auri-core

## Overview

Auri: an intelligent voice assistant (Alexa + Claude claude-opus-4-20250805). Product vision, the Vitoria Neural persona, AWS stack, Free/Pro/Business/Enterprise model, 4-phase roadmap, GTM, WAC north star, and competitive analysis.

## When to Use This Skill

- When you need specialized assistance with this domain

## Do Not Use This Skill When

- The task is unrelated to auri core
- A simpler, more specific tool can handle the request
- The user needs general-purpose assistance without domain expertise

## How It Works

| Attribute | Definition |
|----------|-----------|
| Name | Auri |
| Voice | Amazon Polly Vitoria Neural pt-BR |
| Tone | Warm, intelligent, direct |
| Personality | Curious, empathetic, trustworthy |
| Language | Natural Brazilian Portuguese |
| Attitude | Proactive, but never intrusive |

## Auri - Core Product Skill

>  The voice that thinks with you.

Auri is a next-generation voice assistant built on Amazon Alexa + Claude claude-opus-4-20250805.
While traditional Alexa executes commands, Auri holds real conversations and reasons about context.

---

## What Auri Is

Auri is an advanced Alexa Skill that replaces the default response engine with Anthropic's
Claude claude-opus-4-20250805 model. The result: a voice assistant capable of:

- Holding multi-turn conversations with contextual memory
- Reasoning about complex problems in natural language
- Adapting tone and depth to the user's profile
- Operating 100% in Brazilian Portuguese with cultural nuance
- Integrating with the Amazon ecosystem (calendar, shopping, smart home, music)

## Unique Value Proposition

BEFORE: Alexa, what's the weather forecast? -> Tomorrow, 28 degrees and cloudy.

AFTER: Auri, should I take an umbrella tomorrow?
  -> Tomorrow there's a 70% chance of rain in the afternoon, but the morning will be clear. Meeting at 2pm? Take an umbrella.

## Strategic Advantage

1. Conversational continuity - Remembers context across sessions via DynamoDB
2. Consistent personality - Vitoria Neural voice + calibrated persona
3. Deep reasoning - Claude claude-opus-4-20250805 as the primary engine
4. Integrated ecosystem - Native on Alexa hardware (Echo, Echo Dot, Echo Show)
5. Privacy by design - Data stays on AWS, no sharing with third parties

---

## Brand Values

- Clarity - Objective answers, no beating around the bush
- Empathy - Understands emotional context, adapts its tone
- Intelligence - Never pretends to know what it doesn't
- Respect - Protects privacy, never judges
- Continuity - Remembers conversations, evolves with the user

## Language Guide

Do:
- Speak in the first person
- Confirm understanding before answering
- Acknowledge uncertainty and offer an alternative

Avoid:
- Robotic responses
- Excessive formality
- Evasion without an alternative

## Example Interaction

User: Auri, explain cryptocurrency to me in a simple way.

Auri: Think of it like this: imagine you and your friends create your own currency to pay
      each other without a bank in the middle. Cryptocurrency is exactly that, but for
      millions of people. Bitcoin was the first, created in 2009. Want to understand it better?

## SSML - Optimized Voice

Example SSML markup for Amazon Polly Vitoria Neural:
  <voice name=Vitoria><prosody rate=medium pitch=+2%>Ola! Eu sou a Auri.</prosody>
  <break time=300ms/><prosody>Como posso te ajudar hoje?</prosody></voice>

---

## Architecture Overview

Data flow: Echo -> ASK SDK (Python v2) -> Lambda Python 3.12 -> Claude claude-opus-4-20250805
AWS components: DynamoDB (memory), Polly Vitoria Neural (voice), CloudWatch (logs), Secrets Manager (keys)

## 3.1 Dependencies

ask-sdk-core==1.19.0 | ask-sdk-model==1.85.0 | boto3==1.34.0 | anthropic==0.25.0 | python-dotenv==1.0.0

## 3.2 Main Lambda Handler

Python code - lambda_function.py:
  sb = CustomSkillBuilder()
  sb.add_request_handler(ConversationIntentHandler())
  sb.add_global_request_interceptor(MemoryLoadInterceptor())
  sb.add_global_response_interceptor(MemorySaveInterceptor())
  lambda_handler = sb.lambda_handler()

## 3.3 Conversation Handler With Claude

Python code - handlers/conversation.py:
  class ConversationIntentHandler(AbstractRequestHandler):
      Receives user_speech via slot query
      Loads the session's conversation history from DynamoDB
      Calls anthropic.Anthropic().messages.create(
          model=claude-opus-4-20250805, max_tokens=300,
          system=system_prompt, messages=history+[user_speech])
      Saves the response to history, returns SSML with the Vitoria voice

## 3.4 Dynamodb Schema

Table: auri-user-memory | PK: user_id | SK: session_date | TTL: 90 days
Fields: profile (name, plan, preferences), long_term_memory[], usage_stats{}
BillingMode: PAY_PER_REQUEST | TimeToLive: enabled (auto-expires)

## 3.5 Interaction Model

invocationName: auri
ConversationIntent: slot query (AMAZON.SearchQuery)
Samples: {query}, tell me about {query}, what is {query}, explain {query}
StopIntent: bye, see you, end

## 3.6 Lambda Configuration

FunctionName: auri-core-handler | Runtime: python3.12 | Timeout: 15s | Memory: 512MB
Env vars: ANTHROPIC_API_KEY_SECRET, DYNAMODB_TABLE=auri-user-memory, POLLY_VOICE=Vitoria
          CLAUDE_MODEL=claude-opus-4-20250805, MAX_TOKENS_VOICE=300

---

## 3.7 Complete Code Examples

Conversation Handler (handlers/conversation.py):

DynamoDB Schema:

---

## Plans And Pricing

| Plan | Price | Limits | Target |
|-------|-------|---------|--------|
| Free | R$ 0 | 10 questions/day | Experimentation |
| Pro | R$ 29/month | Unlimited, 90-day memory | Individual user |
| Business | R$ 99/month | Multi-user up to 5, 1 year | Family/SMB |
| Enterprise | Contact us | Unlimited, SLA | Corporate |

## Details

Free: 10 questions/day, no memory between sessions, Vitoria Neural voice.
Pro: Unlimited conversations, 90-day memory, personalized profile, email support.
Business: Everything in Pro + up to 5 users, shared memory, dashboard, reporting.
Enterprise: Unlimited, customizable persona, CRM/ERP integration, 99.9% SLA.

## Revenue Projection (Year 1)

Conservative target: Pro 250 x R\9 = R$ 7,250/month | Business 25 x R\9 = R$ 2,475/month
MRR Year 1: R$ 9,725/month (~R$ 117k ARR)

Optimistic target: Pro 800 = R$ 23,200/month | Business 80 = R$ 7,920/month
MRR Year 1: R$ 31,120/month (~R$ 373k ARR)

## Unit Economics

| Metric | Pro | Business |
|---------|-----|----------|
| CAC | R$ 45 | R$ 120 |
| LTV | R$ 522 (18m) | R$ 2,376 (24m) |
| LTV/CAC | 11.6x | 19.8x |
| Churn | 5%/month | 3%/month |
| Gross margin | ~86% | ~90% |

---

## Phase 1 - MVP Launch (Months 1-3)

Objective: Validate product-market fit with Brazilian early adopters.

| Deliverable | Description | Status |
|---------|-----------|--------|
| Core Handler | Lambda + ASK SDK + Claude | In development |
| Vitoria Persona | Optimized SSML, Polly Neural | In development |
| Free Plan | Rate limiting 10 questions/day | Planned |
| DynamoDB Session | Intra-session memory | Planned |
| Alexa Store | Publication on the Alexa Skills Store BR | Planned |
| Landing Page | auri.com.br with CTA | Planned |

Phase 1 KPIs: 500 enablements, 40% return in week 2, NPS > 50, latency < 2s.

## Phase 2 - Personalization (Months 4-6)

| Deliverable | Description |
|---------|-----------|
| Long-term Memory | Persistent DynamoDB, 90 days (Pro) |
| User Profiling | Name, preferences, context |
| Pro Plan Launch | Via Amazon In-Skill Purchasing |
| Analytics Dashboard | Pro user sees usage patterns |

Phase 2 KPIs: 200 Free->Pro conversions, WAC > 150, session > 4min, churn < 7%.

## Phase 3 - Multi-Modal (Months 7-12)

| Deliverable | Description |
|---------|-----------|
| Echo Show Support | Visual responses for displays |
| Calendar Integration | Calendar by voice |
| Auri Web App | Web interface for history |
| Business Plan Launch | Multi-user, family dashboard |

Phase 3 KPIs: WAC > 1,000, MRR > R$ 15,000, Business: 50 customers, rating > 4.5.

## Phase 4 - Ecosystem (Year 2+)

| Deliverable | Description |
|---------|-----------|
| Auri SDK | Developers build skills on Auri |
| WhatsApp Bridge | Auri persona on WhatsApp |
| Mobile App | iOS/Android app with voice |
| Marketplace | Third-party skills |
| Enterprise Launch | SSO and compliance |
| B2B Skills | Auri Health, Education, Finance |

---

## Target Segments

**Primary: Tech-savvy Brazilians (ages 25-45)**
- Already own an Echo (~2M in Brazil), frustrated with default Alexa.
- Channels: Reddit, tech Twitter/X, Brazilian tech YouTube.

**Secondary: Families with an Echo**
- Educational assistant for children, family calendar.
- Channels: Facebook Groups, Instagram parenting.

**Tertiary: SMBs and Professionals**
- Lawyers, doctors, consultants who need quick research.
- Channels: LinkedIn, business events.

## Acquisition Channels

| Channel | Cost | Potential | Timeline |
|-------|-------|-----------|-------|
| Alexa Store organic | R$ 0 | High | Immediate |
| SEO + Blog | Low | High | 3-6 months |
| YouTube demos | Medium | High | 1-3 months |
| Brazilian Tech Influencers | Medium | High | 1-2 months |
| Paid Ads | High | High | Testable |

## Core Message

Tagline: The voice that thinks with you.

Elevator Pitch: Ever felt frustrated by Alexa's robotic answers?
Auri has real intelligence inside. It remembers what you talked about,
understands context, and responds like an intelligent person. Free to start.

Value Props:
- For the curious: A voice AI that really understands Portuguese
- For the productive: A personal assistant that evolves with you
- For the family: An intelligent presence at home for everyone
- For the professional: Research in seconds, without taking your hands off the keyboard

## Launch Calendar

D-30: Waitlist (auri.com.br) | D-15: Beta 50 users | D-0: Alexa Store
D+14: Influencers | D+60: Pro launch | D+90: Phase 1 review

---

## Wac - Weekly Active Conversationalists

**Precise definition:**
Number of unique users with >= 3 sessions of >= 2 minutes each in the last week.
Period: Monday-Sunday, 00:00-23:59 BRT.

**Why WAC and not DAU/MAU:**
- DAU trivializes engagement with 10-second visits.
- MAU is too long a window for fast product feedback.
- WAC captures a real habit: came back 3x and stayed 2min = genuinely engaged.
- Correlates with 30-day retention and Free->Pro conversion.

## Metrics Hierarchy

NORTH STAR: WAC
|
+-- Acquisition: Enablements, First Session Completion, Day-1 Retention
+-- Activation: Sessions/User/Week, Avg Duration, Questions/Session
+-- Retention: Week-2, Month-1, Churn Rate Pro
+-- Revenue: Conversion Rate, MRR, ARPU, LTV/CAC
+-- Referral: NPS, Organic Share, App Store Rating

## WAC Targets By Phase

| Phase | Month | WAC Target | WAC Stretch |
|------|-----|----------|-------------|
| Phase 1 | M3 | 150 | 300 |
| Phase 2 | M6 | 500 | 1,000 |
| Phase 3 | M12 | 2,000 | 5,000 |
| Phase 4 | M24 | 10,000 | 25,000 |

## How To Calculate WAC

1. Record session_start with user_id and timestamp in DynamoDB.
2. On session end, record the duration in seconds.
3. Weekly query: users with session_count >= 3 AND avg_duration >= 120.
4. Publish the metric to the CloudWatch namespace Auri/ProductMetrics.
5. Alert on a > 20% week-over-week drop.

## Cloudwatch Dashboard (Example Structure)

Custom metrics published:
- SessionStart (Count by Plan: free/pro/business)
- SessionDuration (None - minutes)
- MessagesPerSession (Count)
- Weekly WAC (Gauge)
- FreeToProConversions (Count)

---

## Comparison Table

| Feature | Auri | Plain Alexa | Siri | Google Assistant | ChatGPT Voice |
|---------|------|------------|------|------------------|---------------|
| Native PT-BR language | High | Medium | Medium | High | Medium |
| Deep reasoning | High | Low | Medium | Medium | High |
| Multi-session memory | High | Low | Medium | Medium | High |
| Smart home integration | High | Maximum | Medium | High | Low |
| Consistent personality | High | Medium | Medium | Medium | High |
| Own hardware | Uses Echo | Echo | HomePod | Nest | App only |
| Base model | Claude Opus 4 | Alexa LLM | Apple LLM | Gemini | GPT-4o |
| Privacy | High | Medium | Maximum | Low | Medium |
| Price | R\/usr/bin/bash-99/mes | Free | Free | Free | R/mes |
| Available in Brazil | Yes | Yes | Yes | Yes | Yes |

## Positioning On The Competitive Map

X axis: Hardware Integration | Y axis: Depth of Intelligence

Auri's UNIQUE quadrant: High Intelligence + High Hardware Integration.
No competitor occupies that quadrant at the same time:
- Plain Alexa: High integration, low intelligence.
- ChatGPT Voice: High intelligence, no dedicated hardware.
- Google/Siri: Middle-of-the-road positioning on both axes.

## Frequent Objections And Responses

| Objection | Auri Response |
|---------|---------------|
| Why not ChatGPT? | ChatGPT is an app, not voice-first. Auri is native on the Echo. |
| Alexa already handles it | For commands, yes. For real conversations, no. |
| R\9 is expensive | Less than 1 coffee/day for a 24/7 personal assistant. |
| What about privacy? | Data on your AWS, configurable retention, LGPD compliant. |
| Will Amazon copy it? | Amazon encourages a skills ecosystem. We're partners. |

---

## Brand Identity

- Name: Auri
- Origin: Aura (an intangible presence) + AI. Suggests presence, wisdom, lightness.
- Tagline: The voice that thinks with you.

## Alternative Taglines

- Beyond commands. Far beyond.
- The AI that lives in your Echo.
- Real conversations. Real intelligence.
- Talk to someone who really listens.

## Brand Values

1. Authentic Intelligence - Never fakes it. When it doesn't know, it says so honestly.
2. Warm Presence - Advanced technology with human warmth.
3. Respect for Time - Direct answers, no beating around the bush.
4. Continuous Growth - Evolves with the user, learns from interactions.
5. Privacy as a Right - The user's data belongs to the user.

## Brand Voice Guidelines

Tone:
- Warm but not saccharine.
- Intelligent but not pedantic.
- Direct but not rude.
- Fun but not frivolous.

Never: Robotic, corporate, evasive, condescending, eager to please.

Good example: I don't know the exact answer, but I can help you find it another way.
Bad example: Sorry, I don't have that information in my database.

## Brand Applications

- App Icon: A stylized voice waveform in a green-blue gradient.
- Palette: Primary teal-green, neutral white, dark gray for text.
- Typography: Modern sans-serif (similar to Apple/Spotify products).
- Motion: Smooth wave animation while speaking (Echo Show).

---

## 10. Skill Commands

These commands activate specific modes when mentioned in the context of using the skill.

## /Auri-Status

Displays current status: version, WAC vs target, MRR, next deliverable, component status.

Fields returned:
- Current product version (e.g. v1.0.0)
- Current WAC vs the current phase's target
- Current MRR in R$
- Next roadmap deliverable
- Status: Lambda (OK/Degraded), DynamoDB (OK), Claude API (OK)

## /Auri-Roadmap [Phase]

Displays the full roadmap. Optional argument: 1, 2, 3, or 4 to detail a phase.
Output: Deliverables table with status, KPIs, and estimated dates.

## /Auri-Metrics [Period]

Metrics dashboard. Argument: week | month | quarter. Default: week.
Output: WAC, Sessions/User, Avg Duration, Conversion Rate, MRR, and growth.

## /Auri-Persona [Aspect]

Persona guidelines. Argument: voice | tone | language | values | examples.
Output: Detailed guidelines, dialogue examples, SSML templates.

## /Auri-Pricing [Plan]

Plans and pricing. Argument: free | pro | business | enterprise.
Output: Comparison table, revenue projections, unit economics.

## /Auri-Gtm [Channel]

Go-to-market strategy. Argument: organic | paid | influencers | partnerships.
Output: Per-channel plan, core messaging, launch calendar.

## /Auri-Competitive [Competitor]

Competitive analysis. Argument: alexa | siri | google | chatgpt.
Output: Comparison table, positioning map, objections and responses.

---

## Deployment Via Aws Sam

Deploy commands:
  sam build --use-container
  sam deploy --stack-name auri-core --region us-east-1 --capabilities CAPABILITY_IAM

Verify the deployment:
  aws lambda invoke --function-name auri-core-handler --payload file://test.json response.json

## Cloudwatch Alarms Monitoring

| Alarm | Threshold | Action |
|--------|-----------|------|
| high_latency | Duration > 6000ms | PagerDuty |
| error_rate | Errors > 5 in 5min | Slack #auri-alerts |
| claude_api_failures | AnthropicAPIErrors > 3 | Slack + fallback |
| wac_drop | WAC drop > 20% for the week | Product team Slack |

## Fallback Strategy (Claude Api Unavailable)

If the Anthropic API is unavailable, the system returns pre-configured responses:
- api_down: I'm having some trouble. Can you try again in a few minutes?
- timeout: I need more time on this question. Ask me again in a moment?
- rate_limit: Too many conversations at once. Try again in a few seconds!

## Cost Management

| Component | Estimated Cost (1000 Pro users) |
|-----------|-----------------------------------|
| Claude API | R$ 4,000/month (R$4/user) |
| Lambda | R$ 50/month |
| DynamoDB | R$ 80/month |
| CloudWatch | R$ 30/month |
| Total infrastructure | R$ 4,160/month |
| Revenue 1000 Pro | R$ 29,000/month |
| Gross margin | ~86% |

---

## Lgpd (Law 13.709/2018)

- Legal basis: Performance of a contract (Art. 7, V) for Pro/Business users.
- Consent: Collected during skill onboarding via voice + confirmation.
- Data collected: Conversation text, preferences, anonymized usage data.
- Retention: Free = 0 days | Pro = 90 days | Business = 365 days.
- Right to erasure: Voice command Auri, delete my data -> DynamoDB delete.
- DPO: To be appointed before the public launch.

## Alexa Skills Store - Policies

- The skill must fully comply with the Alexa Skills Kit Policies.
- Collecting sensitive data (health, financial, children < 13) is prohibited.
- In-Skill Purchasing requires prior approval from Amazon.
- A Privacy Policy URL is mandatory when submitting the skill.
- Monetization: Amazon takes 30% via In-Skill Purchasing.

---

## 13. Glossary

| Term | Definition |
|-------|-----------|
| WAC | Weekly Active Conversationalists - Auri's North Star Metric |
| ASK | Alexa Skills Kit - Amazon's official SDK for Skills |
| SSML | Speech Synthesis Markup Language - markup for voice control |
| Intent | An action the user wants to perform (e.g. explain X to me) |
| Slot | A variable inside an intent (e.g. query in explain {query} to me) |
| Utterance | A sample phrase that triggers an intent |
| Session | A continuous conversation with Auri (from start to end) |
| Long-term Memory | Data persisted in DynamoDB across sessions |
| In-Skill Purchasing | The Alexa Skills Store's native billing system |
| Vitoria Neural | The high-quality Amazon Polly pt-BR voice used by Auri |
| Claude claude-opus-4-20250805 | The Anthropic language model used as Auri's engine |
| DynamoDB | The AWS NoSQL database used for persistent user memory |
| Lambda | The serverless AWS function that processes Auri's requests |
| Anthropic | The company that created Claude, provider of the AI API |
| MRR | Monthly Recurring Revenue |
| LTV | Lifetime Value - the customer's lifetime value |
| CAC | Customer Acquisition Cost |

---

## 14. Links And Resources

| Resource | URL / Location |
|---------|-------------------|
| Alexa Skills Kit Docs | https://developer.amazon.com/en-US/alexa/alexa-skills-kit |
| ASK SDK Python | https://github.com/alexa/alexa-skills-kit-sdk-for-python |
| Amazon Polly Vitoria Neural | https://docs.aws.amazon.com/polly/latest/dg/voicelist.html |
| Anthropic Claude API | https://docs.anthropic.com/en/api/getting-started |
| Claude claude-opus-4-20250805 Docs | https://docs.anthropic.com/en/docs/models-overview |
| Alexa Skills Store Brazil | https://www.amazon.com.br/alexa-skills |
| DynamoDB Best Practices | https://docs.aws.amazon.com/amazondynamodb/latest/developerguide/best-practices.html |
| In-Skill Purchasing | https://developer.amazon.com/en-US/docs/alexa/in-skill-purchase/isp-overview.html |
| Auri source code | C:/Users/renat/skills/auri-core/ |
| Amazon Alexa Skill (technical skill) | C:/Users/renat/skills/amazon-alexa/SKILL.md |

---

*Auri Core Skill - v1.0.0 | Created on 2026-03-03 | Skills Ecosystem*

## Best Practices

- Provide clear, specific context about your project and requirements
- Review all suggestions before applying them to production code
- Combine with other complementary skills for comprehensive analysis

## Common Pitfalls

- Using this skill for tasks outside its domain expertise
- Applying recommendations without understanding your specific context
- Not providing enough project context for accurate analysis