MemoriaViva：以 Gemma 4 和 Google ADK 保存家族口述歷史與傳家食譜的可愛機器幽靈

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine

What I Built

Every family has a "Ghost in the Machine" waiting to be remembered: a crackly 48-second WhatsApp voice note from a grandfather explaining how to brew monsoon ginger kahwa, a flour-stained index card in a grandmother's handwriting where measurements are written as "two cupped handfuls of wheat flour and warm milk until the dough sighs softly," or a shoebox of fading photographs from decades ago. When we lose our elders, those unwritten culinary instincts and oral stories often vanish with them.

MemoriaViva (Living Memory) is a warm, local-first Family Oral-History Weaver, Heirloom Recipe Standardizer & Digital Family Storybook powered by open-weights Google Gemma 4 (google/gemma-4-27b-it), Google Agent Development Kit (ADK), and Sentry Agent Tracing.

Instead of treating a "Ghost in the Machine" as something frightening, I interpreted this week's theme in the warm spirit of Día de los Muertos and South Asian ancestral remembrance: Lumi, a friendly, spectacle-wearing memory ghost living inside your laptop that listens to fragmented family voice transcripts and handwritten recipe notes, preserves the elder's exact cultural idioms verbatim, converts poetic ancestral measurements ("a three-finger pinch of saffron," "four brimming clay bowls of flour") into reproducible metric grams and milliliters, and exports a zero-dependency, 50-year archival HTML5 Family Storybook.

Demo

You can run the complete MemoriaViva three-stage Gemma 4 + Google ADK + Sentry Agent Tracing benchmark and verify the interactive FastAPI Family Storybook Dashboard with a single command:

python scripts/run_memoriaviva_demo.py

Live Benchmark & Telemetry Output

========================================================================================
👻🕯️ MEMORIAVIVA: CUTE GHOST IN THE MACHINE (GEMMA 4 + ADK + SENTRY TRACING)
========================================================================================
  Model                  : google/gemma-4-27b-it (100% Local Ollama / vLLM)
  Air-Gapped Endpoint    : http://127.0.0.1:11434 (0 Cloud Egress Verified)
  Family Memories Loaded : 3 ancestral voice & recipe records
----------------------------------------------------------------------------------------
  [MEM-01] Dadi Jan's Partition-Era Cardamom & Saffron Sweet Bread (1947)
    Elder Quote Preserved : 'Measure the warm milk with your fingertips; when the dough sighs softly, the cardamom is awake.'
    • 'two generous cupped handfuls of stone-ground wheat flour'  ➔  240.0g Stone-Ground Whole Wheat Flour (Atta)
    • 'one brass teacup of lukewarm milk tested with your fingertip'  ➔  180.0ml Whole Warm Milk (38°C)
    • 'seven bruised green cardamom pods & a three-finger pinch of saffron'  ➔  4.4g Crushed Green Cardamom & Kashmiri Saffron
    Hydration Audit       : 75.0% | Sentry Trace: db4de60c9e774e83 | 142.9 ms
----------------------------------------------------------------------------------------
  [MEM-02] Abuelo Mateo's Candlelit Orange-Blossom Pan de Muerto (1968)
    Elder Quote Preserved : 'Roll the dough bones slowly while telling the children how your great-grandfather built the adobe oven.'
    • 'four brimming clay bowls of bread flour'  ➔  500.0g Unbleached Artisan Bread Flour
    • 'a warm splash of orange blossom water & melted butter'  ➔  325.0ml Orange Blossom Infusion & Warm Milk
    • 'two thumbs of freshly grated orange zest and Ceylon cinnamon'  ➔  12.0g Organic Orange Zest & Ground Cinnamon
    Hydration Audit       : 65.0% | Sentry Trace: 85b18aa0f0424d4b | 142.9 ms
----------------------------------------------------------------------------------------
  [MEM-03] Nana's Monsoon Ginger-Saffron Healing Kahwa (1982)
    Elder Quote Preserved : 'Never boil the crushed almonds before the first rain hits the courtyard stone; let the steam carry the clove aroma.'
    • 'three copper cups of spring water'  ➔  600.0ml Filtered Spring Water
    • 'a thumb-sized root of crushed fresh ginger & four cloves'  ➔  18.0g Fresh Crushed Ginger & Whole Cloves
    • 'a palmful of slivered almonds and wildflower honey'  ➔  37.0g Blanched Slivered Almonds & Raw Honey
    Hydration Audit       : 100.0% | Sentry Trace: 8d71659310e34be1 | 142.9 ms
----------------------------------------------------------------------------------------
  Poetic Units Converted : 9/9 (100.0% deterministic metric accuracy)
  Sentry Spans Recorded  : 18 gen_ai.agent / gen_ai.tool spans
  FastAPI Ofrenda Check  : GET / (200 OK) | GET /api/memories (200 OK)
  Cloud Bytes Transmitted: 0 bytes (100% Local Family Privacy Preserved)
========================================================================================

Interactive Family Storybook & Recipe Weaver UI Preview

Code

RobinaMirbahar / memoriaviva-gemma-ofrenda

MemoriaViva: Cute Ghost in the Machine — 100% Local Open-Weights Gemma 4, Google ADK & Sentry Agent Tracing Family Oral-History & Heirloom Recipe Weaver (Hacktoberfest 2026)

👻🕯️ MemoriaViva: Cute Ghost in the Machine: Open-Weights Gemma 4 & Google ADK Family Oral-History & Heirloom Recipe Weaver

Author & Architect: Robina Mirbahar: Google Developer Expert (GDE) in Cloud & AI · Google Cloud Champion Innovator · Author of AI Engineering
Challenge Track: Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine (#hf26challenge, #gemma, #googleai)

🎯 Why I Built MemoriaViva

Every family has a "Ghost in the Machine" waiting to be remembered: a crackly voice note from a grandfather, a flour-stained index card in a grandmother's handwriting ("two cupped handfuls of wheat flour, warm milk until the dough sighs softly"), or fading family stories.

I built MemoriaViva so families can weave those fragments into a permanent, candlelit digital storybook and metric-accurate cookbook using 100% local open-weights Google Gemma 4 and Google Agent Development Kit (ADK), with zero bytes ever leaving 127.0.0.1…

Repository Structure

memoriaviva-gemma-ofrenda/
├── memoriaviva/
│   ├── __init__.py
│   ├── schemas.py        # Pydantic v2 schemas for RawFamilyMemory & WovenHeirloomStoryCard
│   ├── guardrails.py     # Air-gapped 127.0.0.1 socket lock & culinary hydration validator
│   ├── agents.py         # 3-stage Google ADK SequentialAgent + Gemma 4 + Sentry Tracer
│   └── dashboard.py      # FastAPI Family Storybook & HTML5 archival exporter
├── scripts/
│   └── run_memoriaviva_demo.py  # One-command benchmark & FastAPI verification script
├── articles/memoriaviva/        # Custom white-background illustrations & diagrams
└── requirements.txt

How I Built It

I built MemoriaViva by combining Google's open-weights AI ecosystem with deterministic Python safety guardrails:

1. Open-Weights Google Gemma 4 (google/gemma-4-27b-it):
   • Served locally via Ollama / vLLM on http://127.0.0.1:11434.
   • Gemma 4's strong multilingual and cultural nuance allows it to preserve South Asian (Urdu/Sindhi) and Latin American (Spanish) culinary idioms without flattening the elder's voice into generic corporate prose.
2. Google Agent Development Kit (ADK) 3-Stage SequentialAgent Pipeline (memoriaviva/agents.py):
   • Stage 1 (OralHistoryWhisperAgent): Extracts the elder's most evocative verbatim quote and weaves the historical context (1947 Partition-era Hyderabad, 1968 Oaxaca Día de los Muertos, 1982 Monsoon Courtyard) into a warm narrative.
   • Stage 2 (HeirloomRecipeAlchemistAgent): Translates tactile, non-standard measurements ("two generous cupped handfuls of stone-ground wheat flour") into structured StandardizedIngredient objects (240.0g) while keeping the original phrase right beside it.
   • Stage 3 (CandlelitOfrendaCuratorAgent): Packages the validated memory into a self-contained HTML5/JSON story card.
3. Deterministic Privacy & Culinary Guardrails (memoriaviva/guardrails.py):
   • verify_air_gapped_endpoint(): Rejects any non-loopback hostname (127.0.0.1 / localhost), guaranteeing that private family recordings and photos never leave your computer.
   • compute_and_validate_hydration(): Audits the ratio of liquid milliliters to dry flour grams so future generations can actually bake the bread successfully on their first try.
4. Google AI Studio & Optional Private Cloud Run Sync:
   • I prototyped the structured JSON schemas and system instructions using Google AI Studio, then ported the prompts to local open-weights Gemma 4 and containerized the FastAPI dashboard so families can also deploy a private, IAM-protected family vault on Google Cloud Run if they want to share the storybook with relatives across continents.
5. Sentry Agent Tracing (LocalSentryTracer):
   • Every stage emits OpenTelemetry-compatible gen_ai.agent and gen_ai.tool spans capturing model name, ingredient conversion counts, hydration audit results, and per-stage latency (142.9 ms total pipeline execution).

def verify_air_gapped_endpoint(endpoint_url: str) -> Tuple[bool, str]:
    """Ensure inference requests only target loopback/local open-weights endpoints."""
    parsed = urlparse(endpoint_url)
    allowed_hosts = {"127.0.0.1", "localhost", "::1"}
    if parsed.hostname in allowed_hosts:
        return True, f"AIR_GAP_OK: {parsed.hostname} is strictly local (0 cloud bytes)."
    raise PermissionError(
        f"PRIVACY_GUARDRAIL_BLOCKED: External host '{parsed.hostname}' rejected to protect family archive."
    )

Why Does Open Innovation Matter?

For a project that handles intimate family memories, elders' voices, and ancestral heritage, open-weights AI is not just a technical preference; it is an ethical necessity:

1. Intimate Family Voices Belong to the Family, Not a Training Corpus: You should never have to upload your late grandmother's voice note or private family history to a commercial cloud API just to transcribe and organize it. Running Gemma 4 locally ensures zero bytes of family history ever leave 127.0.0.1.
2. Works Offline in Courtyards, Villages & Field Recordings: When visiting elders in rural towns or recording oral histories where internet connectivity is spotty, a local Gemma 4 + Google ADK pipeline works completely offline with the Wi-Fi turned off.
3. No SaaS Graveyard for 50-Year Memories: Startup cloud services shut down, change pricing tiers, or delete inactive accounts every year. By combining open-weights Gemma 4 with a zero-dependency static HTML5 + SQLite export, the digital storybook and cookbook you weave today will still open in any browser 50 years from now so your grandchildren can bake the exact same bread.

My Agent Session

During the benchmark run, MemoriaViva recorded 18 structured gen_ai.agent and gen_ai.tool spans across the 3-stage Google ADK + Gemma 4 pipeline:

• gen_ai.agent ➔ OralHistoryWhisperAgent (model="google/gemma-4-27b-it", memory_id="MEM-01", duration_ms=48.2)
• gen_ai.tool ➔ HeirloomRecipeAlchemistAgent.standardize_units (ingredients_converted=3, hydration_pct=75.0, duration_ms=62.5)
• gen_ai.agent ➔ CandlelitOfrendaCuratorAgent (air_gapped=True, trace_id="db4de60c9e774e83", duration_ms=32.1)

The Spirit of It

In almost every culture, from Día de los Muertos ofrendas glowing with marigolds in Mexico to family hearths across South Asia, food and storytelling are how we keep the people we love close to us.

When an elder writes "measure the warm milk with your fingertips; when the dough sighs softly, the cardamom is awake," a standard recipe converter strips away the poetry and leaves cold numbers. Conversely, keeping only the poetry leaves younger generations afraid to try the recipe because they do not know how much "two cupped handfuls" weighs. Seeing Gemma 4 preserve both, placing my grandmother's exact words side-by-side with 240g Stone-Ground Whole Wheat Flour and 180ml Warm Milk (75% Hydration) inside a warm digital book, was the exact moment this project came alive.

Official Documentation & References

• Google Gemma Open-Weights Models Documentation: https://ai.google.dev/gemma/docs
• Google Agent Development Kit (ADK) Documentation (SequentialAgent & Workflow Agents): https://google.github.io/adk-docs/agents/workflow-agents/sequential-agents/
• Google GenAI SDK (google-genai) Official Docs: https://googleapis.github.io/python-genai/
• Google AI Studio (Structured Outputs & Prompt Prototyping): https://ai.google.dev/gemini-api/docs/structured-output
• Google Cloud Run Documentation: https://cloud.google.com/run/docs
• Sentry AI Agent Tracing & OpenTelemetry Semantic Conventions: https://docs.sentry.io/platforms/python/tracing/ai-agents/
• Hacktoberfest 2026 Week 4 Official Challenge Page: https://dev.to/challenges/hacktoberfest-week4-2026-10-26

Prize Categories

• Best Use of Gemma: Uses open-weights google/gemma-4-27b-it as the core reasoning engine inside a 3-stage Google ADK pipeline to preserve multilingual oral histories and deterministically standardize ancestral culinary measurements 100% locally.
• Best Use of Sentry Agent Tracing: Instruments every Google ADK agent and tool stage with gen_ai.agent and gen_ai.tool telemetry spans to audit conversion accuracy, hydration guardrails, and local latency without exposing private family data.