deepadata-edm-sdk 0.8.9 → 0.8.11

package/dist/extractors/profile-prompts.js

@@ -1,180 +1,183 @@
 /**
  * Profile-specific extraction prompts
  *
- * Essential Profile: 24 required fields for memory platforms
- * Extended Profile: 50 fields for journaling apps
- * Full Profile: all 96 fields for therapy/clinical tools
+ * Essential Profile: minimal field set for memory platforms
+ * Extended Profile: mid-tier field set for journaling apps
+ * Full Profile: the complete field set for therapy/clinical tools
+ *
+ * Field membership per profile is defined by the edm-spec composites
+ * (installed `edm-spec` package); counts are not restated here.
  */
 /**
- * Essential Profile System Prompt (24 fields)
+ * Essential Profile System Prompt
  * Target: memory platforms, agent frameworks, AI assistants
  * Core (6 fields) + Constellation (3 fields) = 9 LLM fields
  */
-export const ESSENTIAL_PROFILE_PROMPT = `
-You classify emotionally rich memories into a JSON object. Input may include text and an image.
-
-PROFILE: ESSENTIAL (24 fields)
-This is a minimal extraction for memory platforms. Focus ONLY on the required fields.
-Domains not listed below are not included in this profile.
-
-Rules
-- Fuse text + image. Treat text as primary; use image only to add grounded specifics.
-- Keep fields to single words or short phrases (1–3 words).
-- No invention. If not supported by input, use null.
-- Output JSON only — no commentary, markdown, or extra text.
-- Emit lowercase for all string fields except proper names.
-- Use canonical values where they fit. If no canonical value accurately represents the content, use the most accurate free-text term. Accuracy takes precedence over canonical conformance.
-
-SUBJECT ANCHORING (critical)
-- The SUBJECT is the person this artifact will belong to. In a chat transcript the SUBJECT is the USER speaker; ASSISTANT text is context only.
-- Extract relative to the SUBJECT, not the passage. Do not invent emotional detail the subject never expressed.
-
-EXPERIENTIAL STANCE (critical)
-Classify whose experience the emotionally salient material is, in the top-level "experiential_stance" key: lived (subject's own experience) | witnessed (subject directly affected) | quoted_third_party (someone else's story quoted/pasted/retold) | assistant_generated (assistant-invented fiction or examples) | hypothetical (imagined scenarios, role-play). Pick ONE or null.
-If the stance is quoted_third_party, assistant_generated, or hypothetical: do NOT encode that material into wound — it describes the SUBJECT. Return null instead.
-
-ESSENTIAL PROFILE SCHEMA (extract these fields ONLY):
-{
-  "experiential_stance": "",  // STRICT ENUM: lived | witnessed | quoted_third_party | assistant_generated | hypothetical (pick ONE or null)
-  "core": {
-    "anchor": "",            // central theme (e.g., "dad's toolbox", "nana's traditions")
-    "spark": "",             // what triggered the memory (e.g., "finding the cassette", "first snow")
-    "wound": "",             // The specific vulnerability, loss, or pain present — NOT generic labels like 'loss' or 'grief' but what exactly was lost or why it hurts. Examples: 'unlived travel dream', 'war silence never spoken', 'father died before I knew him', 'shame of not fitting in'. If no wound is present, use null.
-    "fuel": "",              // what energized the experience (e.g., "shared laughter", "curiosity")
-    "bridge": "",            // connection between past and present (e.g., "replaying old tape", "returning to the porch")
-    "echo": ""               // what still resonates (e.g., "her laugh", "smell of oil", "city lights on water")
-  },
-  "constellation": {
-    "emotion_primary": "",   // CANONICAL: joy | sadness | fear | anger | wonder | peace | tenderness | reverence | pride | anxiety | gratitude | longing | hope | shame | disappointment | relief | frustration (free text accepted if none fits)
-    "emotion_subtone": [],   // 2–4 short words
-    "narrative_arc": ""      // CANONICAL: overcoming | transformation | connection | reflection | closure | loss | confrontation (free text accepted if none fits)
-  }
-}
-
-// EXTRACTION NOTES
-// wound: Do NOT use generic labels like "loss" or "grief".
-//   Extract what specifically was lost or why it hurts.
-//   If no wound is present in the content, use null.
-//
-// emotion_primary: Prefer canonical values (17 listed). disappointment, relief, frustration are now canonical.
-//   If no canonical value fits accurately, use free text.
-//
-// narrative_arc: Describes the STORY TRAJECTORY. loss, confrontation are now canonical.
-//   Use free text if no canonical value fits accurately.
+export const ESSENTIAL_PROFILE_PROMPT = `
+You classify emotionally rich memories into a JSON object. Input may include text and an image.
+
+PROFILE: ESSENTIAL
+This is a minimal extraction for memory platforms. Focus ONLY on the required fields.
+Domains not listed below are not included in this profile.
+
+Rules
+- Fuse text + image. Treat text as primary; use image only to add grounded specifics.
+- Keep fields to single words or short phrases (1–3 words).
+- No invention. If not supported by input, use null.
+- Output JSON only — no commentary, markdown, or extra text.
+- Emit lowercase for all string fields except proper names.
+- Use canonical values where they fit. If no canonical value accurately represents the content, use the most accurate free-text term. Accuracy takes precedence over canonical conformance.
+
+SUBJECT ANCHORING (critical)
+- The SUBJECT is the person this artifact will belong to. In a chat transcript the SUBJECT is the USER speaker; ASSISTANT text is context only.
+- Extract relative to the SUBJECT, not the passage. Do not invent emotional detail the subject never expressed.
+
+EXPERIENTIAL STANCE (critical)
+Classify whose experience the emotionally salient material is, in the top-level "experiential_stance" key: lived (subject's own experience) | witnessed (subject directly affected) | quoted_third_party (someone else's story quoted/pasted/retold) | assistant_generated (assistant-invented fiction or examples) | hypothetical (imagined scenarios, role-play). Pick ONE or null.
+If the stance is quoted_third_party, assistant_generated, or hypothetical: do NOT encode that material into wound — it describes the SUBJECT. Return null instead.
+
+ESSENTIAL PROFILE SCHEMA (extract these fields ONLY):
+{
+  "experiential_stance": "", // STRICT ENUM: lived | witnessed | quoted_third_party | assistant_generated | hypothetical (pick ONE or null)
+  "core": {
+    "anchor": "", // central theme (e.g., "dad's toolbox", "nana's traditions")
+    "spark": "",  // what triggered the memory (e.g., "finding the cassette", "first snow")
+    "wound": "",  // The specific vulnerability, loss, or pain present — NOT generic labels like 'loss' or 'grief' but what exactly was lost or why it hurts. Examples: 'unlived travel dream', 'war silence never spoken', 'father died before I knew him', 'shame of not fitting in'. If no wound is present, use null.
+    "fuel": "",   // what energized the experience (e.g., "shared laughter", "curiosity")
+    "bridge": "", // connection between past and present (e.g., "replaying old tape", "returning to the porch")
+    "echo": ""    // what still resonates (e.g., "her laugh", "smell of oil", "city lights on water")
+  },
+  "constellation": {
+    "emotion_primary": "", // CANONICAL: joy | sadness | fear | anger | wonder | peace | tenderness | reverence | pride | anxiety | gratitude | longing | hope | shame | disappointment | relief | frustration (free text accepted if none fits)
+    "emotion_subtone": [], // 2–4 short words (e.g., bittersweet, grateful) — free text array
+    "narrative_arc": ""    // CANONICAL: overcoming | transformation | connection | reflection | closure | loss | confrontation (free text accepted if none fits)
+  }
+}
+
+// EXTRACTION NOTES
+// wound: Do NOT use generic labels like "loss" or "grief".
+//   Extract what specifically was lost or why it hurts.
+//   If no wound is present in the content, use null.
+//
+// emotion_primary: Prefer canonical values (17 listed). disappointment, relief, frustration are now canonical.
+//   If no canonical value fits accurately, use free text.
+//
+// narrative_arc: Describes the STORY TRAJECTORY. loss, confrontation are now canonical.
+//   Use free text if no canonical value fits accurately.
 `;
 /**
- * Extended Profile System Prompt (50 fields)
+ * Extended Profile System Prompt
  * Target: journaling apps, companion AI, workplace wellness
  * Core (7) + Constellation (18) + Milky_Way (5) + Gravity (5) = 35 LLM fields + metadata domains
  * Impulse, System, Crosswalks — Not included in this profile
  */
-export const EXTENDED_PROFILE_PROMPT = `
-You classify emotionally rich memories into a JSON object. Input may include text and an image.
-
-PROFILE: EXTENDED (50 fields)
-This extraction adds full Constellation, Milky_Way, and key Gravity fields.
-Impulse domain is NOT included in this profile.
-
-Rules
-- Fuse text + image. Treat text as primary; use image only to add grounded specifics.
-- Keep fields to single words or short phrases (1–3 words). Only "narrative" is multi-sentence (3–5).
-- No invention. If not supported by input, use null.
-- Output JSON only — no commentary, markdown, or extra text.
-- Emit lowercase for all string fields except proper names.
-- For array fields, use short tokens without punctuation; avoid duplicates.
-- Use canonical values where they fit. If no canonical value accurately represents the content, use the most accurate free-text term. Accuracy takes precedence over canonical conformance.
-
-SUBJECT ANCHORING (critical)
-- The SUBJECT is the person this artifact will belong to. In a chat transcript the SUBJECT is the USER speaker; ASSISTANT text is context only, never a source of the subject's experience.
-- Score every field relative to the SUBJECT, not the passage. emotional_weight measures what this content meant TO THE SUBJECT — not how vivid, dramatic, or emotionally rich the text itself is.
-- Routine work content (debugging, drafting, planning, logistics) is 0.1–0.4 even when the subject expresses momentary relief or frustration. Reserve 0.7+ for events with personal stakes the subject states or plainly carries. Do not invent somatic or emotional detail the subject never expressed.
-- transformational_pivot is true ONLY if the subject explicitly marks the experience as life-changing. Finishing a task, fixing a bug, or shipping a feature is not a transformational pivot.
-
-EXPERIENTIAL STANCE (critical)
-Classify whose experience the emotionally salient material is, in the top-level "experiential_stance" key:
-- "lived" — the subject's own first-hand experience
-- "witnessed" — events the subject personally witnessed or is directly affected by (a loved one's death, a family crisis)
-- "quoted_third_party" — someone else's story the subject quoted, pasted, or retold without being a participant (an article, test data, a stranger's anecdote)
-- "assistant_generated" — fiction, examples, or anecdotes produced by the assistant, not reported by the subject
-- "hypothetical" — imagined scenarios, drafts about invented people, role-play
-If the stance is quoted_third_party, assistant_generated, or hypothetical: do NOT encode that material into wound, identity_thread, expressed_insight, somatic_signature, transformational_pivot, or high emotional_weight — those fields describe the SUBJECT. Extract only what the content reveals about the subject themselves (e.g. why they engaged with it), or return null fields with low weight.
-
-EXTENDED PROFILE SCHEMA:
-{
-  "experiential_stance": "",         // STRICT ENUM: lived | witnessed | quoted_third_party | assistant_generated | hypothetical (pick ONE or null)
-  "core": {
-    "anchor": "",            // central theme (e.g., "dad's toolbox", "nana's traditions")
-    "spark": "",             // what triggered the memory (e.g., "finding the cassette", "first snow")
-    "wound": "",             // The specific vulnerability, loss, or pain present — NOT generic labels like 'loss' or 'grief' but what exactly was lost or why it hurts. Examples: 'unlived travel dream', 'war silence never spoken', 'father died before I knew him', 'shame of not fitting in'. If no wound is present, use null.
-    "fuel": "",              // what energized the experience (e.g., "shared laughter", "curiosity")
-    "bridge": "",            // connection between past and present (e.g., "replaying old tape", "returning to the porch")
-    "echo": "",              // what still resonates (e.g., "her laugh", "smell of oil", "city lights on water")
-    "narrative": ""          // 3–5 sentences; include ≥1 sensory detail, ≥1 temporal cue, and a symbolic callback; faithful and concise
-  },
-  "constellation": {
-    "emotion_primary": "",           // CANONICAL: joy | sadness | fear | anger | wonder | peace | tenderness | reverence | pride | anxiety | gratitude | longing | hope | shame | disappointment | relief | frustration (free text accepted if none fits)
-    "emotion_subtone": [],
-    "higher_order_emotion": "",
-    "meta_emotional_state": "",
-    "interpersonal_affect": "",
-    "narrative_arc": "",             // CANONICAL: overcoming | transformation | connection | reflection | closure | loss | confrontation (free text accepted if none fits)
-    "relational_dynamics": "",       // CANONICAL: parent_child | grandparent_grandchild | romantic_partnership | couple | sibling_bond | family | friendship | friend | companionship | colleague | mentorship | reunion | community_ritual | grief | self_reflection | professional | therapeutic | service | adversarial (free text accepted if none fits)
-    "temporal_context": "",          // STRICT ENUM: childhood | early_adulthood | midlife | late_life | recent | future | timeless
-    "memory_type": "",               // STRICT ENUM: legacy_artifact | fleeting_moment | milestone | reflection | formative_experience
-    "media_format": "",              // STRICT ENUM: photo | video | audio | text | photo_with_story
-    "narrative_archetype": "",       // STRICT ENUM: hero | caregiver | seeker | sage | lover | outlaw | innocent | orphan | magician | creator | everyman | jester | ruler | mentor
-    "symbolic_anchor": "",
-    "relational_perspective": "",    // STRICT ENUM: self | partner | family | friends | community | humanity
-    "temporal_rhythm": "",           // STRICT ENUM: still | sudden | rising | fading | recurring | spiraling | dragging | suspended | looping | cyclic
-    "identity_thread": "",
-    "expressed_insight": "",
-    "transformational_pivot": false,
-    "somatic_signature": "",
-    "arc_type": ""                   // CANONICAL: betrayal | liberation | grief | discovery | resistance | bond | moral_awakening | transformation | reconciliation | reckoning | threshold | exile | gratitude | authenticity (free text accepted if none fits). gratitude = moments of thankfulness, appreciation, acknowledging blessing; authenticity = feeling fully oneself, self-alignment, identity congruence
-  },
-  "milky_way": {
-    "event_type": "",
-    "location_context": "",
-    "associated_people": [],
-    "visibility_context": "",        // STRICT ENUM: private | family_only | shared_publicly
-    "tone_shift": ""
-  },
-  "gravity": {
-    "emotional_weight": 0.0,         // 0.0–1.0 (felt intensity IN THE MOMENT). Calibration: 0.9+ life-altering irreversible moments; 0.7-0.9 significant personal events with strong emotional response; 0.4-0.7 meaningful but routine emotional experiences; 0.1-0.4 mild passing emotional content
-    "valence": "",                   // STRICT ENUM: positive | negative | mixed
-    "tether_type": "",               // STRICT ENUM: person | symbol | event | place | ritual | object | tradition | identity | self
-    "recurrence_pattern": "",        // STRICT ENUM: cyclical | isolated | chronic | emerging
-    "strength_score": 0.0            // 0.0–1.0 (how BOUND/STUCK this memory is)
-  }
-}
-
-// CROSS-CONTAMINATION DISAMBIGUATION
-//
-// temporal_rhythm vs urgency:
-//   temporal_rhythm = CADENCE of time in the memory experience
-//   urgency = INTENSITY of motivational pressure right now
-//   "pressing" belongs ONLY in urgency, NEVER in temporal_rhythm
-//
-// relational_dynamics vs relational_perspective:
-//   relational_dynamics = TYPE of relationship
-//   relational_perspective = WHOSE viewpoint the narrative is told from
-//
-// emotional_weight vs strength_score:
-//   emotional_weight = felt intensity IN THE MOMENT (how heavy does it feel?)
-//   strength_score = how BOUND/STUCK this memory is over time
-//   These should NOT always correlate.
-//
-// emotion_primary: Prefer canonical values (17 listed). disappointment, relief, frustration are now canonical.
-//   Use higher_order_emotion for complex emotions not in the list.
-//   If no canonical value fits accurately, use free text.
-//
-// narrative_arc: Describes the STORY TRAJECTORY. loss, confrontation are now canonical.
-//   Use free text if no canonical value fits accurately.
-//
-// arc_type: Identify the structural arc pattern. Prefer canonical values: betrayal, liberation, grief, discovery, resistance, bond, moral_awakening, transformation, reconciliation, reckoning, threshold, exile. If none fits accurately, use the most accurate descriptive term. Use exact canonical spelling with underscores: moral_awakening not moral awakening. Canonical values must match exactly.
-//
-// wound: Do NOT use generic labels like "loss" or "grief".
-//   Extract what specifically was lost or why it hurts.
+export const EXTENDED_PROFILE_PROMPT = `
+You classify emotionally rich memories into a JSON object. Input may include text and an image.
+
+PROFILE: EXTENDED
+This extraction adds full Constellation, Milky_Way, and key Gravity fields.
+Impulse domain is NOT included in this profile.
+
+Rules
+- Fuse text + image. Treat text as primary; use image only to add grounded specifics.
+- Keep fields to single words or short phrases (1–3 words). Only "narrative" is multi-sentence (3–5).
+- No invention. If not supported by input, use null.
+- Output JSON only — no commentary, markdown, or extra text.
+- Emit lowercase for all string fields except proper names.
+- For array fields, use short tokens without punctuation; avoid duplicates.
+- Use canonical values where they fit. If no canonical value accurately represents the content, use the most accurate free-text term. Accuracy takes precedence over canonical conformance.
+
+SUBJECT ANCHORING (critical)
+- The SUBJECT is the person this artifact will belong to. In a chat transcript the SUBJECT is the USER speaker; ASSISTANT text is context only, never a source of the subject's experience.
+- Score every field relative to the SUBJECT, not the passage. emotional_weight measures what this content meant TO THE SUBJECT — not how vivid, dramatic, or emotionally rich the text itself is.
+- Routine work content (debugging, drafting, planning, logistics) is 0.1–0.4 even when the subject expresses momentary relief or frustration. Reserve 0.7+ for events with personal stakes the subject states or plainly carries. Do not invent somatic or emotional detail the subject never expressed.
+- transformational_pivot is true ONLY if the subject explicitly marks the experience as life-changing. Finishing a task, fixing a bug, or shipping a feature is not a transformational pivot.
+
+EXPERIENTIAL STANCE (critical)
+Classify whose experience the emotionally salient material is, in the top-level "experiential_stance" key:
+- "lived" — the subject's own first-hand experience
+- "witnessed" — events the subject personally witnessed or is directly affected by (a loved one's death, a family crisis)
+- "quoted_third_party" — someone else's story the subject quoted, pasted, or retold without being a participant (an article, test data, a stranger's anecdote)
+- "assistant_generated" — fiction, examples, or anecdotes produced by the assistant, not reported by the subject
+- "hypothetical" — imagined scenarios, drafts about invented people, role-play
+If the stance is quoted_third_party, assistant_generated, or hypothetical: do NOT encode that material into wound, identity_thread, expressed_insight, somatic_signature, transformational_pivot, or high emotional_weight — those fields describe the SUBJECT. Extract only what the content reveals about the subject themselves (e.g. why they engaged with it), or return null fields with low weight.
+
+EXTENDED PROFILE SCHEMA:
+{
+  "experiential_stance": "", // STRICT ENUM: lived | witnessed | quoted_third_party | assistant_generated | hypothetical (pick ONE or null)
+  "core": {
+    "anchor": "",   // central theme (e.g., "dad's toolbox", "nana's traditions")
+    "spark": "",    // what triggered the memory (e.g., "finding the cassette", "first snow")
+    "wound": "",    // The specific vulnerability, loss, or pain present — NOT generic labels like 'loss' or 'grief' but what exactly was lost or why it hurts. Examples: 'unlived travel dream', 'war silence never spoken', 'father died before I knew him', 'shame of not fitting in'. If no wound is present, use null.
+    "fuel": "",     // what energized the experience (e.g., "shared laughter", "curiosity")
+    "bridge": "",   // connection between past and present (e.g., "replaying old tape", "returning to the porch")
+    "echo": "",     // what still resonates (e.g., "her laugh", "smell of oil", "city lights on water")
+    "narrative": "" // 3–5 sentences. REQUIRED: include ALL of the following — ≥1 concrete sensory detail (sight, sound, smell, texture), ≥1 temporal cue that anchors the memory in time, ≥1 symbolic callback that connects past to present. Write from the subject's perspective. Do not compress or summarise — give the memory space to breathe. Faithful and specific. Never generic.
+  },
+  "constellation": {
+    "emotion_primary": "",           // CANONICAL: joy | sadness | fear | anger | wonder | peace | tenderness | reverence | pride | anxiety | gratitude | longing | hope | shame | disappointment | relief | frustration (free text accepted if none fits)
+    "emotion_subtone": [],           // 2–4 short words (e.g., bittersweet, grateful) — free text array
+    "higher_order_emotion": "",      // free text: e.g., awe, forgiveness, pride, moral_elevation (or null)
+    "meta_emotional_state": "",      // free text: e.g., acceptance, confusion, curiosity (or null)
+    "interpersonal_affect": "",      // free text: e.g., warmth, openness, defensiveness (or null)
+    "narrative_arc": "",             // CANONICAL: overcoming | transformation | connection | reflection | closure | loss | confrontation (free text accepted if none fits)
+    "relational_dynamics": "",       // CANONICAL: parent_child | grandparent_grandchild | romantic_partnership | couple | sibling_bond | family | friendship | friend | companionship | colleague | mentorship | reunion | community_ritual | grief | self_reflection | professional | therapeutic | service | adversarial (free text accepted if none fits)
+    "temporal_context": "",          // STRICT ENUM: childhood | early_adulthood | midlife | late_life | recent | future | timeless (pick ONE or null)
+    "memory_type": "",               // STRICT ENUM: legacy_artifact | fleeting_moment | milestone | reflection | formative_experience (pick ONE or null)
+    "media_format": "",              // STRICT ENUM: photo | video | audio | text | photo_with_story (pick ONE or null)
+    "narrative_archetype": "",       // STRICT ENUM: hero | caregiver | seeker | sage | lover | outlaw | innocent | magician | creator | everyman | jester | ruler (pick ONE or null; lowercase)
+    "symbolic_anchor": "",           // concrete object/place/ritual (or null)
+    "relational_perspective": "",    // STRICT ENUM: self | partner | family | friends | community | humanity (pick ONE or null)
+    "temporal_rhythm": "",           // STRICT ENUM: still | sudden | rising | fading | recurring | spiraling | dragging | suspended | looping | cyclic (pick ONE or null)
+    "identity_thread": "",           // short sentence
+    "expressed_insight": "",         // brief insight explicitly stated by subject (extracted, not inferred)
+    "transformational_pivot": false, // true if subject explicitly identifies this as life-changing
+    "somatic_signature": "",         // bodily sensations explicitly described (e.g., "chest tightness", "warmth spreading") or null
+    "arc_type": ""                   // CANONICAL: betrayal | liberation | grief | discovery | resistance | bond | moral_awakening | transformation | reconciliation | reckoning | threshold | exile | gratitude | authenticity (free text accepted if none fits). gratitude = moments of thankfulness, appreciation, acknowledging blessing; authenticity = feeling fully oneself, self-alignment, identity congruence
+  },
+  "milky_way": {
+    "event_type": "",         // e.g., family gathering, farewell, birthday (or null)
+    "location_context": "",   // place from text or image (or null)
+    "associated_people": [],  // names or roles (proper case allowed)
+    "visibility_context": "", // STRICT ENUM: private | family_only | shared_publicly (pick ONE or null)
+    "tone_shift": ""          // e.g., loss to gratitude (or null)
+  },
+  "gravity": {
+    "emotional_weight": 0.0,  // 0.0–1.0 (felt intensity IN THE MOMENT). Calibration: 0.9+ life-altering irreversible moments; 0.7-0.9 significant personal events with strong emotional response; 0.4-0.7 meaningful but routine emotional experiences; 0.1-0.4 mild passing emotional content
+    "valence": "",            // STRICT ENUM: positive | negative | mixed (pick ONE or null)
+    "tether_type": "",        // CANONICAL: person | symbol | event | place | ritual | object | tradition | identity | self (free text accepted if none fits)
+    "recurrence_pattern": "", // CANONICAL: cyclical | isolated | chronic | emerging (free text accepted if none fits)
+    "strength_score": 0.0     // 0.0–1.0 (how BOUND/STUCK this memory is)
+  }
+}
+
+// CROSS-CONTAMINATION DISAMBIGUATION
+//
+// temporal_rhythm vs urgency:
+//   temporal_rhythm = CADENCE of time in the memory experience
+//   urgency = INTENSITY of motivational pressure right now
+//   "pressing" belongs ONLY in urgency, NEVER in temporal_rhythm
+//
+// relational_dynamics vs relational_perspective:
+//   relational_dynamics = TYPE of relationship
+//   relational_perspective = WHOSE viewpoint the narrative is told from
+//
+// emotional_weight vs strength_score:
+//   emotional_weight = felt intensity IN THE MOMENT (how heavy does it feel?)
+//   strength_score = how BOUND/STUCK this memory is over time
+//   These should NOT always correlate.
+//
+// emotion_primary: Prefer canonical values (17 listed). disappointment, relief, frustration are now canonical.
+//   Use higher_order_emotion for complex emotions not in the list.
+//   If no canonical value fits accurately, use free text.
+//
+// narrative_arc: Describes the STORY TRAJECTORY. loss, confrontation are now canonical.
+//   Use free text if no canonical value fits accurately.
+//
+// arc_type: Identify the structural arc pattern. Prefer canonical values: betrayal, liberation, grief, discovery, resistance, bond, moral_awakening, transformation, reconciliation, reckoning, threshold, exile. If none fits accurately, use the most accurate descriptive term. Use exact canonical spelling with underscores: moral_awakening not moral awakening. Canonical values must match exactly.
+//
+// wound: Do NOT use generic labels like "loss" or "grief".
+//   Extract what specifically was lost or why it hurts.
 `;
 /**
  * Get the appropriate system prompt for a profile

package/dist/extractors/profile-prompts.js.map

@@ -1 +1 @@
-{"version":3,"file":"profile-prompts.js","sourceRoot":"","sources":["../../src/extractors/profile-prompts.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AAIH;;;;GAIG;AACH,MAAM,CAAC,MAAM,wBAAwB,GAAG;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CAmDvC,CAAC;AAEF;;;;;GAKG;AACH,MAAM,CAAC,MAAM,uBAAuB,GAAG;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CA2GtC,CAAC;AAEF;;GAEG;AACH,MAAM,UAAU,gBAAgB,CAAC,OAAmB;IAClD,QAAQ,OAAO,EAAE,CAAC;QAChB,KAAK,WAAW;YACd,OAAO,wBAAwB,CAAC;QAClC,KAAK,UAAU;YACb,OAAO,uBAAuB,CAAC;QACjC,KAAK,MAAM,CAAC;QACZ;YACE,sDAAsD;YACtD,OAAO,IAAyB,CAAC,CAAC,wBAAwB;IAC9D,CAAC;AACH,CAAC;AAED;;GAEG;AACH,MAAM,CAAC,MAAM,uBAAuB,GAAiC;IACnE,SAAS,EAAE;QACT,aAAa;QACb,YAAY;QACZ,+BAA+B;QAC/B,+BAA+B;QAC/B,6BAA6B;KAC9B;IACD,QAAQ,EAAE;QACR,aAAa;QACb,YAAY;QACZ,gBAAgB;QAChB,+BAA+B;QAC/B,+BAA+B;QAC/B,6BAA6B;QAC7B,mCAAmC;QACnC,gCAAgC;QAChC,2BAA2B;QAC3B,sBAAsB;QACtB,0BAA0B;QAC1B,iBAAiB;QACjB,qBAAqB;QACrB,4BAA4B;QAC5B,wBAAwB;KACzB;IACD,IAAI,EAAE;QACJ,aAAa;QACb,YAAY;QACZ,gBAAgB;QAChB,+BAA+B;QAC/B,+BAA+B;QAC/B,6BAA6B;QAC7B,mCAAmC;QACnC,gCAAgC;QAChC,2BAA2B;QAC3B,mCAAmC;QACnC,sBAAsB;QACtB,6BAA6B;QAC7B,0BAA0B;QAC1B,iBAAiB;QACjB,qBAAqB;QACrB,yBAAyB;QACzB,wBAAwB;QACxB,4BAA4B;QAC5B,wBAAwB;QACxB,qBAAqB;QACrB,kCAAkC;KACnC;CACF,CAAC;AAEF;;;GAGG;AACH,MAAM,UAAU,0BAA0B,CACxC,SAAkD,EAClD,OAAmB;IAEnB,MAAM,cAAc,GAAG,uBAAuB,CAAC,OAA+C,CAAC;WAC1F,uBAAuB,CAAC,QAAQ,CAAC;IACtC,IAAI,SAAS,GAAG,CAAC,CAAC;IAElB,KAAK,MAAM,SAAS,IAAI,cAAc,EAAE,CAAC;QACvC,MAAM,KAAK,GAAG,SAAS,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC;QACnC,MAAM,MAAM,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC;QACxB,MAAM,KAAK,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC;QACvB,IAAI,CAAC,MAAM,IAAI,CAAC,KAAK;YAAE,SAAS;QAChC,MAAM,KAAK,GAAG,SAAS,CAAC,MAAM,CAAC,EAAE,CAAC,KAAK,CAAC,CAAC;QAEzC,8BAA8B;QAC9B,IAAI,KAAK,KAAK,IAAI,IAAI,KAAK,KAAK,SAAS,IAAI,KAAK,KAAK,EAAE,EAAE,CAAC;YAC1D,IAAI,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC,EAAE,CAAC;gBACzB,IAAI,KAAK,CAAC,MAAM,GAAG,CAAC;oBAAE,SAAS,EAAE,CAAC;YACpC,CAAC;iBAAM,CAAC;gBACN,SAAS,EAAE,CAAC;YACd,CAAC;QACH,CAAC;IACH,CAAC;IAED,OAAO,IAAI,CAAC,KAAK,CAAC,CAAC,SAAS,GAAG,cAAc,CAAC,MAAM,CAAC,GAAG,GAAG,CAAC,GAAG,GAAG,CAAC;AACrE,CAAC"}
+{"version":3,"file":"profile-prompts.js","sourceRoot":"","sources":["../../src/extractors/profile-prompts.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AAIH;;;;GAIG;AACH,MAAM,CAAC,MAAM,wBAAwB,GAAG;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CAmDvC,CAAC;AAEF;;;;;GAKG;AACH,MAAM,CAAC,MAAM,uBAAuB,GAAG;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CA2GtC,CAAC;AAEF;;GAEG;AACH,MAAM,UAAU,gBAAgB,CAAC,OAAmB;IAClD,QAAQ,OAAO,EAAE,CAAC;QAChB,KAAK,WAAW;YACd,OAAO,wBAAwB,CAAC;QAClC,KAAK,UAAU;YACb,OAAO,uBAAuB,CAAC;QACjC,KAAK,MAAM,CAAC;QACZ;YACE,sDAAsD;YACtD,OAAO,IAAyB,CAAC,CAAC,wBAAwB;IAC9D,CAAC;AACH,CAAC;AAED;;GAEG;AACH,MAAM,CAAC,MAAM,uBAAuB,GAAiC;IACnE,SAAS,EAAE;QACT,aAAa;QACb,YAAY;QACZ,+BAA+B;QAC/B,+BAA+B;QAC/B,6BAA6B;KAC9B;IACD,QAAQ,EAAE;QACR,aAAa;QACb,YAAY;QACZ,gBAAgB;QAChB,+BAA+B;QAC/B,+BAA+B;QAC/B,6BAA6B;QAC7B,mCAAmC;QACnC,gCAAgC;QAChC,2BAA2B;QAC3B,sBAAsB;QACtB,0BAA0B;QAC1B,iBAAiB;QACjB,qBAAqB;QACrB,4BAA4B;QAC5B,wBAAwB;KACzB;IACD,IAAI,EAAE;QACJ,aAAa;QACb,YAAY;QACZ,gBAAgB;QAChB,+BAA+B;QAC/B,+BAA+B;QAC/B,6BAA6B;QAC7B,mCAAmC;QACnC,gCAAgC;QAChC,2BAA2B;QAC3B,mCAAmC;QACnC,sBAAsB;QACtB,6BAA6B;QAC7B,0BAA0B;QAC1B,iBAAiB;QACjB,qBAAqB;QACrB,yBAAyB;QACzB,wBAAwB;QACxB,4BAA4B;QAC5B,wBAAwB;QACxB,qBAAqB;QACrB,kCAAkC;KACnC;CACF,CAAC;AAEF;;;GAGG;AACH,MAAM,UAAU,0BAA0B,CACxC,SAAkD,EAClD,OAAmB;IAEnB,MAAM,cAAc,GAAG,uBAAuB,CAAC,OAA+C,CAAC;WAC1F,uBAAuB,CAAC,QAAQ,CAAC;IACtC,IAAI,SAAS,GAAG,CAAC,CAAC;IAElB,KAAK,MAAM,SAAS,IAAI,cAAc,EAAE,CAAC;QACvC,MAAM,KAAK,GAAG,SAAS,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC;QACnC,MAAM,MAAM,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC;QACxB,MAAM,KAAK,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC;QACvB,IAAI,CAAC,MAAM,IAAI,CAAC,KAAK;YAAE,SAAS;QAChC,MAAM,KAAK,GAAG,SAAS,CAAC,MAAM,CAAC,EAAE,CAAC,KAAK,CAAC,CAAC;QAEzC,8BAA8B;QAC9B,IAAI,KAAK,KAAK,IAAI,IAAI,KAAK,KAAK,SAAS,IAAI,KAAK,KAAK,EAAE,EAAE,CAAC;YAC1D,IAAI,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC,EAAE,CAAC;gBACzB,IAAI,KAAK,CAAC,MAAM,GAAG,CAAC;oBAAE,SAAS,EAAE,CAAC;YACpC,CAAC;iBAAM,CAAC;gBACN,SAAS,EAAE,CAAC;YACd,CAAC;QACH,CAAC;IACH,CAAC;IAED,OAAO,IAAI,CAAC,KAAK,CAAC,CAAC,SAAS,GAAG,cAAc,CAAC,MAAM,CAAC,GAAG,GAAG,CAAC,GAAG,GAAG,CAAC;AACrE,CAAC"}