CryptoPrism On-chain Analytics

On-chain analytics is the gold standard of crypto intelligence. Metrics like active addresses, whale movements, NVT ratios, and exchange flows reveal what is actually happening on blockchains — information that price charts alone cannot provide.

The problem: access costs a fortune. Glassnode charges $800+/month for their professional tier. Nansen, Santiment, and IntoTheBlock charge similar premiums. For a bootstrapped project like CryptoPrism, these are non-starters.

But Google makes raw blockchain data freely available through BigQuery public datasets — every transaction, every block, every address, for 10+ chains. The data is there; the challenge is building the intelligence layer on top of it.

The system follows a five-stage pipeline: raw blockchain data flows from BigQuery through a FastAPI-based processing engine, gets stored in PostgreSQL, cached in Redis for fast reads, and served through 17 REST API endpoints.

On-Chain Score Formula

Every chain gets a composite score from 0 to 100, computed from four equally-weighted dimensions. This score feeds into the broader CryptoScore system used by the Saarthi AI advisor.

On-Chain Score (0–100): Network Activity (25%) — active address trend + tx count trend Whale Behavior (25%) — whale tx trend + exchange flow direction Holder Conviction (25%) — MVRV zone + CDD normalized Usage Growth (25%) — NVT ratio + DEX volume trend

Build Phases

Phase 1–2 (BTC Core): Infrastructure + active addresses, whale txs, volume. Phase 3–4 (BTC Advanced): MVRV, CDD, supply distribution, exchange flow. Phase 5–6 (ETH Suite): Gas, ERC-20, DEX volume, staking flows. Phase 7–8 (Multi-Chain): LTC, Polygon, Optimism, Tron. Phase 9–10 (Production): CryptoScore, middleware, 17 endpoints. All phases complete.

Six blockchains are live and returning real data from BigQuery. Every metric below was validated with actual queries during the build session — these are not estimates.

Pending Chains

The system implements 18 distinct on-chain metrics across two metric suites: BTC (UTXO-native) and ETH (EVM-native). Every value shown below is a real measurement from BigQuery, not a placeholder.

Bitcoin — Full Suite (10 Metrics)

Ethereum — Full Suite (8 Metrics)

The REST API is organized into four categories. Every endpoint returns JSON with standardized error handling, request logging, and Redis caching.

Health (1 endpoint)

GET /health — BigQuery + PostgreSQL + Redis status

On-Chain Metrics (13 endpoints)

GET /api/v1/onchain/{chain}/active-addresses — Daily unique addresses GET /api/v1/onchain/{chain}/whale-transactions — Large transfers GET /api/v1/onchain/{chain}/tx-volume — Tx count + volume GET /api/v1/onchain/{chain}/nvt — NVT ratio GET /api/v1/onchain/{chain}/exchange-flow — Exchange inflow/outflow GET /api/v1/onchain/{chain}/mvrv — MVRV ratio (UTXO chains) GET /api/v1/onchain/{chain}/supply-distribution — Holder cohorts (UTXO chains) GET /api/v1/onchain/{chain}/cdd — Coin Days Destroyed (UTXO) GET /api/v1/onchain/{chain}/gas-trends — Gas price trends (EVM chains) GET /api/v1/onchain/{chain}/token-transfers — ERC-20 top tokens (EVM chains) GET /api/v1/onchain/{chain}/dex-volume — DEX swap counts (EVM chains) GET /api/v1/onchain/{chain}/staking-flows — Beacon Chain deposits (ETH only) GET /api/v1/onchain/chains — Supported chains list

Scores (2 endpoints)

GET /api/v1/scores/onchain/{chain}/{token_id} — On-Chain Score (0–100) GET /api/v1/scores/context/{chain}/{token_id} — AI-ready context for Saarthi

Pipeline (2 endpoints)

POST /api/v1/pipeline/run/{chain} — Trigger single chain pipeline POST /api/v1/pipeline/run-all — Trigger all chains

Every BigQuery query was metered during the build session. The total cost to validate all metrics across all chains was $12 for ~1,600 GB scanned. Below is the full breakdown.

Session Query Costs (Actual)

Projection: 90-Day Backfill

Backfill estimate: $67.07 for 10.7 TB. Full 90-day historical backfill across all 8 chains (including estimated Arbitrum and Avalanche). BTC dominates at $45.93 due to community dataset full-table scans. ETH + all EVM chains combined cost only $2.96.

Monthly ongoing: $30/month (~$1/day). Daily scheduled pipeline scanning 1-day windows. BTC: ~$0.75/day, ETH + EVM chains: ~$0.05/day, LTC: ~$0.20/day. Well within $1,000 GCP credits budget.

Budget Analysis

The entire system was built in a single session. Here is the quantitative breakdown of what was produced.

Build Output by Phase

P1-2 BTC Core: 18 files, 3 metrics. P3-4 BTC Advanced: 14 files, 5 metrics. P5-6 ETH Suite: 16 files, 6 metrics. P7-8 Multi-Chain: 12 files, 0 additional metrics. P9 CryptoScore: 10 files, 1 metric. P10 Production: 16 files.

Module Breakdown

UTXO Metrics (6 modules): active_addresses, whale_txs, tx_volume, mvrv, cdd, supply_distribution.

EVM Metrics (4 modules): gas_trends, token_transfers, dex_volume, staking_flows.

Shared Metrics (2 modules): nvt, exchange_flow (work across both UTXO and EVM chains).

Chain Configs (10 chains): BTC, ETH, LTC, DOGE, Polygon, Arbitrum, Optimism, Avalanche, Tron, Solana. Each config maps dataset names, table schemas, and supported metric types.

Exchange Labels (39 labeled addresses): 19 BTC + 20 ETH addresses across 10 exchanges (Binance, Coinbase, Kraken, Bitfinex, Gemini, Huobi, OKX, Bitstamp, KuCoin, Gate.io).

Every component was chosen for cost-effectiveness and operational simplicity. The entire system runs on Google Cloud Platform within the social-data-pipeline-and-push project.

Infrastructure Components

Data Source — Google BigQuery: Public blockchain datasets for 10 chains. Pay-per-query pricing at $6.25/TB. Google-managed datasets use efficient columnar storage for sub-cent queries.

Database — PostgreSQL (Cloud SQL): Persistent storage at 34.55.195.199. Stores computed metrics, exchange labels, chain configs, and score history for the CryptoScore system.

Cache — Redis: Get-or-compute pattern: check cache first, query BigQuery only on miss. TTL-based expiration aligned with pipeline schedule (daily refresh).

API Framework — FastAPI + Uvicorn: Async Python API with automatic OpenAPI docs, request validation via Pydantic, structured logging middleware, and health checks.

Deployment — Cloud Run + Docker: Containerized deployment with auto-scaling. CI/CD via GitHub Actions triggers on push to main. Cloud Scheduler for daily pipeline runs at 06:00 UTC.

Testing — pytest (20 tests): Unit tests for all metric modules, API endpoints, score calculations, and chain config validation. All passing as of build completion.

The project delivered a production-ready on-chain analytics backend in a single build session, covering 6 live chains with 18 metrics and 17 API endpoints — at a fraction of the cost of commercial alternatives.

Next Steps