Skip to main content

Knowledge > Processes > Theological Lens Matching

Theological Lens Matching

How the system adapts to each church's theological tradition across both the voice agent and chatbot. The lens controls vocabulary, doctrinal rules, RAG filtering, and agent personality.

Theological Lens Taxonomy​

17 theological lenses (IDs 1-17) plus 1 universal (ID 0), stored in sai_theological_lenses.

ID  Lens Name
──  ──────────────────────
 0  Universal (shows in all traditions)
 1  Evangelical / Restoration
 2  Progressive
 3  Missional
 4  Reformed / Presbyterian
 5  Wesleyan / Methodist
 6  Lutheran
 7  Catholic
 8  Anabaptist
 9  Pentecostal
10  Christocentric (DEFAULT — non-denominational)
11  Orthodox
12  Black Church
13  Anglican / Episcopal
14  Baptist
15  Charismatic
16  Dispensational
17  Liberation

Lens Resolution (Priority Order)​

The system resolves a church's theological lens through a cascade of sources. The first match wins.

Chatbot Resolution (route.ts)​

PRIORITY 1: CLIENT-SIDE OVERRIDE (demo mode)
  IF lensOverride is a valid number (1-17)
    AND lensNameOverride is a non-empty string (<=50 chars):
    lensId = lensOverride
    lensName = sanitized lensNameOverride
    # Used in demo/preview when user selects a tradition

PRIORITY 2: EXPLICIT CHURCH SETTING (church_theological_lenses)
  query church_theological_lenses
    WHERE church_id = church.id
    JOIN sai_theological_lenses for lens_name
    LIMIT 1
  IF found:
    lensId = theological_lens_id
    lensName = lens_name

PRIORITY 3: DENOMINATION AUTO-DETECT (DENOMINATION_TO_LENS mapping)
  IF church.denomination is not null:
    lensId = DENOMINATION_TO_LENS[church.denomination] ?? 10
    query sai_theological_lenses for lens_name WHERE lens_id = lensId

PRIORITY 4: DEFAULT
  lensId = 10 (Christocentric)
  lensName = "Christocentric"

Voice Agent Resolution (core/rag.py)​

FUNCTION get_lens_id(denomination: str | None) -> int:
  IF denomination is None or empty:
    RETURN 10 (Christocentric — DEFAULT_LENS_ID)
  RETURN DENOMINATION_TO_LENS.get(denomination, 10)

Denomination-to-Lens Mapping​

70+ denomination strings mapped to 17 lens IDs. This mapping is maintained in sync across three files:

  • churchwiseai-web/voice-agent-livekit/core/rag.py (DENOMINATION_TO_LENS)
  • churchwiseai-web/src/lib/rag.ts (DENOMINATION_TO_LENS)
  • Legacy: voice-agent/rag.js (reference only)
BAPTIST FAMILY → Lens 14 (Baptist)
  Baptist, Southern Baptist, American Baptist, Free Will Baptist

CATHOLIC → Lens 7
  Catholic, Roman Catholic

WESLEYAN/METHODIST FAMILY → Lens 5
  Methodist, United Methodist, Free Methodist, AME
  Church of the Nazarene, Nazarene, Wesleyan

REFORMED/PRESBYTERIAN → Lens 4
  Presbyterian, PCA, PCUSA, Reformed

LUTHERAN → Lens 6
  Lutheran, ELCA, LCMS, Missouri Synod

PENTECOSTAL → Lens 9
  Pentecostal, Assembly of God, Assemblies of God
  Church of God, Foursquare

ANGLICAN/EPISCOPAL → Lens 13
  Episcopal, Anglican

ORTHODOX → Lens 11
  Orthodox, Greek Orthodox, Eastern Orthodox, Russian Orthodox

ANABAPTIST → Lens 8
  Mennonite, Amish, Anabaptist, Brethren

NON-DENOMINATIONAL → Lens 10 (Christocentric, DEFAULT)
  Non-denominational, Nondenominational, Community Church, Bible Church

CHARISMATIC → Lens 15
  Charismatic

EVANGELICAL / RESTORATION → Lens 1
  Evangelical, Church of Christ, Disciples of Christ, Christian Church

BLACK CHURCH → Lens 12
  AME Zion, CME, COGIC, National Baptist

DISPENSATIONAL → Lens 16
  Dispensational

PROGRESSIVE → Lens 2
  Progressive, UCC, United Church of Christ

MISSIONAL → Lens 3
  Missional

LIBERATION → Lens 17
  Liberation

What the Lens Controls​

1. RAG Filtering​

Both voice agent and chatbot filter search results by lens ID.

VOICE AGENT (fetch_session_rag in core/rag.py):
  lens_id = get_lens_id(denomination)
  search_unified_rag(embedding, lens_ids=[lens_id], match_count=5)
  # Returns content tagged with this lens OR universal (lens_id=0)

CHATBOT (route.ts):
  searchRAG({ queryEmbedding: embedding, lensIds: [lensId], matchCount: 8 })
  # Supabase RPC search_unified_rag_content filters by p_theological_lens_ids

RESULT:
  A Baptist church gets Baptist-tradition illustrations
  A Catholic church gets Catholic-tradition content
  Universal content (lens_id=0) appears for all traditions

2. Doctrinal Rules (theological_contradictions)​

Per-lens rules that constrain what the AI must/must-not say.

QUERY: theological_contradictions WHERE lens_id = lensId

FIELDS PER RULE:
  doctrine_category: e.g., "baptism", "eucharist", "ordination"
  primary_position: "Believers' baptism by immersion only"
  contrary_positions: ["infant baptism", "sprinkling"]
  must_include_terms: ["credobaptism", "immersion"]
  must_exclude_terms: ["christening", "infant sprinkling"]
  explanation: context for the AI

INJECTED AS:
  "IMPORTANT DOCTRINAL REQUIREMENTS:
   BAPTISM:
   - This church's position: Believers' baptism by immersion only
   - MUST include: credobaptism, immersion
   - MUST NEVER mention: christening, infant sprinkling
   - DO NOT recommend: infant baptism, sprinkling
   This means your answers should strictly follow the Baptist
   theological positions listed above."

CHURCH-SPECIFIC OVERRIDES (organization_settings.doctrinal_overrides):
  A church can override standard lens positions:
    { "baptism": { custom_practice: true,
        explanation: "We practice both infant dedication and believer's baptism" }}
  Appended as "CUSTOM CHURCH PRACTICES" section

3. Vocabulary (lens_knowledge → lens-vocabulary.ts)​

Per-lens preferred and avoided vocabulary injected into the chatbot prompt.

EXAMPLES:
  Baptist (lens 14):
    PREFER: "baptism by immersion", "believer's baptism", "Sunday school"
    AVOID: "christening", "infant baptism", "catechism"

  Catholic (lens 7):
    PREFER: "Holy Mass", "Eucharist", "parish", "sacrament", "priest"
    AVOID: "worship service", "communion table", "pastor" (use "Father")

  Orthodox (lens 11):
    PREFER: "Divine Liturgy", "Theotokos", "Holy Tradition"
    AVOID: "worship service", "Bible study" (use "catechesis")

  Lutheran (lens 6):
    PREFER: "the Lord's Supper", "Real Presence", "sacrament"
    AVOID: "mere symbol", "just symbolic"

IMPLEMENTATION:
  fetchLensVocabulary(lensId) → queries lens_knowledge table
  buildChatVocabularyBlock(vocab, lensName) → formats for prompt injection

4. Agent Personality​

The theological lens influences the chatbot's conversational style.

FORMAL TRADITIONS (Catholic, Orthodox, Anglican, Lutheran):
  More formal language
  Liturgical calendar awareness
  Proper titles (Father, Reverend, Bishop)
  Reference to sacraments, tradition, church fathers

INFORMAL TRADITIONS (Non-denominational, Community, Charismatic):
  Conversational tone
  Contemporary language
  First-name basis
  Focus on personal relationship with God

REFORMED TRADITIONS (Presbyterian, Reformed, Baptist):
  Theological precision
  Scripture-first approach
  Doctrinal clarity
  Respect for confessional standards

BLACK CHURCH TRADITION:
  Rich call-and-response patterns
  Social justice awareness
  Community emphasis
  Musical/worship tradition references

Cross-Channel Consistency​

The lens resolution produces identical results across both channels because:

  1. Both use the same DENOMINATION_TO_LENS mapping (kept in sync manually)
  2. Both search the same unified_rag_content table with the same lens filter
  3. Both read the same theological_contradictions rules
  4. Default is always Christocentric (lens 10) when denomination is unknown

The voice agent resolves lens once at session initialization (fetch_session_rag) and caches it for the call. The chatbot resolves lens on every request (stateless API).