Precon Sizing Workflow
Summary
Precon sizing produces early, evidence-backed package understanding from the published document corpus. It extracts trade and scope signals, applies heuristics and guardrails, records citations, and exposes results as estimator evidence rather than final estimating authority.
The concept page Precon Sizing System is the primitive orientation. This thread explains how sizing fits into the opportunity workflow.
Reader Question
How does Precon turn documents into useful sizing signals while keeping uncertainty, citations, and guardrails visible?
Surface Or Workflow
Operators see sizing in opportunity detail, workspace summaries, sizing panels, readiness signals, and diagnostics. Engineers enter through sizing services, shared sizing contracts, processing tasks, preview jobs, quality reports, benchmark jobs, comparison jobs, and a large targeted test suite.
Lifecycle
Sizing starts after documents have been published or indexed. Processing tasks select the relevant corpus, classify pages, extract signals, build or reuse indexes, apply selection and synthesis logic, persist snapshots, and attach evidence references. Quality and benchmark jobs compare outputs across opportunities and help tune the workflow.
Guardrails matter because a sizing result can be useful without being complete. The workflow should show sparse recall, stale indexes, low confidence, missing evidence, or scope ambiguity instead of hiding it behind a confident summary.
Child Threads
- Precon Document Publication Workflow: provides the corpus sizing depends on.
- Precon Estimator Workspace Workflow: presents sizing evidence to operators.
- AI In Operational Workflows: explains model-assisted evidence handling.
Implementation Boundaries
Sizing domain services live under services/precon/sizing; shared contracts and guardrails live under libs/precon/sizing; runtime integration lives under services/precon/processing/runtime and services/precon/processing/tasks. Preview, quality, benchmark, compare, and reuse jobs live under jobs/precon.
Tradeoffs
Sizing is treated as evidence because early document understanding is valuable but incomplete. That posture prevents the system from over-claiming and gives estimators enough citations and diagnostics to decide how much to trust the output.
Visual
The current visual is a graph neighborhood. The intended visual is a sizing lineage diagram from canonical documents to page classification, selected evidence, synthesis, guardrails, snapshot, diagnostics, and workspace display.
Source Evidence
docs/engineering/concepts/precon-sizing-system.mdservices/precon/sizinglibs/precon/sizingservices/precon/processing/runtime/sizing_tasks.pyservices/precon/processing/tasks/sizing.pyservices/precon/diagnostics/sizing_operations.pyjobs/precon/preview_opportunity_sizing/run.pyjobs/precon/report_opportunity_sizing_quality/run.pytests/precon/test_precon_sizing_synthesis_contracts.py