Milestones
List view
## Goal Add evidence-driven calibration and extended scenario families for quantitative R&D. Target release: **v2.7**. ## Scope - Real-data statistical profile extraction. - Calibration/training/holdout semantics. - Serializable fit and validation reports. - Evaluate empirical/block-bootstrap, scripted shock and regime scenarios. - Maintain model registry and suitability matrix. ## Exit Criteria - Calibration artifacts are reproducible and versioned. - Fit and validation evidence are separated where required. - Each added model has a concrete use case, stochastic contract and validation evidence. - No model is admitted merely because it is a familiar finance technique.
No due date•0/8 issues closed## Goal Modernize packaging, CI and release delivery for the 2.x line. Target release: **v2.6**. ## Scope - pyproject-based packaging and one canonical version source. - Explicit supported Python matrix. - Core/optional dependency classification. - GitHub Actions PR quality gate. - Human-gated release artifact verification/publication. - Current Fuzzy Technologies project metadata. ## Exit Criteria - Supported environments build, install and test cleanly. - PR quality gates are deterministic and blocking. - sdist/wheel artifacts are verified before publication. - Release publication requires explicit human approval. - Active metadata no longer presents obsolete Travis/personal-account infrastructure as current.
No due date•0/8 issues closed## Goal Make large deterministic scenario generation measurable and efficient. Target release: **v2.5**. ## Scope - Benchmark the actual package from small samples through 100k/1M rows. - Reduce avoidable generation allocation overhead. - Add chunked or streaming generation/export where justified. ## Exit Criteria - Runtime and peak-memory evidence is reproducible. - Optimizations demonstrate material improvement against the baseline. - Promised seeded sequence semantics are preserved. - Large-series workflows do not require excessive high-overhead intermediates.
No due date•0/4 issues closed## Goal Establish the intentional PriceGenerator 2.x public architecture. Target release: **v2.4**. ## Scope - Explicit v2 configuration and public API. - Pure generation core and market-data validation boundary. - Separate I/O and analysis boundaries. - Optional rendering/indicator integrations. - Remove import-time and shell/process side effects. ## Exit Criteria - Core generation is independent of optional rendering/TA dependencies. - v2 public contracts are documented and tested. - Analysis does not unexpectedly mutate canonical OHLCV data. - 1.x compatibility helpers exist only where they provide concrete migration value.
No due date•0/8 issues closed## Goal Establish correct, versioned stochastic-model semantics. Target release: **v2.3**. ## Scope - Preserve historical stochastic behavior as an explicitly named reference model. - Define direction, doji, boundary and rounding semantics. - Replace endpoint-forced trend behavior in a new model. - Define outlier, negative-price and volume-process semantics. ## Exit Criteria - Model parameters have explicit measurable meanings. - Model-specific invariant/statistical tests pass. - Trend generation does not violate declared candle constraints. - Behaviorally different models have distinct model identities.
No due date•0/8 issues closed## Goal Make synthetic market-data quality quantitatively measurable and public claims evidence-based. Target release: **v2.2**. ## Scope - Return, candle, volume, dependence and drawdown diagnostics. - Reproducible real-vs-synthetic comparison. - Statistical tolerance/confidence policy. - Realism levels and suitability limits. ## Exit Criteria - Validation metrics have precise definitions and deterministic tests. - Comparison reports are reproducible and machine-readable. - Realism/similarity claims map to named evidence. - Unsupported marketing wording is removed or qualified.
No due date•0/8 issues closed## Goal Establish deterministic scenario identity and a canonical cross-language market-data contract. Target release: **v2.1**. ## Scope - Explicit RNG ownership, seed and model identity. - Scenario manifests and deterministic fingerprints. - Canonical OHLCV validation. - Datetime, timezone, timeframe and CSV semantics. ## Exit Criteria - Same supported model/config/seed identity is reproducible. - Generated and loaded OHLCV share one executable validation contract. - Python/CSV/MQL5 interchange semantics are explicit. - Host-local time or process-global RNG state is not an implicit deterministic input.
No due date•0/9 issues closed## Goal Establish the reproducible engineering baseline for PriceGenerator 2.x before behavior-changing work. Target release: **v2.0**. ## Scope - Snapshot repository, package, release and transfer state. - Characterize published 1.x generator behavior. - Codify existing invariants and known defects. - Establish clean-install, test and benchmark evidence. ## Exit Criteria - The baseline is reproducibly identified and executable. - Characterization/invariant tests provide a trusted comparison point. - Clean installation and baseline performance evidence are recorded. - No production generator semantics are changed as part of baseline establishment.
No due date•2/9 issues closed