Skip to content

Latest commit

 

History

246 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Angavu Intelligence Backend — Rust CFO Platform

Rust PostgreSQL Docker 35+ Tools Federated Learning License

Africa's Economic Nervous System. Rust-powered backend that transforms anonymized worker data into economic intelligence — 35+ tools, 7 superagent modules, 15 revenue engines, a federated learning pipeline, and a unified knowledge graph.

New in v0.3 (26+ Councils): SHAP explainability (EU AI Act), fairness testing, privacy budget (ε-DP), k-anonymity enforcement, data retention & right-to-erasure, OpenTelemetry distributed tracing, CI/CD pipeline, inequality tracker (Gini/Theil/Palma), model registry with A/B testing, classical algorithm implementations, response compression, security headers, JWT revocation, usage alerts, structured JSON logging.

Key claims (with citations):

  • Kenya's informal sector employs 83.4% of the workforce (Kenya National Bureau of Statistics, 2023 Economic Survey) [1]
  • Mobile money (M-Pesa) processed KES 35.5 trillion in 2023 (Central Bank of Kenya, 2023) [2]
  • k-Anonymity (k≥10) prevents re-identification in cohort-level data releases (Sweeney, 2002) [3]
  • Federated learning enables privacy-preserving model training across devices (McMahan et al., 2017) [4]
  • HNSW indexes achieve 95-99% recall@10 vs 85-95% for IVFFlat (Malkov & Yashunin, 2018) [5]
  • Brandes' algorithm computes betweenness centrality in O(VE) time (Brandes, 2001) [6]

Built by Angavu Intelligence Ltd. — Migori, Kenya


Architecture

Rust (Axum) = Primary backend — API, business logic, security, OODA loop
Python = LLM inference ONLY — DeepSeek, Qwen, XGBoost

OODAOrchestrator (Continuous Intelligence Loop)
├── 35+ Rust Tools (analysis, privacy, security, infrastructure, monitoring, credit explainability)
├── 7 Superagent Modules (flywheel, guardrails, intelligence, memory, sync, unified knowledge, service price discovery)
├── UnifiedKnowledgeLayer (in-memory + PostgreSQL + Harness graph bridge)
├── Billing Engine (4 tiers, API keys, invoicing, usage metering, M-Pesa)
├── REST API (Axum + WebSocket + GraphQL)
├── OpenTelemetry Distributed Tracing (26 instrumented spans)
├── CI/CD Pipeline (8-job GitHub Actions)
└── Structured JSON Logging (correlation IDs)

Tech Stack: Rust (Axum + Tokio) · PostgreSQL 16 + pgvector (HNSW) · ClickHouse 24 · Redis 7 · Docker · Nginx · OpenTelemetry


Tool Inventory — 35+ Tools + 7 Superagent Modules

Intelligence & Analysis

Tool Purpose
OODAOrchestrator Continuous Observe-Orient-Decide-Act intelligence loop
MarketAnalyzer Aggregated demand pattern analysis across worker cohorts (PostgreSQL-backed)
CreditScorer Alama Score (300–850) — credit scoring for informal economy
DemandForecaster Time-series demand forecasting for products and markets
EconomicAnalyzer Macro/micro economic indicator computation
FMCGIntelligence Fast-moving consumer goods market intelligence (PostgreSQL-backed)
CompositeIndexBuilder Multi-factor composite economic indices
AnomalyDetector Statistical anomaly detection in transaction streams
ScenarioModeler What-if scenario simulation and stress testing
PolicyImpactAnalyzer Government policy impact measurement
InequalityTracker Gini, Theil, Palma, D9/D1 ratio, median/mean income, trend analysis
ServicePriceDiscoveryEngine Service pricing intelligence from broadcast aggregation

Privacy & Security

Tool Purpose
DifferentialPrivacyEngine ε-differential privacy with Laplace mechanism (ε=0.1 default), privacy budget tracking, working noise injection API
KAnonymityEnforcer k≥10 anonymity cohort enforcement (MIN_K_ANONYMITY constant), wired into market_analysis + demand_forecast endpoints
FederatedAggregator Privacy-preserving FedProx gradient aggregation (μ=0.01 proximal term), Gaussian noise + gradient clipping + convergence monitoring
ModelDistributor Secure model distribution to edge devices
SHAPExplainer NEW: Shapley value computation for credit score explainability (EU AI Act compliance)
FairnessTester NEW: Demographic parity, equalized odds, predictive parity testing with Z-test
ClassicalCreditScorer NEW: IRLS logistic regression with Ridge regularization
SimulatedAnnealingOptimizer NEW: Quantum-inspired optimization for QUBO problems
DataRetentionEngine NEW: 10-category retention policies with right-to-erasure (Kenya DPA 2019)

Infrastructure

Tool Purpose
ApiGateway REST API gateway with auth, compression (gzip/brotli/deflate), and routing
CircuitBreaker Fault isolation and cascade failure prevention
RateLimiter Token-bucket rate limiting per API key
AuditLogger Immutable audit trail for all operations
SyncReceiver Device data reception and conflict resolution
SecurityHeaders NEW: HSTS, X-Content-Type-Options, X-Frame-Options, CSP, Referrer-Policy, Permissions-Policy
ModelRegistry NEW: Model versioning, A/B testing, algorithm tier selection
TelemetryModule NEW: OpenTelemetry spans (26 instrumented functions), structured JSON logging, correlation IDs
HealthChecker NEW: 3-tier health endpoints (liveness, readiness, detailed) with DB/Redis/ClickHouse checks

Monitoring & Reporting

Tool Purpose
HealthMetrics System health and performance metrics
ReportEngine Automated business and intelligence reports
AlertGenerator Threshold-based alerting and notification
WhatsAppSender WhatsApp Business API integration for alerts
MobileMoneySignalExtractor M-Pesa/mobile money transaction signal extraction

Superagent Modules

Module Purpose
FlywheelEngine Data network effects — more data → better models → more users
GuardrailsEngine Financial integrity checks, advice validation, compliance
IntelligenceEngine LLM-powered insight generation and reasoning
MemoryEngine 5-layer memory hierarchy (working → knowledge)
SyncEngine Bidirectional device-cloud synchronization
UnifiedKnowledgeLayer Bridges 3 disconnected knowledge graphs (in-memory, PostgreSQL, Harness) into single interface with cross-reference indices
ServicePriceDiscoveryEngine NEW: Aggregates service price broadcasts into market signals (k-anonymity enforced, PostgreSQL-backed)

Academic Formulas

The platform implements peer-reviewed statistical and econometric methods:

Formula Module Application
Bayesian Inference CreditScorer Prior belief updating with transaction evidence
Maximum Likelihood Estimation (MLE) CreditScorer Parameter calibration for score model
Gini Coefficient InequalityTracker Income/expenditure concentration measurement
Theil Index InequalityTracker Decomposable inequality (within/between groups)
Monte Carlo Simulation ScenarioModeler Risk scenario generation and stress testing
Ordinary Least Squares (OLS) PolicyImpactAnalyzer Causal effect estimation for policy interventions
Interrupted Time Series (ITS) PolicyImpactAnalyzer Pre/post policy impact measurement
Difference-in-Differences (DiD) PolicyImpactAnalyzer Quasi-experimental treatment effect estimation
Herfindahl-Hirschman Index (HHI) MarketAnalyzer Market concentration and competition analysis
Shapley Values (SHAP) SHAPExplainer Feature attribution for credit score explainability
Iteratively Reweighted Least Squares (IRLS) ClassicalCreditScorer Logistic regression training with Ridge regularization
Simulated Annealing SimulatedAnnealingOptimizer Quantum-inspired optimization for routing & matching
Gini Coefficient InequalityTracker Income/expenditure concentration measurement
Theil Index (GE(1)) InequalityTracker Decomposable inequality (within/between groups)
Palma Ratio InequalityTracker Top 10% / bottom 40% income share

Billing System

Tier Price Queries/mo Reports/mo Data Exports
Free $0 100 2
Starter $299/mo 5,000 20 5
Pro $1,499/mo 50,000 100 50
Enterprise Custom Unlimited Unlimited Unlimited

Features: API key management, usage metering, invoice generation, subscription lifecycle.


API Endpoints

Tools (/api/v1/tools)

Method Endpoint Description
GET /tools List all available tools
GET /tools/health System health metrics
POST /tools/credit Compute Alama credit score
GET /tools/market Market analysis
GET /tools/market/demand Demand forecast
POST /tools/economic Economic indicators
GET /tools/distribution Distribution gap analysis
GET /tools/fmcg FMCG intelligence report
POST /tools/privacy/noise Inject differential privacy noise
POST /tools/anonymize Apply k-anonymity
GET /tools/federated Federated learning status
GET /tools/sync Sync status
GET /tools/model Model distribution status
POST /tools/report Generate report
POST /tools/alert Generate alert
POST /tools/whatsapp Send WhatsApp message
GET /tools/gateway API gateway status
GET /tools/audit Audit log status
GET /tools/circuit-breaker Circuit breaker status
GET /tools/rate-limiter Rate limiter status

Billing (/api/v1/billing)

Method Endpoint Description
GET /billing/tiers List pricing tiers
POST /billing/subscriptions Create subscription
GET /billing/subscriptions/{org_id} Get subscription
PUT /billing/subscriptions/{id}/tier Change tier
POST /billing/subscriptions/{id}/cancel Cancel subscription
POST /billing/api-keys Create API key
GET /billing/api-keys/{org_id} List API keys
DELETE /billing/api-keys/{id} Revoke API key
GET /billing/usage/{org_id} Get usage metrics
GET /billing/invoices/{org_id} List invoices
GET /billing/invoices/detail/{id} Get invoice detail
POST /billing/invoices/{id}/finalize Finalize invoice
POST /billing/invoices/{id}/pay Pay invoice

Superagent (/api/v1/superagent)

Method Endpoint Description
GET /api/v1/superagent/status Orchestrator status
POST /api/v1/superagent/cycles Trigger OODA cycle
POST /api/v1/superagent/invocations Invoke superagent
POST /api/v1/superagent/alert/respond Respond to alert
GET /api/v1/superagent/alerts List alerts
GET /api/v1/superagent/history Cycle history

WebSocket

Endpoint Description
ws://host/ws Real-time data stream

15 Revenue Engines

Engine Buyer Price
Soko Pulse FMCG companies $2K–$12K/month
Alama Score Banks, MFIs $0.05–$0.50/query
Angavu Pulse Government Contact for pricing
+ 12 more Various Various

Infrastructure

  • Dockerfile — Multi-stage build (Rust builder → Python LLM → minimal runtime)
  • Dockerfile.oracle — ARM64-optimized for Oracle Free Tier (4 OCPU, 24GB RAM). No Python/ClickHouse/monitoring.
  • docker-compose.yml — Full stack: PostgreSQL 16, Redis 7, ClickHouse 24, Nginx, API
  • docker-compose.oracle.yml — Resource-constrained for Oracle Free Tier (PG 8GB, Redis 1GB, API 4GB)
  • docker-compose.production.yml — Production with WAL archiving, automated failover, replication
  • nginx/nginx.conf — Reverse proxy with rate limiting, SSL termination, WebSocket support
  • .github/workflows/deploy.yml — CI/CD: test → build → push GHCR → deploy to Oracle Cloud
  • OpenTelemetry — Distributed tracing for OODA loops and all critical paths (26 instrumented spans, OTLP exporter, Jaeger/Tempo compatible). Structured JSON logging with correlation IDs.

Oracle Free Tier Deployment

# SSH into Oracle ARM Ampere A1 instance
ssh ubuntu@your-oracle-ip

# Clone and configure
git clone https://github.com/ovalentine964/angavu-intelligence-backend.git
cd angavu-intelligence-backend
cp .env.oracle.example .env.oracle
nano .env.oracle  # set passwords and secrets

# Deploy
chmod +x scripts/deploy-oracle.sh
./scripts/deploy-oracle.sh

# Verify
./scripts/deploy-oracle.sh --status
curl http://localhost:8000/health

See scripts/deploy-oracle.sh for full deployment automation with health checks, backup rotation, and zero-downtime updates.


Quick Start

# Clone and configure
cp .env.example .env
# Edit .env with your API keys and secrets

# Start full stack
docker compose up -d

# Verify
curl http://localhost:8000/health

Documentation


Company

Angavu Intelligence Ltd. — Africa's Economic Nervous System


References

[1] Kenya National Bureau of Statistics. (2023). Economic Survey 2023. Nairobi: KNBS. The informal sector accounts for 83.4% of total employment.

[2] Central Bank of Kenya. (2023). Mobile Payments Statistics — Annual Report 2023. M-Pesa processed KES 35.5 trillion across 1.6 billion transactions.

[3] Sweeney, L. (2002). k-Anonymity: A Model for Protecting Privacy. International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems, 10(5), 557-570. https://doi.org/10.1142/S0218488502001648

[4] McMahan, B., Moore, E., Ramage, D., Hampson, S., & y Arcas, B. A. (2017). Communication-Efficient Learning of Deep Networks from Decentralized Data. Proceedings of the 20th International Conference on Artificial Intelligence and Statistics (AISTATS). arXiv:1602.05629

[5] Malkov, Y. A., & Yashunin, D. A. (2018). Efficient and Robust Approximate Nearest Neighbor Search Using Hierarchical Navigable Small World Graphs. IEEE Transactions on Pattern Analysis and Machine Intelligence, 42(4), 824-836. arXiv:1603.09320

[6] Brandes, U. (2001). A faster algorithm for betweenness centrality. Journal of Mathematical Sociology, 25(2), 163-177. https://doi.org/10.1080/0022250X.2001.9990249

[7] Dwork, C., & Roth, A. (2014). The Algorithmic Foundations of Differential Privacy. Foundations and Trends in Theoretical Computer Science, 9(3-4), 211-407. https://doi.org/10.1561/0400000042

[8] Kairouz, P., et al. (2021). Advances and Open Problems in Federated Learning. Foundations and Trends in Machine Learning, 14(1-2), 1-210. arXiv:1912.04977

[9] Brier, G. W. (1950). Verification of Forecasts Expressed in Terms of Probability. Monthly Weather Review, 78(1), 1-3. https://doi.org/10.1175/1520-0493(1950)078<0001:VOFEIT>2.0.CO;2

[10] Banerjee, A., Duflo, E., Glennerster, R., & Kinnan, C. (2015). The Miracle of Microfinance? Evidence from a Randomized Evaluation. American Economic Journal: Applied Economics, 7(1), 22-53. https://doi.org/10.1257/app.20130533

About

Angavu Intelligence Backend — Rust CFO platform. 20 superagent tools, federated learning, 15 revenue engines. By Angavu Intelligence Ltd.

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages