Skip to content

INDB: The Epistemological Engine

Inhale. Exhale. Axiom.

"When everyone lies, INDB remembers what you saw."

Version: 0.7.0
Status: ✅ Production-Ready
Last Updated: March 4, 2026

🎯 What is INDB?

INDB is not a database. It is an Epistemological Engine — a system designed to preserve truth in the age of manipulation.

The Problem: How do you maintain objective reality when collective gaslighting can rewrite history?

INDB's Solution: Treat data as a biological process — metabolism, not storage. Modeled after human memory: inhale, exhale, axiom, hidden zones.

🫁 The Core Philosophy

Inhale → Exhale → Axiom

1. INHALE (Ingestion)

Accept reality as it is: raw, chaotic, unfiltered.

  • Multi-protocol intake (UDP, gRPC, HTTP, WebSocket)
  • No judgment, only observation
  • Performance: <1ms latency, 1000+ events/sec

2. EXHALE (Fusion & Decay)

Compress truth. Purge noise. Reputation is gravity.

Mass = Count × Reputation²
  • 1000 lies from bots (rep 0.01) → Mass = 0.1
  • 1 truth from sensor (rep 0.99) → Mass = 0.98

The system "exhales" the lies as waste heat.

3. AXIOM (Irrefutable Core)

What remains is cryptographically signed, immutable truth.

🧠 Intelligence Modules

Ways of recalling: Prism interprets, Echo resonates, Instinct reflexes, Deduction chains — Blind is what we store but never read.

Prism — Paradox-aware Synthesis

Transforms raw events into meaning. Returns not a single answer — returns its confidence and all competing readings.

curl -X POST http://localhost:8000/api/v2/prism/synthesize \
  -H "Content-Type: application/json" \
  -d '{
    "query": "would have died",
    "observer_context": {
      "token_weights": {"silence": 2.0},
      "vocabulary": {"Subjective": ["scream", "felt"]},
      "harmonic_weights": {"emotion": 0.8}
    }
  }'

Response includes the Perception Paradox:

Field Meaning
meaning Primary reading
alternative_readings Competing readings — preserved, not discarded
perception_gap 0.0 = clear · 1.0 = maximum ambiguity
is_paradoxical True when gap ≥ 0.25 — the system cannot decide
has_unread_essence True when blind_payload present — reality may differ

What you see may not be what it is.

Status: ✅ Production (100% success, ~20ms latency)

Echo — Resonance Detection

Finds similar events through harmonic analysis.

Harmonic Weights: - Token similarity: 20% - Emotion similarity: 30% - Meta (location): 50%

Status: ✅ Production (<30ms latency)

Instinct — Adrenaline Reflexes

Adaptive response based on urgency.

Three Modes: - Analytical (0.0): Deep analysis, ~10% confidence - Alert (0.5): Balanced, ~46% confidence - Instinctive (1.0): Meta-only, ~82% confidence

Reflex Triggers: - Location Match: +0.7 - Owner Signature: +0.5 - Critical Tokens: +0.3

Status: ✅ Production (100% success, ~25ms latency)

Deduction — Sherlock Holmes Thinking

Infers conclusions from evidence chains. Given facts (events), what can we deduce?

Pipeline: 1. Resolve seeds (event IDs or text query) 2. Expand via Echo resonance → evidence cloud 3. Build chain (temporal, spatial, token inference) 4. Synthesize conclusion

curl -X POST http://localhost:8000/api/v2/deduction/deduce \
  -H "Content-Type: application/json" \
  -d '{"query": "birthday cheese"}'

Or with explicit seeds: 

curl -X POST http://localhost:8000/api/v2/deduction/deduce \
  -H "Content-Type: application/json" \
  -d '{"seed_event_ids": ["evt-1", "evt-2"]}'

Status: ✅ Production

🏗️ Architecture

Distributed Raft Cluster

graph TD
    Client[Client] -->|HTTP/gRPC/UDP| LB[HAProxy :8080]
    LB --> Node1[Node 1 :8000<br/>Leader]
    LB --> Node2[Node 2 :8004<br/>Follower]
    LB --> Node3[Node 3 :8005<br/>Follower]

    Node1 <-.->|Raft Consensus| Node2
    Node2 <-.->|Raft Consensus| Node3
    Node3 <-.->|Raft Consensus| Node1

    Node1 --> Data1[(data-node-1/)]
    Node2 --> Data2[(data-node-2/)]
    Node3 --> Data3[(data-node-3/)]

Features: - ✅ Automatic leader election - ✅ Data replication across all nodes - ✅ Synchronous commits - ✅ Fault tolerance (survives single node failures) - ✅ Stable: 4800% improvement (zero re-elections over 5+ minutes)

📊 Verified Performance

Cluster Performance (3-Node Raft)

Metric Value
Throughput 19-28 events/sec sustained
Success Rate 99.7% (997/1000 events)
Latency 50-250ms per request
Data Persistence 100% (survives restarts)
Stability Zero re-elections (5+ min)

Module Performance

Module Success Rate Latency Status
Prism 100% (50/50) ~20ms ✅ Production
Echo 100% <30ms ✅ Production
Instinct 100% (50/50) ~25ms ✅ Production
Deduction 100% <50ms ✅ Production
Fusion Real-time N/A ✅ Production

Standalone Performance

  • Ingestion: <1ms per event
  • Query: <10ms for 1000 events
  • Fusion: Real-time semantic deduplication

🚀 Quick Start

# Start 3-node cluster
docker-compose up -d

# Wait for initialization
sleep 20

# Verify cluster
curl http://localhost:8080/raft/status

# Ingest event
curl -X POST http://localhost:8080/api/v2/events \
  -H "Content-Type: application/json" \
  -d '{
    "raw_data_anchor": ["test", "event"],
    "location": "test/demo",
    "ttl": 86400
  }'

Standalone Mode

# Start single node
docker-compose -f docker-compose.standalone.yml up -d

# Access at http://localhost:8000

🔌 Protocol Stack

Protocol Port Type Use Case
HTTP/REST 8000 Interface Standard API, queries
gRPC 50051 Stream Binary streams, service-to-service
UDP 9001 Inhale Fire-and-forget IoT ingestion
WebSocket 8001 Live Real-time event broadcasting
GraphQL 8000 Flexible Schema introspection
MCP 8000 AI Model Context Protocol

🛠 Key Features

✅ Distributed Consensus (Raft)

  • 3-node cluster with automatic leader election
  • Data replication across all nodes
  • Synchronous commits
  • Fault tolerance
  • Recent Fix: 4800% stability improvement

🔍 Contextual Lens

  • Adjustable spectrum ratios (10/90, 30/70, 50/50, 70/30, 90/10)
  • Dynamic context blending (recent/historical, simple/complex)
  • Cognitive zoom (child vs adult mode)
  • Zero duplication (virtual "Third Half")

🧠 Neural Fusion Engine

  • Semantic deduplication
  • Adaptive frequency tracking
  • Penalty system (prevents fusion spam)
  • Temporal awareness

🔐 Security

  • Encryption: AES-256-GCM (at rest), TLS/SSL (in transit)
  • Signatures: Ed25519
  • Auth: API keys + RBAC

📚 API Examples

Prism (Paradox-aware Synthesis)

# With observer context
curl -X POST http://localhost:8000/api/v2/prism/synthesize \
  -H "Content-Type: application/json" \
  -d '{
    "query": "6dede7d7-82fb-41cc-95fe-8c534f82ad4e",
    "observer_context": {
      "token_weights": {"pain": 3.0},
      "vocabulary": {"Personal": ["felt", "almost"]},
      "harmonic_weights": {"emotion": 0.9}
    }
  }'

# Without observer context (system defaults apply)
curl -X POST http://localhost:8000/api/v2/prism/synthesize \
  -H "Content-Type: application/json" \
  -d '{"query": "amsterdam bicycle evening"}'

Instinct (Adrenaline Reflexes)

curl -X POST http://localhost:8000/api/v2/prism/instinct \
  -H "Content-Type: application/json" \
  -d '{
    "seed_event_id": "6dede7d7-82fb-41cc-95fe-8c534f82ad4e",
    "adrenaline": 0.7,
    "cloud_size": 10
  }'

Echo (Resonance Detection)

curl -X POST http://localhost:8000/api/v2/echo/resonate \
  -H "Content-Type: application/json" \
  -d '{
    "seed_event_id": "6dede7d7-82fb-41cc-95fe-8c534f82ad4e",
    "cloud_size": 10
  }'

Contextual Lens (Adaptive Queries)

# Balanced view (50/50)
curl -X POST http://localhost:8000/api/v2/lens/query \
  -H "Content-Type: application/json" \
  -d '{
    "context_a": "recent",
    "context_b": "historical",
    "ratio": [0.5, 0.5],
    "mode": "sharp",
    "limit": 20
  }'

🎯 Use Cases

1. Reality Guard ("Truth Mirror")

Coordinated attack claims server is down.

  • Crowd: 10k messages "Down!" (Rep 0.01) → Mass 10
  • Axiom: 1 heartbeat "Up." (Rep 0.99) → Mass 0.98
  • Verdict: Data Conflict. Axiom stands.

2. Digital Ghost ("Black Box")

Autonomous agent records life history.

  • Routine logs fuse into "Habit" events
  • Critical anomalies remain distinct Axioms
  • Memory remembers meaning, not just files

3. Literary Analysis

Analyze themes across Dostoevsky's works.

  1. Ingest all books as events
  2. Prism finds significant passages
  3. Echo finds thematically similar content
  4. Instinct for rapid pattern matching

4. Threat Detection

Monitor system for anomalies.

  1. Ingest logs in real-time
  2. Fusion compresses normal patterns
  3. Instinct (adrenaline=1.0) triggers on anomalies
  4. Prism analyzes significance

📖 Interactive Documentation

API Documentation

MkDocs Site

# Start documentation server
docker-compose up -d docs

# View at http://localhost:8008

🧪 Testing

# Run all tests
pytest

# Run specific suites
pytest tests/unit/
pytest tests/integration/
pytest tests/e2e/

# With coverage
pytest --cov=core --cov-report=html

🏆 Recent Achievements (March 2026 · v0.7.0)

Paradox of Perception ✅

  • Prism no longer resolves contradictions — it surfaces them
  • perception_gap, is_paradoxical, alternative_readings, has_unread_essence
  • What you see may not be what it is

Observer Context ✅

  • Meaning is now observer-dependent, not hardcoded
  • Pass token_weights, vocabulary, harmonic_weights to Prism
  • Same event → different readings for different observers

Signed Memory ✅

  • Every API response signed with Ed25519
  • System cannot disown what it said — mathematically
  • Stable key via SIGNING_PRIVATE_KEY env variable

Black Box Contract ✅

  • blind_payload validated at protocol level
  • location + timestamp mandatory when blind_payload is set
  • No backdoor: engine has no key — architecturally, not by policy

Zero Hardcoded Values ✅

  • All thresholds and labels extracted to core/constants.py
  • Observer context replaces fixed context_label defaults

📐 Conceptual Alignment

See docs/CONCEPTUAL_ALIGNMENT.md for the integrity check: Inhale-Exhale-Axiom mapping, Hermes vs REST flow, constants-driven design, Interpret vs Lens.

🗺️ Roadmap

Q1 2026 (Current)

  • ✅ Raft stability fixes
  • ✅ Prism/Echo/Instinct modules
  • ✅ Production deployment
  • 🔄 Performance optimization
  • 🔄 Monitoring dashboard

Q2 2026

  • Advanced fusion patterns
  • ML-based pattern detection
  • Multi-region replication
  • Real-time analytics

Q3 2026

  • Blockchain integration (optional)
  • Advanced encryption modes
  • Federated learning
  • Edge deployment

💡 Core Principles

1. Truth Through Metabolism

Data is not static. It lives, breathes, fuses, and decays.

2. Reputation as Gravity

Not all data is equal. Trust shapes reality.

3. Axioms Resist Gaslighting

What is signed and witnessed cannot be rewritten.

📝 License

Proprietary - All Rights Reserved

🏆 Summary

INDB is: - ✅ A distributed epistemological engine - ✅ Production-ready (verified under load) - ✅ Semantically intelligent (Prism/Echo/Instinct) - ✅ Fault-tolerant (Raft consensus) - ✅ Secure (encryption + signatures) - ✅ Fast (<1ms ingestion, <30ms queries)

INDB solves: - Information overload (Fusion) - Collective gaslighting (Axioms) - Context loss (Prism) - Pattern blindness (Echo) - Slow response (Instinct)

Built with ❤️ for Truth in the Age of Manipulation

"Inhale the chaos. Exhale the noise. Remember the Axiom."