Skip to content

Passivbot v7.9.0

Choose a tag to compare

View all tags
@enarjord enarjord released this 04 Apr 17:53
· 401 commits to master since this release
v7.9.0
8df57ea
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Release Notes for v7.9.0

These notes describe the user-facing changes from master to v7.9.0.

Highlights

  • Equity hard-stop loss is now a full live + backtest feature with Rust-owned runtime handling, richer metrics, and deterministic fake-live replay coverage.
  • Optimization now supports pymoo as a first-class backend and uses it by default, with explicit objective goals and live NSGA-II / NSGA-III configuration.
  • A new passivbot tool pareto explorer makes it easier to inspect and select Pareto candidates from local optimize results.
  • Config loading now runs through a canonical staged pipeline with in-code schema defaults, compatibility migrations for older config keys, and a new canonical example config path.
  • Passivbot now ships a unified passivbot CLI with clearer default help, install-profile separation, stronger environment mismatch detection, and stricter Rust freshness checks.
  • A built-in monitor stack is now available: publisher, relay, browser dashboard, TUI, and helper wrappers.

Upgrade Notes

  • Reinstall after upgrading.
    python3 -m pip install -e .
    or
    python3 -m pip install -e ".[full]"
  • If Rust looks stale or the wrong binary is being used, run:
    maturin develop --release
  • optimize.backend now defaults to pymoo, so optimize users need the full install profile with the new dependency set.
  • configs/template.json is no longer the canonical starting point. Use configs/examples/default_trailing_grid_long_npos10.json or omit the config path to start from schema defaults.
  • The canonical schema enables the local monitor by default. Set monitor.enabled = false if you do not want snapshot and event files written locally.
  • live.max_realized_loss_pct now defaults to 1.0, so the realized-loss gate is opt-in unless you set a tighter value explicitly.

What's New

Equity Hard Stop Loss

  • Added the account-level equity hard-stop framework to live and backtest, with Rust-managed drawdown state, tier tracking, RED latching, cooldown restart logic, and no-restart handling.
  • Added richer HSL metrics and plots, including per-year trigger and restart counts, time-share metrics, halt and restart metrics, panic-close metrics, and hard_stop_drawdown.png.
  • Added deterministic fake-live replay tooling and scenarios so RED halt, cooldown behavior, manual intervention, and restart policies can be tested locally.
  • Clarified HSL semantics around no-restart threshold clamping, sub-interval EMA fallback, and market panic execution.

Optimization and Pareto Tooling

  • Added the pymoo backend alongside DEAP and made it the default optimizer backend.
  • Activated real runtime support for nested optimize.pymoo.* settings, including algorithm, shared crossover and mutation settings, and NSGA-III reference-direction configuration.
  • Changed optimizer scoring to explicit {metric, goal} specs instead of implicit sign-based weights, while keeping legacy string-list configs readable.
  • Added auto-sized NSGA-III population defaults when optimize.population_size is null.
  • Added passivbot tool pareto, a CLI Pareto explorer for filtering local Pareto candidates with optimizer-style limit expressions and selecting a single candidate using methods such as knee, reference-point, ideal-point, weighted utility, lexicographic, or outranking selection.
  • The Pareto explorer can now default to the newest local optimize_results/.../pareto, accept either a run directory or a pareto/ directory, show the retained front's ideal point, and use stored metrics outside the original optimize.scoring list when their direction is known.

Configs, CLI, and Install Flow

  • Moved canonical defaults to src/config/schema.py and made schema defaults the no-config-path behavior for live, backtest, and optimize.
  • Added staged config normalization, runtime compilation helpers, migration logging, and backward-compatibility renames for older config keys.
  • Replaced the old implicit template workflow with configs/examples/default_trailing_grid_long_npos10.json.
  • Added the unified passivbot CLI, curated default help, --help-all, and install-profile guidance for live, full, and dev.
  • Restored passivbot live --user / -u as the curated shorthand for live.user, and added a curated shorthand for live.pnls_max_lookback_days as --pnls-max-lookback-days / -pmld in the default live help.
  • Added environment mismatch detection so stale shell shims and wrong-entrypoint installs fail loudly or re-exec into the active environment.
  • Tightened Rust extension freshness checks to reduce silent stale-binary runs.

Monitoring and Diagnostics

  • Added the local monitor publisher for bot snapshots, event streams, and retained fill, price-tick, and candle history.
  • Added the read-only monitor relay with websocket streaming and recent-message replay.
  • Added the browser dashboard, terminal TUI, monitor-web, and monitor-dev.
  • Added standalone trailing diagnostics tooling for recomputing trailing-entry and trailing-close behavior from saved snapshots or manual input.
  • Added repro and sync sidecar tools for investigation and deployment workflows.

Backtest and Runtime Behavior

  • Rust now owns more of the market-vs-limit execution intent, and live plus backtest now consume the same shared logic.
  • Market fills in backtests now use taker fees, optional taker-fee override support, and explicit maker/taker liquidity labeling in fills.csv.
  • Backtest BTC metrics now always use BTC equity instead of mirroring USD analysis when btc_collateral_cap = 0.
  • ADG terminal smoothing now uses the last up to three daily samples instead of an EMA over the full run.
  • Executable min-cost filtering now uses actual rounded executable size rather than raw market metadata, improving forager tradability filtering.
  • First-timestamp cache handling for newly listed coins is more robust, avoiding fetches from invalid early dates.

Exchange and Market Coverage

  • Hyperliquid HIP-3 handling is more robust across margin-mode detection, state sync, and source-dir resolution for stock-perp backtests.
  • Isolated HIP-3 live trading is explicitly blocked for now instead of partially supported.

Discord Forum Announcement Draft

Passivbot v7.9.0 is out.

This release covers the full user-facing diff from the current master branch to v7.9.0, with major changes across live trading, backtesting, optimization, config handling, CLI UX, and monitoring.

Important upgrade note first:

  • After pulling, reinstall Passivbot in your active environment.
  • Live-only: python3 -m pip install -e .
  • Backtest / optimize / research: python3 -m pip install -e ".[full]"
  • If Rust looks stale, run maturin develop --release

Main changes in v7.9.0:

  • Equity hard-stop loss is now a full live + backtest feature with Rust-owned runtime state, cooldown handling, improved RED supervision, richer metrics, and deterministic fake-live replay scenarios.
  • Optimization now supports pymoo as a first-class backend and uses it by default. NSGA-II / NSGA-III settings under optimize.pymoo.* are now actually honored, and NSGA-III population sizing can be auto-derived from reference directions.
  • Optimizer scoring is now explicit via {metric, goal} entries instead of implicit signed weights. Legacy scoring configs still load.
  • A new passivbot tool pareto explorer can filter local Pareto candidates with optimizer-style limit expressions and select a single config using knee, reference-point, ideal-point, weighted utility, lexicographic, or outranking methods.
  • Config loading now uses a canonical staged pipeline with in-code schema defaults, compatibility migrations for older keys, and a new canonical example config at configs/examples/default_trailing_grid_long_npos10.json.
  • Passivbot now has a unified passivbot CLI with cleaner default help, --help-all, install-profile separation, restored passivbot live -u, and much stronger environment / stale-extension detection.
  • A new local monitoring stack is included: publisher, relay, browser dashboard, TUI, monitor-web, and monitor-dev.
  • Rust and Python execution behavior are better aligned across live and backtest, especially for market-vs-limit order intent, HSL panic behavior, and market-fill fee modeling.
  • Backtests now include better HSL plots and metrics, cleaner BTC-relative analysis, taker-fee handling for market executions, and improved newly-listed-coin timestamp handling.
  • Hyperliquid HIP-3 handling is more robust, while isolated HIP-3 live trading is now blocked explicitly until it is properly supported.

Behavior changes to be aware of:

  • optimize.backend now defaults to pymoo
  • configs/template.json is no longer the canonical starting point
  • the canonical schema enables the local monitor by default
  • live.max_realized_loss_pct now defaults to 1.0, so the realized-loss gate is opt-in unless you set it explicitly

Telegram Announcement Draft

Passivbot v7.9.0 is out.

Key changes:

  • full live + backtest equity hard-stop framework with richer metrics and fake-live replay tooling
  • pymoo optimizer backend added and now the default
  • new passivbot tool pareto explorer for filtering and selecting Pareto candidates
  • canonical staged config pipeline with schema defaults and new example config path
  • unified passivbot CLI with better help and stricter environment / Rust freshness checks
  • new monitor stack: publisher, relay, web dashboard, and TUI
  • better live/backtest alignment for market-vs-limit execution and market-fill fee handling

Important after upgrading:

  • reinstall Passivbot in your active env
  • use python3 -m pip install -e ".[full]" if you run optimize/backtest
  • rebuild Rust with maturin develop --release if needed

Also note:

  • optimize.backend now defaults to pymoo
  • configs/template.json is no longer the canonical starting point
  • monitor is enabled by default in the schema
  • live.max_realized_loss_pct now defaults to 1.0
Assets 2
Loading