Changelog¶
All notable changes to this project will be documented in this file.
[0.2.0] - In Progress¶
Added¶
- Pure-Python solver backend (
fppy) with dual-engine parity validation - MCP server expanded to 44 tools with 9 resources and 6 structured prompts
- Agent-first workspace authoring: packs, cards, recipes, managed workspaces
- FRED, BEA, and BLS data ingestion pipelines with safe-lane update workflow
- Official Fair bundle fetch and verification (
fp data fetch-fair-bundle) - Scenario DSL compiler (
fp dsl compile) - Parity regression testing with golden baselines (
fp parity --regression) - Triage tools for fppy reports and parity hard-fails
- Variable dictionary with source coverage and quality reports
- Sensitivity analysis with fan charts and response tables
- Historical fit comparison page in dashboard
- GitHub Pages static model-runs export (
fp export pages) - 500+ tests across 81 test files
Changed¶
- Dashboard expanded to 12 pages (from initial 3)
- Input parser now uses canonical snake_case keys throughout
- Scenario runner supports backend selection (
fpexe/fppy/both)
[0.1.0] - Unreleased¶
Added¶
- Scenario YAML configuration with Pydantic validation (
fp run,fp validate). - FP output parser: forecast tables, estimation results (31 equations), solve iterations.
- FP input file parser: full DSL support with normalized canonical keys.
- Exogenous override generation: single-value and multi-period series.
- Run comparison with ranked variable deltas (
fp diff). - Equation dependency graph analysis with networkx (
fp graph). - Batch scenario runner with golden-output regression comparison (
fp batch). - Diff export to CSV and Excel (
fp diff --export csv|excel). - Matplotlib forecast and comparison charts (
fp viz plot). - Interactive Streamlit dashboard with Plotly charts (Explore, Compare, New Run pages).
- Run artifact discovery from filesystem (
fp history). - MCP server with 8 tools:
run_fp_scenario,parse_fp_output,diff_runs,list_output_variables,list_output_equations,validate_scenario,list_scenarios,run_batch_scenarios. - Dashboard: equation graph visualization page.
- Dashboard: report download on Explore page, diff CSV download on Compare page.
- Markdown run report generation (
fp report). - 96 tests covering parsers, CLI, dashboard, writer, batch, graph modules.
- CLI with subcommands: run, validate, diff, graph, batch, report, history, dashboard, io, viz.
- Example scenarios: baseline, higher_growth, tight_monetary, recession, supply_shock, fiscal_expansion.
Known limitations¶
fp.exerequires Wine on macOS/Linux (Windows PE32 binary).- FM/ model data files are not distributable.
- Dashboard caching may serve stale data on scenario re-run to same directory.
- Equation graph requires optional
networkxdependency (pip install fp-wraptr[graph]). - Excel export requires optional
openpyxldependency.
Requirements¶
- Python 3.11+
- Wine (for
fp.exeexecution on non-Windows)