Technical Stack & Integration Guides

Silsilat is a modular, agent-based architecture combining AI, DLT, and liquidity automation. The stack is designed to be interoperable, composable, and secure, allowing easy integration by pawnshops, liquidity providers, regulators, and third-party applications.

---

1. System Architecture Overview

Key Layers

Layer	Purpose	Technology Stack
Frontend	User dashboards for pawnshops, investors, and regulators	Next.js / React / TailwindCSS
API Gateway	Unified access layer for REST & WebSocket endpoints	FastAPI (Python)
Agent Layer	Independent AI and policy agents running evaluation logic	Python 3.11 / OpenTelemetry / Arize Phoenix
Consensus Layer	Immutable event ledger and state synchronization	Hedera SDK / HCS-10 / HCS-11
Data Layer	Artifact storage and model traces	Postgres, IPFS / Arize Phoenix local collector
Liquidity Layer	On-chain fund management and yield distribution	Hedera SDK
Security Layer	Authentication, encryption, and signatures	JWT / HMAC / Ed25519

---

2. Development Environment Setup

Prerequisites

• Python 3.11+

• Node.js 20+

• Docker & Docker Compose

• Hedera SDK (Python + JS)

• Pinata API

• Arize Phoenix Local Instance

• .env.local file with required keys

Example .env.local

HEDERA_OPERATOR_ID=0.0.987654
HEDERA_OPERATOR_KEY=302e020100300506032b657004220420abc...
METALPRICE_API_KEY=mpa_xxxxxxx
FASTFOREX_API_KEY=ffx_xxxxxxx
PHOENIX_COLLECTOR_ENDPOINT=http://localhost:6006/v1/traces
HCS_INPUT_TOPIC_ID=0.0.1001
HCS_OUTPUT_TOPIC_ID=0.0.1002
HCS_OVERRIDE_TOPIC_ID=0.0.1003
IPFS_GATEWAY_URL=https://ipfs.io/ipfs/

Run agents locally using Docker:

docker-compose up --build

---

3. Core Services

Gold Evaluator Agent

Evaluates LTV ratios, applies haircut policies, and produces valuation artifacts.

python3 agents/gold_evaluator.py

Key Methods:

• fetch_gold_price() → calls MetalPriceAPI

• convert_fx() → calls FastForex

• apply_policy() → loads haircut and LTV limits

• publish_trace() → sends artifact CID to HCS

Output Example:

{
  "trace_id": "3f78a4b2-1c4d-4fd2-9c1b-98ab3a2e32c1",
  "loan_value_myr": 7820.45,
  "ltv": 0.78,
  "haircut": 0.12,
  "policy_id": "AR_RAHNU_STD_2025",
  "ipfs_cid": "bafybeigoldtracehash"
}

---

Liquidity Node

Executes pool operations via Hbarsuite Smart Node logic.

python3 nodes/liquidity_agent.py

Functions:

• buy_sag_token()

• mint_lqt_token()

• redeem_lqt_token()

• calculate_yield()

• publish_transaction() (to HCS)

Smart node state transitions are deterministic and replicated across all liquidity node validators.

---

Policy Agent

Monitors transactions and applies KYC/AML/Shariah compliance logic.

python3 agents/policy_agent.py

Data Sources:

• AML/KYC screening API (WorldCheck / Dow Jones)

• Silsilat Policy JSON schema

• Regulator node public key list

Compliance Outputs:

{
  "trace_id": "c1a76f81-77da-49a7-98a1-fb5a123da9c7",
  "kyc_status": "verified",
  "aml_screening": "clear",
  "shariah_result": "compliant",
  "result": "pass"
}

---

4. APIs for Integration Partners

Base URL

https://api.silsilat.io/v1/

Endpoints

Method	Endpoint	Description
POST	/pawnshops/register	Onboard pawnshop node with KYC
POST	/sag/mint	Create new Secured Asset Gold token
GET	/sag/{sag_id}	Retrieve SAG metadata
POST	/evaluate	Trigger gold evaluation
GET	/evaluation/{trace_id}	Fetch AI trace result
POST	/liquidity/request	Pawnshop requests funding
POST	/liquidity/deposit	Investor deposits funds
GET	/liquidity/pool	Query pool metrics
POST	/override/submit	Submit admin override decision

All endpoints require JWT tokens issued via /auth/login. Payloads are signed using HMAC-SHA256.

---

5. SDKs and Developer Tools

Python SDK

from silsilat import Client

client = Client(api_key="your_api_key")

# Mint SAG
sag = client.mint_sag(
    pawnshop_id="ARHNU-001",
    gold_weight=22.5,
    purity=0.916,
    appraisal_value=9500
)

# Request liquidity
response = client.request_liquidity(sag_id=sag["sag_id"], amount=7800)

JavaScript SDK

import { SilsilatClient } from 'silsilat-sdk';

const client = new SilsilatClient({ apiKey: 'your_api_key' });

const evaluation = await client.evaluateGold({
  weight: 15.2,
  purity: 0.916,
  currency: 'MYR',
});

---

6. Topic Configuration for Developers

Topic	Purpose	Example ID
Input Topic	Loan initiation and pawnshop messages	0.0.1001
Output Topic	AI evaluator and compliance responses	0.0.1002
Override Topic	Admin or regulator overrides	0.0.1003
Policy Topic	Policy updates and version control	0.0.1004
Treasury Topic	Fee and yield settlement records	0.0.1005

Message Format Example:

{
  "event": "loan_evaluation",
  "trace_id": "12345",
  "actor": "GoldEvaluatorAgent",
  "data": {
    "ltv_ratio": 0.78,
    "haircut": 0.12,
    "loan_value_myr": 7800
  },
  "timestamp": "2025-10-26T07:25:00Z"
}

---

7. Integration Workflow

(a) For Pawnshops

1. Register via /pawnshops/register

2. Submit pledged gold metadata → /sag/mint

3. Trigger AI evaluation → /evaluate

4. Receive LTV → /evaluation/{trace_id}

5. Request liquidity → /liquidity/request

6. Receive instant payout to bank / wallet

(b) For Investors

1. Register investor account

2. Deposit funds into liquidity pool

3. Receive LQT tokens representing pool share

4. Track yield via /liquidity/pool

5. Redeem principal + yield after maturity

(c) For Regulators

1. Access audit dashboard (read-only HCS node)

2. Query transactions by trace_id or policy_id

3. Review override events

4. Validate compliance packs (via /policy/validate)

---

8. Integration with External Systems

Integration Type	Partner Example	Purpose
KYC / AML	Refinitiv / WorldCheck	Automated screening
Gold Pricing	MetalPriceAPI	Real-time commodity data
FX Conversion	FastForex	USD → MYR rate conversion
Storage Verification	IoT Sensors via MQTT	Vault monitoring data
Payment Rails	PayNet / Bank Negara Open API	Liquidity settlement
Phoenix Observability	Arize Phoenix	Trace monitoring for AI explainability

---

9. Security & Privacy Architecture

Security Layer	Mechanism
Encryption in Transit	TLS 1.3
Encryption at Rest	AES-256
Identity & Roles	DID registry (HCS-11) + JWT
Payload Integrity	HMAC-SHA256 signatures
Data Privacy	IPFS pointers only, raw data off-chain
ZKP Ready	Supports Zero-Knowledge Attestation for private KYC proofs

Example: ZKP Verification Stub

zkp = generate_proof(identity_hash, kyc_passed=True)
verify_proof(zkp, public_key)

---

10. Deployment Options

Environment	Purpose	Hosting Option
Local Dev	Testing with mock APIs	Docker Compose
Testnet	Integration testing	Hedera Testnet
Production	Live transaction network	Hedera Mainnet
Private Instance	Regulated node deployment	HSuite Smart Node Cluster

Deployment is CI/CD-ready via GitHub Actions and can be containerized for Kubernetes or Hbarsuite orchestration.

---

Summary

Silsilat’s technical stack is designed to be:

• Composable: Modular agent-based components that can be replaced or extended.

• Observable: Every decision logged through Arize Phoenix and IPFS.

• Secure: Multi-layered authentication, encryption, and consensus anchoring.

• Interoperable: REST, SDKs, and GraphQL APIs for easy external integration.

Together, this stack enables a trusted, auditable bridge between real-world collateral and digital liquidity, ready for institutional adoption.