Skip to content

FTE_SEED.md

Latest commit

 

History

History
49 lines (41 loc) · 3.29 KB

FTE_SEED.md

File metadata and controls

49 lines (41 loc) · 3.29 KB
Raw

FTE capacity report — seed & data model

The first test module. Reports team staffing capacity from Objective issues + a Projects v2 board. FTE is summed per person per Program Increment (PI); > 1.0 = over-allocated. A duration-weighted FTE adjusts for objectives that cover only part of a PI.

Data split (important)

  • Issue body holds the ## LOE/FTE table — | Person | Role | FTE | Notes |. Roles are a fixed set (PM, Frontend, Backend, Geospatial, Data Curator, Comms, ML, Designer, Jupyterhub); any other role or non-numeric FTE is skipped + warned.
  • Board fields hold everything else, per objective: Program Increment, Start/End dates, and Project / Initiative / Team. The generator joins body + board by issue number.

Board fields (created by seed/bootstrap_board.py)

Field Type Notes
Program Increment single-select PI 26.4, PI 27.2 — add more to test PI changes. Windows come from PI_WINDOWS in the generator; the parser also handles a real iteration PI field.
Start Date / End Date date the objective's own window inside the PI
Project / Initiative / Team single-select grouping fields; Initiative drives the dashboard's Capacity Matrix grouping

Seed order (seed/)

generate_sample_issues.py (→ sample_issues.json) → create_issues.py (real issues) → setup_project.py (add to board + PI/Start/End) → setup_board_grouping.py (Project/Initiative/Team). Tracking files (created_issues.json, project_items.json) stay in seed/. Board = the personal project kyle-lesinger/projects/2 (isolated from the org project list).

Generator + workflow

.github/scripts/generate_fte_report.py (stdlib) accepts --project-json (board JSON), --issues-json (offline sample), or --from-dir; --pi "PI 26.4" filters. It emits PI-slugged fte_allocations_<slug>.csv (with project,initiative,team), fte_by_person_<slug>.csv, fte_by_role_<slug>.csv, and fte_summary_<slug>.md (slug = pi-26.4, or all-pis when unfiltered), plus a fixed fte_manifest.json. The workflow fte-report.yml runs it against the board and publishes to fte-report/<pi> (+ artifact + run summary). Adding a board field ⇒ no code change: the parser reads any field generically via _item_field(it, "<name>").

Reconciliation invariant (tested)

Σ allocation FTE == Σ by_person == Σ by_role, per PI, for both raw and weighted (weighted is rounded per-allocation, so it reconciles to the cent). Sample assignment of PI / Initiative / Project / Team is deterministic (seeded + index-based) so reports diff cleanly across runs.

Gotchas

  • PROJECT_TOKEN_FOR_BOARD_READS (repo secret) = classic PAT repo + read:org + project. The default GITHUB_TOKEN cannot read Projects v2. unknown owner type from gh project ⇒ missing read:org.
  • GraphQL rate limit is 5,000/hr — a full reseed makes 150+ board mutations; space them out or you'll get throttled (the Action's board read will then fail until the hour resets).
  • Dashboard reads the public fte-report/all-pis branch at runtime (fte-dashboard/src/data.ts RAW_BASE), with the bundled public/data/ snapshot as fallback.

Full dashboard internals: source repo Disasters-Learning-Portal/disasters-aws-conversion, PR #74.