizes safety, privacy, accountability, transparency, inclusion, child well-being, and child-centred governance for AI systems affecting children. It also flags AI companions used by children and accessibility for children with disabilities as emerging issues.<\/p>\n

    So the Guardian Gate should not be optional.<\/p>\n

    It is core infrastructure.<\/p>\n

    4. Care Memory Ledger<\/h2>\n

    The Care Memory Ledger stores care-relevant memory.<\/p>\n

    It does not store everything.<\/p>\n

    Memory entries should be typed:<\/p>\n

    raw event
    \ncaregiver note
    \nschool note
    \ntherapist note
    \nclinician note
    \nAI-generated draft
    \ncandidate pattern
    \napproved care rule
    \nsensitive incident
    \nhandoff summary<\/p>\n

    Each entry should include:<\/p>\n

    care purpose
    \nsource
    \ndate
    \nsetting
    \nreview status
    \nprivacy level
    \nretention status
    \npermissions
    \nwhether AI access is allowed
    \nwhether training use is forbidden by default<\/p>\n

    The default should be:<\/p>\n

    no model training from private care data.<\/strong><\/p>\n

    Care memory serves the person.<\/p>\n

    It does not feed the system.<\/p>\n

    5. AI Assistance Layer<\/h2>\n

    This is where the model enters.<\/p>\n

    But the model should not sit directly on raw private data with unlimited freedom.<\/p>\n

    The AI Assistance Layer should be narrow, gated, and task-specific.<\/p>\n

    Allowed v0.1 AI tasks:<\/p>\n

    summarize caregiver notes
    \ndraft handoff summaries
    \ndetect possible repeated patterns
    \nsuggest review questions
    \nflag missing source trace
    \nsuggest possible sensory trigger categories
    \norganize notes into Care Profile sections
    \ncompare current event with approved care rules
    \nprepare export summaries
    \ngenerate plain-language caregiver instructions from approved rules<\/p>\n

    Not allowed v0.1 AI tasks:<\/p>\n

    diagnose
    \nprescribe
    \nchange care plans automatically
    \ninterpret consent alone
    \nspeak as the child without confirmation
    \nshare records without approval
    \ntrain on care data by default
    \nmake emergency decisions independently
    \noverride guardians or clinicians
    \npromote its own guesses into truth<\/p>\n

    NIST\u2019s AI Risk Management Framework is intended to help developers, users, and evaluators manage risks to individuals, organizations, and society from AI systems, and it centers trustworthiness considerations such as safety, accountability, transparency, explainability, privacy, and fairness. (NIST<\/a>)<\/p>\n

    For Amanah Companions, this means the AI layer should be designed as a controlled assistant.<\/p>\n

    Not a free-roaming agent.<\/p>\n

    6. Integration Layer<\/h2>\n

    The prototype may eventually connect to other systems.<\/p>\n

    Possible integrations:<\/p>\n

    AAC apps or exports
    \ncalendar routines
    \nvisual schedule tools
    \nschool\/therapy note uploads
    \nsmart-home alerts
    \nwearables, if appropriate
    \nemergency contacts
    \ncloud storage export
    \nhealthcare records, only with strict boundaries
    \nFHIR-compatible health data, if clinical integration is ever pursued<\/p>\n

    FHIR is a healthcare data exchange standard published by HL7, designed to support electronic exchange of healthcare information using modular \u201cResources.\u201d It is widely used as an interoperability framework, but an Amanah Companion should only touch clinical data with proper consent, clear scope, and expert implementation. (HL7<\/a>)<\/p>\n

    For v0.1, I would not begin with deep clinical integration.<\/p>\n

    Start with manual import\/export and source trace.<\/p>\n

    Clinical interoperability can come later.<\/p>\n

    7. Audit \/ Source Trace Layer<\/h2>\n

    Every meaningful action should leave a trail.<\/p>\n

    The system should log:<\/p>\n

    who viewed a record
    \nwho edited a record
    \nwho approved a care rule
    \nwho rejected an AI suggestion
    \nwho exported data
    \nwhich AI tool accessed which records
    \nwhat was retrieved
    \nwhat was generated
    \nwhat was promoted
    \nwhat was archived
    \nwhat was deleted
    \nwhat permissions changed<\/p>\n

    Source trace should attach to every care entry:<\/p>\n

    parent observation
    \nguardian-approved rule
    \nteacher note
    \ntherapist note
    \nclinician note
    \nAI-generated draft
    \ndevice log
    \nmanual caregiver entry
    \nunknown \/ needs review<\/p>\n

    No care system should allow invisible memory mutation.<\/p>\n

    If the system changes what it \u201cknows\u201d about a vulnerable person, the care circle should be able to see how and why.<\/p>\n

    8. Safety and Privacy Layer<\/h2>\n

    This layer sets hard boundaries.<\/p>\n

    It should include:<\/p>\n

    encryption at rest and in transit
    \nrole-based access
    \nsensitive-record restrictions
    \nlocal-first or private-hosting options where possible
    \nexport controls
    \ndeletion \/ archive rules
    \nguardian approval for sharing
    \nno third-party training by default
    \nage-appropriate protections
    \nconsent logs
    \nemergency contact rules
    \nrestricted handling of media
    \nredaction tools
    \ndata minimization<\/p>\n

    The UN Convention on the Rights of Persons with Disabilities states that persons with disabilities have the right to protection from arbitrary or unlawful interference with privacy, family, home, correspondence, or communications, and that personal, health, and rehabilitation information must be protected. (World Health Organization<\/a>)<\/p>\n

    That means privacy is not a feature.<\/p>\n

    It is a right.<\/p>\n

    Suggested v0.1 System Diagram<\/h2>\n
    Amanah Companion v0.1\n\n[Parent \/ Guardian Dashboard]\n        |\n        v\n[Guardian Gate + Permissions]\n        |\n        v\n[Care Profile]\n   |       |       |\n   v       v       v\n[Communication Map] [Sensory Map] [Routine \/ Safety Maps]\n        |\n        v\n[Care Memory Ledger]\n        |\n        v\n[Review Queue]\n        |\n        v\n[AI Assistance Layer]\n   |       |       |\n   v       v       v\nSummaries  Candidate Patterns  Handoff Drafts\n        |\n        v\n[Human Review Required]\n        |\n        v\n[Approved Care Rules \/ Archived Notes]\n\nParallel layers:\n- Source Trace\n- Audit Log\n- Privacy Controls\n- Export \/ Backup\n- Emergency Summary\n<\/code><\/pre>\n
    The key is that the AI does not bypass the review queue.<\/p>\n
    It may assist.<\/p>\n
    It may not quietly govern.<\/p>\n
    Database Objects<\/h2>\n
    A simple v0.1 schema might include:<\/p>\n
    users\n- id\n- name\n- role\n- contact\n- authentication_status\n\ncare_subjects\n- id\n- name\n- date_of_birth\n- preferred_address\n- dignity_notes\n- guardian_id\n\ncare_profiles\n- id\n- care_subject_id\n- status\n- version\n- created_at\n- updated_at\n\nprofile_sections\n- id\n- care_profile_id\n- section_type\n- content\n- sensitivity_level\n- review_status\n- source_id\n\ncare_memory_entries\n- id\n- care_subject_id\n- entry_type\n- care_purpose\n- content\n- source_id\n- setting\n- privacy_level\n- review_status\n- retention_status\n- created_at\n\nguardian_permissions\n- id\n- user_id\n- care_subject_id\n- permission_type\n- granted_by\n- expires_at\n\nai_suggestions\n- id\n- care_subject_id\n- source_entries\n- suggestion_type\n- content\n- confidence_label\n- status\n- reviewed_by\n- reviewed_at\n\naudit_logs\n- id\n- actor_id\n- action_type\n- target_type\n- target_id\n- timestamp\n- details\n\nsources\n- id\n- source_type\n- source_person\n- role\n- date_observed\n- setting\n- confidence\n<\/code><\/pre>\n
    This is not final.<\/p>\n
    But it shows the philosophy:<\/p>\n
    memory is typed
    \nauthority is explicit
    \nAI suggestions are separate
    \nreview status matters
    \nsource trace is structural
    \naudit logs are mandatory<\/p>\n
    AI Workflow Example<\/h2>\n
    A caregiver enters:<\/p>\n
    \u201cBath was hard today. He cried when tablet was taken.\u201d<\/p>\n
    The system stores it as a raw caregiver note.<\/p>\n
    The AI may generate:<\/p>\n
    \u201cPossible transition issue: tablet removal before bath. Check whether warning was given, visual timer used, water temperature, soap smell, and fatigue.\u201d<\/p>\n
    The system marks this as:<\/p>\n
    AI-generated draft
    \nunreviewed
    \ncandidate pattern
    \nnot active care rule<\/p>\n
    If similar notes appear several times, the AI may suggest:<\/p>\n
    \u201cBath transition may be difficult when tablet use ends abruptly. Guardian review recommended.\u201d<\/p>\n
    The guardian reviews and approves:<\/p>\n
    \u201cUse five-minute visual timer before bath. Do not remove tablet suddenly. Offer first\/then script.\u201d<\/p>\n
    Only then does it become an approved care rule.<\/p>\n
    That is the Guardian Gate working.<\/p>\n
    Human Roles<\/h2>\n
    The prototype should define human roles clearly.<\/p>\n
    Parent \/ Legal Guardian<\/h3>\n
    Can approve rules, manage permissions, export records, review AI suggestions, and control sensitive data.<\/p>\n
    Approved Caregiver<\/h3>\n
    Can add notes, view approved care instructions, receive handoff summaries, and flag concerns.<\/p>\n
    Therapist \/ Clinician<\/h3>\n
    Can contribute professional notes within scope, review relevant patterns, and suggest care plan updates.<\/p>\n
    Teacher \/ School Support<\/h3>\n
    Can add setting-specific notes, view approved school-relevant instructions, and contribute observations.<\/p>\n
    AI Assistant<\/h3>\n
    Can summarize, organize, suggest, flag, and draft.<\/p>\n
    Cannot approve, diagnose, prescribe, override, or silently promote memory.<\/p>\n
    Model Design<\/h2>\n
    The AI layer should not rely on one huge prompt.<\/p>\n
    It should use task-specific functions.<\/p>\n
    Examples:<\/p>\n
    SummarizeHandoff
    \nDetectCandidatePattern
    \nClassifyCareMemory
    \nCheckSourceTrace
    \nDraftGuardianReviewQuestion
    \nCompareWithApprovedRoutine
    \nSuggestSensoryCategory
    \nRedactSensitiveExport
    \nGenerateEmergencySummary<\/p>\n
    Each function should have:<\/p>\n
    input scope
    \nallowed sources
    \nforbidden outputs
    \nrequired uncertainty labels
    \nreview requirements
    \naudit logging<\/p>\n
    This is where the Ahd Nucleus idea of Stem Cells<\/strong> becomes practical.<\/p>\n
    Stem Cells are not magic.<\/p>\n
    They are small governed behaviors.<\/p>\n
    For Amanah Companions, each Stem Cell should do one narrow care-support task and know its boundary.<\/p>\n
    Safety Rules for AI Output<\/h2>\n
    Every AI output should be labeled.<\/p>\n
    Possible labels:<\/p>\n
    Draft
    \nUnreviewed
    \nPossible pattern
    \nNeeds guardian review
    \nClinician review recommended
    \nApproved care rule
    \nSuperseded
    \nSensitive
    \nEmergency only<\/p>\n
    The AI should use careful language:<\/p>\n
    \u201cMay indicate\u2026\u201d
    \n\u201cPossible pattern\u2026\u201d
    \n\u201cCheck with guardian\u2026\u201d
    \n\u201cDo not assume\u2026\u201d
    \n\u201cNeeds human review\u2026\u201d
    \n\u201cOutside AI authority\u2026\u201d<\/p>\n
    It should avoid overconfident language:<\/p>\n
    \u201cHe always\u2026\u201d
    \n\u201cHe is manipulating\u2026\u201d
    \n\u201cHe definitely wants\u2026\u201d
    \n\u201cThis proves\u2026\u201d
    \n\u201cChange the plan\u2026\u201d<\/p>\n
    In care, phrasing is safety.<\/p>\n
    Embodied Layer Later<\/h2>\n
    Only after the software architecture is stable should embodiment be tested.<\/p>\n
    A future embodied Amanah Companion might be allowed to:<\/p>\n
    display visual schedule
    \nplay approved calming audio
    \nbring AAC device closer
    \nalert caregiver
    \nmove away during overload
    \nsupport transition cue
    \ncarry safe object
    \nadjust lights through approved smart-home controls<\/p>\n
    But it should not be allowed to:<\/p>\n
    restrain
    \ndiscipline
    \nblock movement except under carefully governed emergency designs
    \nremove communication tools
    \nphysically force transitions
    \nrecord private spaces by default
    \ninterpret consent alone
    \nact as sole supervisor<\/p>\n
    The embodied layer should inherit the Guardian Gate.<\/p>\n
    Not bypass it.<\/p>\n
    Minimum Viable Prototype<\/h2>\n
    A realistic MVP would not include robotics.<\/p>\n
    It would include:<\/p>\n
    Care Profile editor
    \nGuardian Gate permissions
    \nCommunication Map
    \nSensory Map
    \nRoutine Map
    \nSafety Map
    \nCare Memory Ledger
    \nReview Queue
    \nAI summary drafts
    \nAI candidate-pattern detection
    \nSource Trace
    \nAudit Log
    \nExport function
    \nPrivacy settings
    \nNo training by default<\/p>\n
    That is enough to test the core question:<\/p>\n
    Can continuity governance improve care documentation and handoff without becoming surveillance?<\/p>\n
    Pilot Study Shape<\/h2>\n
    A careful pilot could involve:<\/p>\n
    small number of families
    \nopt-in participation
    \nno private model training
    \nlocal or private deployment
    \nmanual entry first
    \nguardian-controlled data
    \ntherapist\/clinician advisory input
    \nmeasured caregiver burden
    \nmeasured handoff clarity
    \nmeasured usefulness of AI summaries
    \ntracking false or harmful suggestions
    \nreviewing privacy comfort
    \ntesting exportability
    \ntesting whether the system reduces or increases stress<\/p>\n
    Success should not be measured by engagement.<\/p>\n
    It should be measured by care usefulness and trust.<\/p>\n
    Possible pilot questions:<\/p>\n
    Did caregivers find handoffs clearer?
    \nDid the system reduce repeated explanation burden?
    \nWere AI suggestions useful or noisy?
    \nDid the review queue feel manageable?
    \nDid families feel in control of data?
    \nWere privacy boundaries understandable?
    \nDid the system avoid overclaiming?
    \nDid it help preserve meaningful care patterns?<\/p>\n
    What Not to Build First<\/h2>\n
    Do not build the humanoid first.<\/p>\n
    Do not build a child-facing AI friend first.<\/p>\n
    Do not build emotional attachment loops first.<\/p>\n
    Do not build always-on home recording first.<\/p>\n
    Do not build automatic behavioral scoring first.<\/p>\n
    Do not build \u201cAI knows what your child wants\u201d first.<\/p>\n
    Do not build training-data pipelines first.<\/p>\n
    Build the care spine first.<\/p>\n
    Everything else must answer to it.<\/p>\n
    Where Ahd Nucleus Fits<\/h2>\n
    Ahd Nucleus gives the parent architecture.<\/p>\n
    It already asks:<\/p>\n
    What is the source of truth?
    \nWho has authority?
    \nWhat is draft?
    \nWhat is canon?
    \nWhat is sensitive?
    \nWhat should be retrieved?
    \nWhat should be reviewed?
    \nWhat should be logged?
    \nWhat should never be flattened?<\/p>\n
    Amanah Companions applies that same continuity governance to care.<\/p>\n
    The difference is the stakes.<\/p>\n
    In creative work, bad memory can damage a project.<\/p>\n
    In care, bad memory can misread a person.<\/p>\n
    That is why the architecture must be stricter.<\/p>\n
    Closing<\/h2>\n
    The Amanah Companion prototype should not begin with a face.<\/p>\n
    It should begin with a governed care system.<\/p>\n
    A Care Profile.
    \nA Guardian Gate.
    \nA Communication Map.
    \nA Sensory Map.
    \nA Care Memory Ledger.
    \nSource trace.
    \nAudit logs.
    \nHuman review.
    \nPrivacy by design.
    \nAI suggestions that know they are only suggestions.<\/p>\n
    If that foundation cannot hold, the system should not be embodied.<\/p>\n
    A body without governance is not care.<\/p>\n
    It is theatre with sensors.<\/p>\n
    But a governed continuity system \u2014 one that helps families, caregivers, therapists, and teachers preserve what matters without surrendering dignity or privacy \u2014 could become something genuinely useful.<\/p>\n
    Not because it imitates a person.<\/p>\n
    Because it helps people care better.<\/p>\n
    That is the prototype worth building.<\/p>\n<\/div><\/div><\/div><\/div><\/div>\n","protected":false},"excerpt":{"rendered":"","protected":false},"author":1,"featured_media":41017,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"advanced_seo_description":"","jetpack_seo_html_title":"","jetpack_seo_noindex":false,"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_feature_clip_id":0,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_publicize_message":"","jetpack_publicize_feature_enabled":true,"jetpack_social_post_already_shared":true,"jetpack_social_options":{"image_generator_settings":{"template":"highway","default_image_id":0,"font":"","enabled":false},"version":2},"jetpack_post_was_ever_published":false},"categories":[2],"tags":[7,22],"class_list":["post-41521","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-articles","tag-amanah-companions","tag-research"],"jetpack_publicize_connections":[],"jetpack_featured_media_url":"","jetpack_sharing_enabled":true,"jetpack_shortlink":"https:\/\/wp.me\/phjhRs-aNH","_links":{"self":[{"href":"https:\/\/mithaqpraxis.com\/wp-json\/wp\/v2\/posts\/41521","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/mithaqpraxis.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/mithaqpraxis.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/mithaqpraxis.com\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/mithaqpraxis.com\/wp-json\/wp\/v2\/comments?post=41521"}],"version-history":[{"count":0,"href":"https:\/\/mithaqpraxis.com\/wp-json\/wp\/v2\/posts\/41521\/revisions"}],"wp:attachment":[{"href":"https:\/\/mithaqpraxis.com\/wp-json\/wp\/v2\/media?parent=41521"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/mithaqpraxis.com\/wp-json\/wp\/v2\/categories?post=41521"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/mithaqpraxis.com\/wp-json\/wp\/v2\/tags?post=41521"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}