dykij/dmarket_bot

Purpose and entry points

DMarket Quantitative Engine is an autonomous Python trading system for DMarket marketplace items. It scans order-book-derived demand signals, evaluates candidates, executes purchases/resales (or dry runs), and manages positions. Primary runtime entry point is src/__main__.py, which initializes and runs the application via src/core/application.py; operational startup is also supported by scripts/start.sh and Docker definitions.

Application lifecycle and orchestration

The core application owns configuration, startup/shutdown, recovery, signals, and notifications. It composes long-running scanner and trading services rather than exposing an HTTP application interface.

• src/core/application.py

• src/core/app_initialization.py

• src/core/app_lifecycle.py

Control flow is: load environment/configuration → initialize persistence, API, risk, telemetry, and notifications → run scheduled/autonomous scan cycles → recover or shut down safely on lifecycle events.

Market-access boundary

DMarket REST v2 interaction is isolated behind a signed client with account, market, offer, target, fee, retry/backoff, and rate-limit concerns. Rust/PyO3 provides Ed25519 signing support, creating a language boundary between Python orchestration and cryptographic performance code.

• src/api/dmarket_api_client/core.py

• src/api/dmarket_api_client/rate_limiter.py

• src/rust_core/src/lib.rs

External dependencies are DMarket API credentials/services and, optionally, Telegram. API responses become normalized market/candidate inputs for trading stages.

Target-sniping trading pipeline

target_sniping is the architecture’s execution-critical pipeline. Its cycle orchestrator schedules scanner output through filtering, microstructure analysis, demand/price ranking, validation, guarded execution, and resale management.

• src/core/target_sniping/cycle_orchestrator.py

• src/core/target_sniping/scanner.py

• src/core/target_sniping/execution.py

Primary data flow: listings and aggregated bid/ask data → candidate items → OBI/demand and price calculations → ranked, validated targets → purchase order or sandbox result → inventory position → resale/exit action.

Signal, valuation, and strategy layer

Microstructure modules calculate OBI-related signals and volatility; the algorithm pack supplies statistical models. The demand strategy applies adaptive thresholds, and pricing/fair-value logic establishes expected economics before a target is eligible.

• src/core/target_sniping/microstructure_pipeline.py

• src/core/target_sniping/demand_strategy.py

• src/core/target_sniping/pricing.py

This is intentionally separated from API transport and execution, allowing signal generation and strategy tuning without directly placing orders. Additional strategies exist under src/strategies/, but the target-sniping demand pipeline is the stated primary path.

Risk, positions, and persistence

Risk controls form a gate around both entry and holding: price validation, liquidity/pump checks, concentration/portfolio rules, dynamic controls, and circuit breaking. Position guards and resale modules determine stops, profit targets, holding limits, and exit execution. SQLite in WAL mode persists price history, inventory, targets, state, and analytics.

• src/risk/risk_manager.py

• src/core/target_sniping/position_guard.py

• src/db/price_history/core.py

The persistence boundary should preserve trade/position state across restarts; recovery logic is in the core lifecycle rather than the database package.

Operations, observability, and control

Telegram is the operator command and notification surface, while Prometheus metrics and health endpoints support machine monitoring. Docker Compose provides deployment composition; Python 3.13+ is the main runtime, with Rust built through Cargo/Maturin.

• src/telegram/control_bot/bot.py

• src/monitoring/prometheus_metrics.py

• src/utils/health_server.py

Dry-run and live-shadow components provide non-production execution boundaries before real DMarket order placement.

• src/core/live_shadow.py

• src/core/target_sniping/sandbox.py