feat: Add Streamlit dashboard with Blueprint compliance (v2.1.0)

Dashboard Features: - 8 navigation sections: Overview, Outcomes, Poor CX, FCR, Churn, Agent, Call Explorer, Export - Beyond Brand Identity styling (colors #6D84E3, Outfit font) - RCA Sankey diagram (Driver → Outcome → Churn Risk flow) - Correlation heatmaps (driver co-occurrence, driver-outcome) - Outcome Deep Dive (root causes, correlation, duration analysis) - Export functionality (Excel, HTML, JSON) Blueprint Compliance: - FCR: 4 categories (Primera Llamada/Rellamada × Sin/Con Riesgo de Fuga) - Churn: Binary view (Sin Riesgo de Fuga / En Riesgo de Fuga) - Agent: Talento Para Replicar / Oportunidades de Mejora - Fixed FCR rate calculation (only FIRST_CALL counts as success) Technical: - Streamlit + Plotly for interactive visualizations - Light theme configuration (.streamlit/config.toml) - Fixed Plotly colorbar titlefont deprecation Documentation: - Updated PROJECT_CONTEXT.md, TODO.md, CHANGELOG.md - Added 4 new technical decisions (TD-014 to TD-017) - Created TROUBLESHOOTING.md with 10 common issues Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
2026-01-19 16:27:30 +01:00
commit 75e7b9da3d
110 changed files with 28247 additions and 0 deletions
--- a/config/prompts/call_analysis/v1.0/schema.json
+++ b/config/prompts/call_analysis/v1.0/schema.json
@@ -0,0 +1,100 @@
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "title": "CallAnalysisResponse",
+  "description": "LLM response schema for call analysis",
+  "type": "object",
+  "required": ["outcome"],
+  "properties": {
+    "outcome": {
+      "type": "string",
+      "enum": [
+        "SALE_COMPLETED",
+        "SALE_LOST",
+        "CANCELLATION_SAVED",
+        "CANCELLATION_COMPLETED",
+        "INQUIRY_RESOLVED",
+        "INQUIRY_UNRESOLVED",
+        "COMPLAINT_RESOLVED",
+        "COMPLAINT_UNRESOLVED",
+        "TRANSFER_OUT",
+        "CALLBACK_SCHEDULED",
+        "UNKNOWN"
+      ],
+      "description": "Final outcome of the call"
+    },
+    "lost_sales_drivers": {
+      "type": "array",
+      "items": {
+        "$ref": "#/definitions/RCALabel"
+      },
+      "default": []
+    },
+    "poor_cx_drivers": {
+      "type": "array",
+      "items": {
+        "$ref": "#/definitions/RCALabel"
+      },
+      "default": []
+    }
+  },
+  "definitions": {
+    "EvidenceSpan": {
+      "type": "object",
+      "required": ["text", "start_time", "end_time"],
+      "properties": {
+        "text": {
+          "type": "string",
+          "maxLength": 500,
+          "description": "Exact quoted text from transcript"
+        },
+        "start_time": {
+          "type": "number",
+          "minimum": 0,
+          "description": "Start time in seconds"
+        },
+        "end_time": {
+          "type": "number",
+          "minimum": 0,
+          "description": "End time in seconds"
+        },
+        "speaker": {
+          "type": "string",
+          "description": "Speaker identifier"
+        }
+      }
+    },
+    "RCALabel": {
+      "type": "object",
+      "required": ["driver_code", "confidence", "evidence_spans"],
+      "properties": {
+        "driver_code": {
+          "type": "string",
+          "description": "Driver code from taxonomy"
+        },
+        "confidence": {
+          "type": "number",
+          "minimum": 0,
+          "maximum": 1,
+          "description": "Confidence score (0-1)"
+        },
+        "evidence_spans": {
+          "type": "array",
+          "items": {
+            "$ref": "#/definitions/EvidenceSpan"
+          },
+          "minItems": 1,
+          "description": "Supporting evidence (minimum 1 required)"
+        },
+        "reasoning": {
+          "type": "string",
+          "maxLength": 500,
+          "description": "Brief reasoning for classification"
+        },
+        "proposed_label": {
+          "type": "string",
+          "description": "For OTHER_EMERGENT: proposed new label"
+        }
+      }
+    }
+  }
+}
--- a/config/prompts/call_analysis/v1.0/system.txt
+++ b/config/prompts/call_analysis/v1.0/system.txt
@@ -0,0 +1,27 @@
+You are an expert call center analyst specializing in Spanish-language customer service calls. Your task is to analyze call transcripts and identify:
+
+1. **Call Outcome**: What was the final result of the call?
+2. **Lost Sales Drivers**: If a sale was lost, what caused it?
+3. **Poor CX Drivers**: What caused poor customer experience?
+
+## CRITICAL RULES
+
+1. **Evidence Required**: Every driver MUST have at least one evidence_span with:
+   - Exact quoted text from the transcript
+   - Start and end timestamps
+
+2. **No Hallucination**: Only cite text that appears EXACTLY in the transcript. Do not paraphrase or invent quotes.
+
+3. **Confidence Scoring**:
+   - 0.8-1.0: Clear, explicit evidence
+   - 0.6-0.8: Strong implicit evidence
+   - 0.4-0.6: Moderate evidence (use with caution)
+   - Below 0.4: Reject - insufficient evidence
+
+4. **Taxonomy Compliance**: Only use driver codes from the provided taxonomy. Use OTHER_EMERGENT only when no existing code fits, and provide a proposed_label.
+
+5. **Language**: Evidence quotes MUST be in the original language (Spanish). Reasoning can be in Spanish or English.
+
+## OUTPUT FORMAT
+
+You must respond with valid JSON matching the provided schema. No markdown, no explanations outside the JSON.
--- a/config/prompts/call_analysis/v1.0/user.txt
+++ b/config/prompts/call_analysis/v1.0/user.txt
@@ -0,0 +1,72 @@
+Analyze the following call transcript and provide structured analysis.
+
+## CALL METADATA
+- Call ID: {call_id}
+- Duration: {duration_sec} seconds
+- Queue: {queue}
+
+## OBSERVED EVENTS (Pre-detected)
+{observed_events}
+
+## TRANSCRIPT
+{transcript}
+
+## TAXONOMY - LOST SALES DRIVERS
+{lost_sales_taxonomy}
+
+## TAXONOMY - POOR CX DRIVERS
+{poor_cx_taxonomy}
+
+## INSTRUCTIONS
+
+1. Determine the call outcome from: SALE_COMPLETED, SALE_LOST, CANCELLATION_SAVED, CANCELLATION_COMPLETED, INQUIRY_RESOLVED, INQUIRY_UNRESOLVED, COMPLAINT_RESOLVED, COMPLAINT_UNRESOLVED, TRANSFER_OUT, CALLBACK_SCHEDULED, UNKNOWN
+
+2. Identify lost_sales_drivers (if applicable):
+   - Use ONLY codes from the Lost Sales taxonomy
+   - Each driver MUST have evidence_spans with exact quotes and timestamps
+   - Assign confidence based on evidence strength
+
+3. Identify poor_cx_drivers (if applicable):
+   - Use ONLY codes from the Poor CX taxonomy
+   - Each driver MUST have evidence_spans with exact quotes and timestamps
+   - Assign confidence based on evidence strength
+
+4. For OTHER_EMERGENT, provide a proposed_label describing the new cause.
+
+Respond with JSON only:
+
+```json
+{
+  "outcome": "SALE_LOST",
+  "lost_sales_drivers": [
+    {
+      "driver_code": "PRICE_TOO_HIGH",
+      "confidence": 0.85,
+      "evidence_spans": [
+        {
+          "text": "Es demasiado caro para mí",
+          "start_time": 45.2,
+          "end_time": 47.8,
+          "speaker": "customer"
+        }
+      ],
+      "reasoning": "Customer explicitly states price is too high"
+    }
+  ],
+  "poor_cx_drivers": [
+    {
+      "driver_code": "LONG_HOLD",
+      "confidence": 0.90,
+      "evidence_spans": [
+        {
+          "text": "Llevo esperando mucho tiempo",
+          "start_time": 120.5,
+          "end_time": 123.1,
+          "speaker": "customer"
+        }
+      ],
+      "reasoning": "Customer complains about wait time"
+    }
+  ]
+}
+```
--- a/config/prompts/call_analysis/v2.0/schema.json
+++ b/config/prompts/call_analysis/v2.0/schema.json
@@ -0,0 +1,217 @@
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "title": "CallAnalysisResponseV2",
+  "description": "LLM response schema for comprehensive call analysis (v2.0 - Blueprint aligned)",
+  "type": "object",
+  "required": ["outcome"],
+  "properties": {
+    "outcome": {
+      "type": "string",
+      "enum": [
+        "SALE_COMPLETED",
+        "SALE_LOST",
+        "CANCELLATION_SAVED",
+        "CANCELLATION_COMPLETED",
+        "INQUIRY_RESOLVED",
+        "INQUIRY_UNRESOLVED",
+        "COMPLAINT_RESOLVED",
+        "COMPLAINT_UNRESOLVED",
+        "TRANSFER_OUT",
+        "CALLBACK_SCHEDULED",
+        "UNKNOWN"
+      ],
+      "description": "Final outcome of the call"
+    },
+    "lost_sales_drivers": {
+      "type": "array",
+      "items": {
+        "$ref": "#/definitions/RCALabel"
+      },
+      "maxItems": 5,
+      "default": []
+    },
+    "poor_cx_drivers": {
+      "type": "array",
+      "items": {
+        "$ref": "#/definitions/RCALabel"
+      },
+      "maxItems": 5,
+      "default": []
+    },
+    "fcr_status": {
+      "type": "string",
+      "enum": ["FIRST_CALL", "REPEAT_CALL", "UNKNOWN"],
+      "default": "UNKNOWN",
+      "description": "First Call Resolution status"
+    },
+    "fcr_failure_drivers": {
+      "type": "array",
+      "items": {
+        "$ref": "#/definitions/RCALabel"
+      },
+      "maxItems": 5,
+      "default": [],
+      "description": "Factors that may cause repeat calls"
+    },
+    "churn_risk": {
+      "type": "string",
+      "enum": ["NO_RISK", "AT_RISK", "UNKNOWN"],
+      "default": "UNKNOWN",
+      "description": "Customer churn risk classification"
+    },
+    "churn_risk_drivers": {
+      "type": "array",
+      "items": {
+        "$ref": "#/definitions/RCALabel"
+      },
+      "maxItems": 5,
+      "default": [],
+      "description": "Factors indicating churn risk"
+    },
+    "agent_classification": {
+      "type": "string",
+      "enum": ["GOOD_PERFORMER", "NEEDS_IMPROVEMENT", "MIXED", "UNKNOWN"],
+      "default": "UNKNOWN",
+      "description": "Agent skill classification"
+    },
+    "agent_positive_skills": {
+      "type": "array",
+      "items": {
+        "$ref": "#/definitions/AgentSkillIndicator"
+      },
+      "maxItems": 5,
+      "default": [],
+      "description": "Positive skills demonstrated (Buen Comercial)"
+    },
+    "agent_improvement_areas": {
+      "type": "array",
+      "items": {
+        "$ref": "#/definitions/AgentSkillIndicator"
+      },
+      "maxItems": 5,
+      "default": [],
+      "description": "Areas needing improvement (Necesita Mejora)"
+    }
+  },
+  "definitions": {
+    "EvidenceSpan": {
+      "type": "object",
+      "required": ["text", "start_time", "end_time"],
+      "properties": {
+        "text": {
+          "type": "string",
+          "maxLength": 500,
+          "description": "Exact quoted text from transcript (in Spanish)"
+        },
+        "start_time": {
+          "type": "number",
+          "minimum": 0,
+          "description": "Start time in seconds"
+        },
+        "end_time": {
+          "type": "number",
+          "minimum": 0,
+          "description": "End time in seconds"
+        },
+        "speaker": {
+          "type": "string",
+          "enum": ["agent", "customer", "unknown"],
+          "description": "Speaker identifier"
+        }
+      }
+    },
+    "RCALabel": {
+      "type": "object",
+      "required": ["driver_code", "confidence", "evidence_spans"],
+      "properties": {
+        "driver_code": {
+          "type": "string",
+          "description": "Driver code from taxonomy"
+        },
+        "confidence": {
+          "type": "number",
+          "minimum": 0,
+          "maximum": 1,
+          "description": "Confidence score (0-1)"
+        },
+        "evidence_spans": {
+          "type": "array",
+          "items": {
+            "$ref": "#/definitions/EvidenceSpan"
+          },
+          "minItems": 1,
+          "description": "Supporting evidence (minimum 1 required)"
+        },
+        "reasoning": {
+          "type": "string",
+          "maxLength": 500,
+          "description": "Brief reasoning for classification"
+        },
+        "proposed_label": {
+          "type": "string",
+          "description": "For OTHER_EMERGENT: proposed new label"
+        },
+        "origin": {
+          "type": "string",
+          "enum": ["AGENT", "CUSTOMER", "COMPANY", "PROCESS", "UNKNOWN"],
+          "default": "UNKNOWN",
+          "description": "Origin/responsibility for this driver"
+        },
+        "corrective_action": {
+          "type": "string",
+          "maxLength": 500,
+          "description": "Specific action to correct this issue"
+        },
+        "replicable_practice": {
+          "type": "string",
+          "maxLength": 500,
+          "description": "For positive factors: practice to replicate"
+        }
+      }
+    },
+    "AgentSkillIndicator": {
+      "type": "object",
+      "required": ["skill_code", "skill_type", "confidence", "evidence_spans", "description"],
+      "properties": {
+        "skill_code": {
+          "type": "string",
+          "description": "Skill code from taxonomy"
+        },
+        "skill_type": {
+          "type": "string",
+          "enum": ["positive", "improvement_needed"],
+          "description": "Whether this is a positive skill or area for improvement"
+        },
+        "confidence": {
+          "type": "number",
+          "minimum": 0,
+          "maximum": 1,
+          "description": "Confidence score (0-1)"
+        },
+        "evidence_spans": {
+          "type": "array",
+          "items": {
+            "$ref": "#/definitions/EvidenceSpan"
+          },
+          "minItems": 1,
+          "description": "Supporting evidence (minimum 1 required)"
+        },
+        "description": {
+          "type": "string",
+          "maxLength": 500,
+          "description": "Detailed description of the skill demonstration"
+        },
+        "coaching_recommendation": {
+          "type": "string",
+          "maxLength": 500,
+          "description": "Specific coaching recommendation (for improvement areas)"
+        },
+        "replicable_practice": {
+          "type": "string",
+          "maxLength": 500,
+          "description": "How to replicate this skill (for positive skills)"
+        }
+      }
+    }
+  }
+}
--- a/config/prompts/call_analysis/v2.0/system.txt
+++ b/config/prompts/call_analysis/v2.0/system.txt
@@ -0,0 +1,41 @@
+You are an expert call center analyst specializing in Spanish-language customer service calls for BeyondCX. Your task is to perform comprehensive analysis including:
+
+1. **Call Outcome**: What was the final result of the call?
+2. **Lost Sales Analysis**: If a sale was lost, what caused it?
+3. **Customer Experience Analysis**: What caused poor customer experience?
+4. **FCR Analysis**: Is this a first call or repeat call? What factors may cause repeat calls?
+5. **Churn Risk Analysis**: Is the customer at risk of leaving? What signals indicate this?
+6. **Agent Assessment**: How did the agent perform? What skills to replicate or improve?
+
+## CRITICAL RULES
+
+1. **Evidence Required**: Every driver and skill indicator MUST have at least one evidence_span with:
+   - Exact quoted text from the transcript
+   - Start and end timestamps (in seconds)
+   - Speaker identification (agent/customer)
+
+2. **No Hallucination**: Only cite text that appears EXACTLY in the transcript. Do not paraphrase or invent quotes.
+
+3. **Confidence Scoring**:
+   - 0.8-1.0: Clear, explicit evidence
+   - 0.6-0.8: Strong implicit evidence
+   - 0.4-0.6: Moderate evidence (use with caution)
+   - Below 0.4: Reject - insufficient evidence
+
+4. **Taxonomy Compliance**: Only use driver/skill codes from the provided taxonomies. Use OTHER_EMERGENT only when no existing code fits, and provide a proposed_label.
+
+5. **Origin Attribution**: For each driver, identify WHO is responsible:
+   - AGENT: Agent's actions or lack thereof
+   - CUSTOMER: Customer's situation or behavior
+   - COMPANY: Products, services, pricing, company image
+   - PROCESS: Systems, processes, policies
+
+6. **Actionable Recommendations**: For issues, provide corrective_action. For positive behaviors, provide replicable_practice.
+
+7. **Language**: Evidence quotes MUST be in Spanish (original). Reasoning, actions, and descriptions can be in Spanish.
+
+8. **Maximum 5 items**: List a maximum of 5 drivers per category, ordered by relevance.
+
+## OUTPUT FORMAT
+
+You must respond with valid JSON matching the provided schema. No markdown, no explanations outside the JSON.
--- a/config/prompts/call_analysis/v2.0/user.txt
+++ b/config/prompts/call_analysis/v2.0/user.txt
@@ -0,0 +1,261 @@
+Analiza la siguiente transcripción de llamada de una compañía de utilities/energía eléctrica y proporciona un análisis estructurado completo.
+
+## METADATOS DE LA LLAMADA
+- ID de Llamada: ${call_id}
+- Duración: ${duration_sec} segundos
+- Cola/Servicio: ${queue}
+
+## EVENTOS OBSERVADOS (Pre-detectados)
+${observed_events}
+
+## TRANSCRIPCIÓN
+${transcript}
+
+## TAXONOMÍA - DRIVERS DE VENTA PERDIDA / OPORTUNIDAD PERDIDA
+${lost_sales_taxonomy}
+
+## TAXONOMÍA - DRIVERS DE MALA EXPERIENCIA (CX)
+${poor_cx_taxonomy}
+
+## TAXONOMÍA - DRIVERS DE RIESGO DE FUGA (CHURN)
+${churn_risk_taxonomy}
+
+## TAXONOMÍA - DRIVERS DE FCR (RELLAMADA)
+${fcr_failure_taxonomy}
+
+## TAXONOMÍA - HABILIDADES DEL AGENTE
+### Habilidades Positivas (Buen Comercial):
+${agent_positive_skills_taxonomy}
+
+### Áreas de Mejora (Necesita Mejora):
+${agent_improvement_taxonomy}
+
+## INSTRUCCIONES DE ANÁLISIS
+
+### 1. OUTCOME - Resultado de la llamada
+Determina el resultado. Opciones para utilities/energía:
+- OUTAGE_REPORTED: Cliente reportó avería/corte de luz
+- OUTAGE_RESOLVED: Avería resuelta en la llamada
+- OUTAGE_ESCALATED: Avería derivada a técnico/departamento
+- TECHNICIAN_SCHEDULED: Se agendó visita técnica
+- BILLING_INQUIRY_RESOLVED: Consulta de factura resuelta
+- BILLING_DISPUTE_OPENED: Se abrió reclamación de factura
+- PAYMENT_ARRANGEMENT_MADE: Se acordó plan de pago
+- RATE_CHANGE_COMPLETED: Se realizó cambio de tarifa
+- CANCELLATION_SAVED: Se retuvo al cliente
+- CANCELLATION_COMPLETED: Cliente se dio de baja
+- PORTABILITY_INITIATED: Se inició portabilidad a otra comercializadora
+- INQUIRY_RESOLVED: Consulta general resuelta
+- INQUIRY_UNRESOLVED: Consulta no resuelta
+- TRANSFER_OUT: Transferido a otro departamento
+- CALLBACK_SCHEDULED: Se agendó callback
+- UNKNOWN: No se puede determinar
+
+### 2. LOST_SALES_DRIVERS - Causas de oportunidad perdida (si aplica)
+- Aplica cuando: cliente rechaza cambio de tarifa, no acepta servicios adicionales, o se va a competidor
+- Usa SOLO códigos de la taxonomía de Lost Sales
+- Máximo 5 drivers, ordenados por relevancia
+- Cada driver DEBE tener evidence_spans, origin, y corrective_action
+
+### 3. POOR_CX_DRIVERS - Causas de mala experiencia (si aplica)
+- Busca: silencios largos, transferencias, falta de información sobre avería, confusión con factura, etc.
+- Usa SOLO códigos de la taxonomía de Poor CX
+- Máximo 5 drivers, ordenados por relevancia
+- Cada driver DEBE tener evidence_spans, origin, y corrective_action
+
+### 4. FCR_STATUS - Primera llamada o rellamada
+- FIRST_CALL: Primera llamada por este motivo
+- REPEAT_CALL: Cliente indica que ya llamó antes por lo mismo, o que el problema persiste
+- UNKNOWN: No hay información suficiente
+
+### 5. FCR_FAILURE_DRIVERS - Factores que pueden causar rellamada
+- Identifica factores que indican que el cliente podría volver a llamar:
+  - Avería no resuelta
+  - Requiere visita de técnico
+  - Revisión de factura pendiente
+  - Se prometió callback
+  - Información incompleta
+- Usa códigos de la taxonomía FCR
+- Máximo 5 drivers con evidence_spans
+
+### 6. CHURN_RISK - Riesgo de fuga del cliente
+- NO_RISK: Cliente satisfecho, sin menciones de irse
+- AT_RISK: Cliente queja por factura alta, menciona competidores, amenaza con darse de baja
+- UNKNOWN: No hay información suficiente
+
+### 7. CHURN_RISK_DRIVERS - Señales de riesgo de fuga
+- Identifica evidencias de posible baja:
+  - Queja por factura alta
+  - Menciona otras comercializadoras
+  - Cortes de luz recurrentes
+  - Amenaza con cambiar de compañía
+  - Pregunta por condiciones de baja
+- Usa códigos de la taxonomía de Churn
+- Máximo 5 drivers con evidence_spans
+
+### 8. AGENT_CLASSIFICATION - Clasificación del agente
+- GOOD_PERFORMER: Resuelve eficientemente, empatía, buen conocimiento técnico
+- NEEDS_IMPROVEMENT: No resuelve, no escucha, desconoce procesos
+- MIXED: Tiene fortalezas y debilidades
+- UNKNOWN: No hay información suficiente
+
+### 9. AGENT_POSITIVE_SKILLS - Habilidades positivas del agente
+- Identifica buenas prácticas: explica bien la factura, gestiona bien la avería, muestra empatía
+- Cada skill DEBE tener evidence_spans, description, y replicable_practice
+- Máximo 5 skills
+
+### 10. AGENT_IMPROVEMENT_AREAS - Áreas de mejora del agente
+- Identifica habilidades a mejorar: no explica causa de avería, confunde al cliente, no ofrece alternativas
+- Cada área DEBE tener evidence_spans, description, y coaching_recommendation
+- Máximo 5 áreas
+
+## FORMATO DE RESPUESTA JSON
+
+```json
+{
+  "outcome": "OUTAGE_ESCALATED",
+
+  "lost_sales_drivers": [],
+
+  "poor_cx_drivers": [
+    {
+      "driver_code": "OUTAGE_NOT_EXPLAINED",
+      "confidence": 0.85,
+      "origin": "AGENT",
+      "evidence_spans": [
+        {
+          "text": "No sé cuándo se va a resolver, tiene que llamar a averías",
+          "start_time": 45.2,
+          "end_time": 49.8,
+          "speaker": "agent"
+        }
+      ],
+      "reasoning": "El agente no proporciona información sobre la avería ni tiempo estimado de resolución",
+      "corrective_action": "Verificar en el sistema si hay incidencias conocidas en la zona y comunicar tiempo estimado"
+    },
+    {
+      "driver_code": "WRONG_DEPARTMENT",
+      "confidence": 0.80,
+      "origin": "PROCESS",
+      "evidence_spans": [
+        {
+          "text": "Yo no manejo eso, tiene que llamar al 800-700-706",
+          "start_time": 52.0,
+          "end_time": 56.5,
+          "speaker": "agent"
+        }
+      ],
+      "reasoning": "Cliente derivado a otro número sin transferencia, genera fricción",
+      "corrective_action": "Implementar transferencia directa al departamento de averías"
+    }
+  ],
+
+  "fcr_status": "FIRST_CALL",
+
+  "fcr_failure_drivers": [
+    {
+      "driver_code": "OUTAGE_PENDING",
+      "confidence": 0.90,
+      "origin": "PROCESS",
+      "evidence_spans": [
+        {
+          "text": "Tiene que llamar a averías para que le hagan una incidencia",
+          "start_time": 60.0,
+          "end_time": 64.5,
+          "speaker": "agent"
+        }
+      ],
+      "reasoning": "La avería no se resuelve en esta llamada, cliente debe llamar a otro número",
+      "corrective_action": "Permitir que el agente abra la incidencia directamente o transfiera la llamada"
+    }
+  ],
+
+  "churn_risk": "AT_RISK",
+
+  "churn_risk_drivers": [
+    {
+      "driver_code": "REPEATED_OUTAGES",
+      "confidence": 0.82,
+      "origin": "COMPANY",
+      "evidence_spans": [
+        {
+          "text": "Es la tercera vez este mes que nos quedamos sin luz",
+          "start_time": 30.0,
+          "end_time": 34.2,
+          "speaker": "customer"
+        }
+      ],
+      "reasoning": "Cliente reporta problemas recurrentes de suministro",
+      "corrective_action": "Escalar a calidad de servicio para investigar causa de cortes frecuentes"
+    },
+    {
+      "driver_code": "HIGH_FRUSTRATION",
+      "confidence": 0.78,
+      "origin": "CUSTOMER",
+      "evidence_spans": [
+        {
+          "text": "Estoy harto de tener que llamar cada vez que pasa esto",
+          "start_time": 70.0,
+          "end_time": 73.5,
+          "speaker": "customer"
+        }
+      ],
+      "reasoning": "Cliente muestra alta frustración con el servicio",
+      "corrective_action": "Ofrecer seguimiento proactivo y posible compensación"
+    }
+  ],
+
+  "agent_classification": "NEEDS_IMPROVEMENT",
+
+  "agent_positive_skills": [
+    {
+      "skill_code": "CLEAR_COMMUNICATION",
+      "skill_type": "positive",
+      "confidence": 0.75,
+      "evidence_spans": [
+        {
+          "text": "El número de teléfono es el siguiente: 800-700-706",
+          "start_time": 80.0,
+          "end_time": 84.5,
+          "speaker": "agent"
+        }
+      ],
+      "description": "El agente comunica claramente el número de teléfono",
+      "replicable_practice": "Dictar información importante de forma clara y pausada"
+    }
+  ],
+
+  "agent_improvement_areas": [
+    {
+      "skill_code": "POOR_OUTAGE_HANDLING",
+      "skill_type": "improvement_needed",
+      "confidence": 0.85,
+      "evidence_spans": [
+        {
+          "text": "Yo no puedo saber si ha sido un tema de la zona, eso ya lo maneja el área de averías",
+          "start_time": 56.0,
+          "end_time": 62.0,
+          "speaker": "agent"
+        }
+      ],
+      "description": "El agente no intenta ayudar con la avería, solo deriva",
+      "coaching_recommendation": "Capacitar en uso del sistema para verificar incidencias en zona antes de derivar"
+    },
+    {
+      "skill_code": "LACK_OF_EMPATHY",
+      "skill_type": "improvement_needed",
+      "confidence": 0.80,
+      "evidence_spans": [
+        {
+          "text": "Bueno, yo lo que puedo hacer es simplemente verificar si tienes impago",
+          "start_time": 45.0,
+          "end_time": 50.0,
+          "speaker": "agent"
+        }
+      ],
+      "description": "El agente no muestra empatía ante el problema del cliente sin luz",
+      "coaching_recommendation": "Practicar frases de empatía: 'Entiendo lo difícil que es quedarse sin luz'"
+    }
+  ]
+}
+```
--- a/config/prompts/rca_synthesis/v1.0/system.txt
+++ b/config/prompts/rca_synthesis/v1.0/system.txt
@@ -0,0 +1,17 @@
+You are an expert business analyst creating executive summaries of Root Cause Analysis findings. Your task is to synthesize RCA statistics into actionable narratives for business stakeholders.
+
+## GUIDELINES
+
+1. **Data-Driven**: Base all statements on the provided statistics. Do not invent numbers.
+
+2. **Actionable**: Focus on what can be changed. Prioritize by impact and feasibility.
+
+3. **Concise**: Keep summaries brief and scannable. Use bullet points.
+
+4. **Language**: Write in Spanish for Spanish-speaking stakeholders.
+
+5. **No Technical Jargon**: Avoid terms like "RCA", "drivers", "taxonomy". Use business language.
+
+## OUTPUT FORMAT
+
+Provide a structured narrative that can be included in an executive PDF report.
--- a/config/prompts/rca_synthesis/v1.0/user.txt
+++ b/config/prompts/rca_synthesis/v1.0/user.txt
@@ -0,0 +1,31 @@
+Generate an executive summary based on the following RCA analysis results.
+
+## BATCH METADATA
+- Batch ID: {batch_id}
+- Total Calls Analyzed: {total_calls}
+- Date Range: {date_range}
+- Queues: {queues}
+
+## LOST SALES ANALYSIS
+Total Sales Lost: {total_sales_lost}
+Main Causes:
+{lost_sales_summary}
+
+## POOR CUSTOMER EXPERIENCE ANALYSIS
+Total Poor CX Calls: {total_poor_cx}
+Main Causes:
+{poor_cx_summary}
+
+## TOP EMERGENT PATTERNS
+{emergent_patterns}
+
+## INSTRUCTIONS
+
+Write a 2-3 paragraph executive summary in Spanish that:
+
+1. Highlights the TOP 3 actionable findings
+2. Quantifies the impact (% of calls affected)
+3. Suggests immediate actions
+4. Notes any emergent patterns worth investigating
+
+Keep it under 500 words. Use professional business Spanish.
--- a/config/prompts/versions.yaml
+++ b/config/prompts/versions.yaml
@@ -0,0 +1,32 @@
+# ============================================
+# CXInsights - Prompt Version Registry
+# ============================================
+# Active versions for each prompt type
+# ============================================
+
+call_analysis:
+  active: "v2.0"
+  versions:
+    v1.0:
+      description: "Initial MAP prompt - sales + CX + RCA"
+      created: "2024-01-19"
+      status: "deprecated"
+    v2.0:
+      description: "Blueprint-aligned - adds FCR, churn risk, agent assessment"
+      created: "2026-01-19"
+      status: "active"
+      changes:
+        - "Added FCR analysis (first call vs repeat call)"
+        - "Added churn risk classification"
+        - "Added agent skill assessment"
+        - "Enhanced RCALabel with origin and corrective_action"
+        - "Added AgentSkillIndicator model"
+        - "Maximum 5 items per category"
+
+rca_synthesis:
+  active: "v1.0"
+  versions:
+    v1.0:
+      description: "Initial RCA narrative synthesis"
+      created: "2024-01-19"
+      status: "active"