Skip to main content
Mind Layer

AI Grounded in Real Data

Integrate your domain data once, and your AI employees reference it forever. Per-user facts, preferences, and inventory are tracked automatically from conversations — no manual logging. Your AI stays grounded in real data.

Mind Layer · Memory
User fact timeline
6 facts
Session · 2 days ago
Feeling burned out from current work
emotionwellbeing
91%
Restored by a hike in nature last weekend
experiencerecovery
94%
Current role misaligned with growth expectations
beliefcareer
87%
Session · last week
Prefers work without constant digital interruption
preferenceenvironment
82%
Processes difficult feelings through physical activity
behaviorcoping
89%
Values autonomy and physical presence in work
identityvalues
78%

Knowledge Base

Structured Knowledge Your AI Can Reference

Power your AI employees with grounded, real-time data for accurate and reliable interactions. Upload documents or insert entities into a knowledge graph, and your AI references it in every conversation automatically — citing real data instead of guessing.

Real Estate

Property listings, prices, sizes, and availability — all queryable by your AI. Help buyers find their perfect home instantly.

Personal Assistant

Track collectible values, personal inventories, and niche data like Pokémon card prices — your AI becomes a domain expert.

E-Commerce

Product catalogs, live pricing, and stock levels fed directly to your AI. Customers get accurate answers, not guesses.

Raw Data Sources

PDFs & DocsDatabasesAPIs & Tools

Structured Knowledge Graph

EntitiesRelationshipsFacts

AI Agent

Real-time UpdatesAudit Trails

Value Proposition

Zero Hallucinations

Ground your AI in verified data. Eliminate factual errors and unreliable responses by referencing structured sources.

Real-time Updates

Keep knowledge current automatically. Connect to live data sources to ensure your AI always has the latest information.

Audit Trails

Track every AI response to its source. Provide transparency and accountability with detailed citation logs.

Code Example

// Insert domain knowledge as entities
client.Knowledge.InsertFacts(ctx, projectID,
    sonzai.InsertFactsParams{
        Entities: []sonzai.Entity{
            {
                Label: "Espresso",
                Type:  "product",
                Properties: map[string]any{
                    "price": 4.50, "category": "coffee",
                    "origin": "Italy",
                },
            },
        },
    },
)

// Semantic + full-text search across the knowledge base
results, _ := client.Knowledge.Search(ctx, projectID,
    sonzai.KBSearchParams{
        Query: "Italian coffee drinks under $5",
    },
)
Knowledge Base · Graph view
Real estate domain
resolving
Query
“Got any 2-bedrooms near good schools in Pac Heights?”
interested_incontainscontainslistedlistedreferenced_byasked_aboutPriyalookingPac Heightsneighborhood2BR · $1.4Mlisting3BR · $2.1MlistingMaya K.agentSchools.pdfdoc
Answer — grounded, cited
Yes — 2BR at $1.4Min Pac Heights, and here's the latest on Rooftop Elementary from the schools briefing.

Per-User Facts

Accurate Facts for Every Individual

Every conversation automatically builds a personal fact profile for each user — preferences, history, relationships, and interests — all extracted, scored, and categorized without any manual work. Query what the AI has learned, or just let it use these facts to personalize future conversations on its own.

// Facts are extracted and verified automatically from conversations.
// Query what the AI knows about any user:
facts, _ := client.Agents.Memory.ListFacts(ctx, agentID,
    sonzai.ListFactsParams{
        UserID:   "user_123",
        Category: "preferences",
    },
)
// → [{fact: "Allergic to shellfish", importance: 0.95}, ...]

// Search across everything the AI has learned
results, _ := client.Agents.Memory.Search(ctx, agentID,
    sonzai.MemorySearchParams{
        Query:  "dietary restrictions",
        UserID: "user_123",
    },
)

// Browse the timeline of what was learned and when
timeline, _ := client.Agents.Memory.Timeline(ctx, agentID,
    sonzai.TimelineParams{
        UserID: "user_123",
        From:   "2025-01-01",
        To:     "2025-03-01",
    },
)
Knowledge · Per-user facts
user_8f2c · Priya R.
6 facts
P
Priya Ramanathanverified
14 sessions · first seen 3 months ago · last seen 3 days ago
Dietary98
Allergic to shellfish · EpiPen carrier
Session 3
Dietary82
Vegetarian Mon/Wed, otherwise pescatarian
Session 7
Preferences74
Prefers window seats on flights
Session 4
Preferences81
Morning person · no calls before 9am
Session 11
History68
Ran the Berlin marathon in 2024
Session 9
Work90
Principal engineer at a fintech startup
Session 1

Inventory

Track Items & Assets Per User

Inventory tracking works out of the box when enabled — your AI automatically manages items mentioned in conversations. You can also add items programmatically, batch import up to 1,000 at once, or query with aggregations. Items auto-resolve against your knowledge base for market valuations.

// Add items to a user's inventory
client.Agents.Inventory.Update(ctx, agentID, "user_123",
    sonzai.InventoryUpdateOptions{
        Action: "add",
        Items: []sonzai.InventoryWriteItem{
            {Name: "Diamond Ring", Quantity: 1,
             Properties: map[string]any{
                "material": "platinum", "carat": 2.5,
            }},
        },
    },
)

// Query with aggregation
summary, _ := client.Agents.Inventory.Query(ctx, agentID,
    "user_123",
    sonzai.InventoryQueryOptions{
        Mode:  "aggregate",
        Query: "total value of jewelry",
    },
)
Knowledge · Inventory ledger
user_8f2c · trading cards
auto-resolving
1
Charizard · 1st ed. holomentioned
PSA 9 · shadowless
$18,400
1
Blastoise · 1st ed. holomentioned
PSA 8 · no crease
$6,200
1
Venusaur · 1st ed. holoimported
PSA 8 · centered
$4,800
1
Pikachu Illustratormentioned
unverified
2
Base set booster boximported
sealed · wotc
$52,000
× 2
3
Gengar VMAX · alt artadded
NM · Japanese
$1,650
× 3
Auto-resolved total
6 items · valued against live knowledge base
$138,350

Analytics

Built-In Recommendations & Trends

Define recommendation and trend rules once, and the system continuously computes scores from interaction data. Your AI employees surface personalized recommendations in conversations automatically. Query trends, track conversions, and feed back results to improve accuracy over time.

// Get personalized recommendations for a user
recs, _ := client.Knowledge.GetRecommendations(ctx, projectID,
    sonzai.RecommendationParams{
        UserID:  "user_123",
        AgentID: agentID,
        Limit:   5,
    },
)

// Track trending items over a time window
trends, _ := client.Knowledge.GetTrends(ctx, projectID,
    sonzai.TrendParams{Window: "7d"},
)

// Record feedback to improve future recommendations
client.Knowledge.RecordFeedback(ctx, projectID,
    sonzai.FeedbackParams{
        UserID: "user_123",
        NodeID: recs[0].NodeID,
        Action: "clicked",
    },
)
Analytics · Recs & trends
Podcast companion agent
learning
For user_8f2c
Stoic Coffee #417 — on discipline94
because listens 3× a week
How I Built This · Patagonia87
because ESG interest
Huberman Lab · sleep protocol81
because asked about jet lag
Lex Fridman · Andrej Karpathy76
because building an LLM
Trending · 7d window
The Diary of a CEO+38%
1,824 hits
Dwarkesh Patel+27%
1,392 hits
Acquired · NVIDIA Pt 3+19%
981 hits
Founders · Charlie Munger+12%
702 hits
Feedback loop
surface
user acts
score ↑↓
re-rank