Added a staged roadmap to implement the small changes and refactors identified by the audit

This commit is contained in:
2026-06-16 09:51:22 -05:00
parent a27e870522
commit 0884eb0ce5
4 changed files with 481 additions and 1366 deletions

481
docs/roadmap/cleanup.md Normal file
View File

@@ -0,0 +1,481 @@
# Cleanup Roadmap
## Purpose
This roadmap defines the staged cleanup work recommended by
`docs/roadmap/audit.md`. It is written for an LLM coding agent that will
implement each stage in order.
The cleanup sequence is intentionally narrow. It should reduce duplication and
clarify package responsibilities before the next major release without changing
public CLI syntax, user configuration, report identities, managed artifact
paths, or generated report behavior.
This file may describe planned work because it lives under `docs/roadmap/`.
## Cleanup Principles
- Preserve public behavior unless a stage explicitly says otherwise.
- Prefer small behavior-preserving refactors over broad rewrites.
- Keep domain policy in the package that owns the relevant contract.
- Keep external system details behind adapter boundaries.
- Keep report, module, template, schema, path, and artifact identity explicit.
- Add tests before or during cleanup where they protect public behavior or
important internal invariants.
- Update implemented documentation only after code behavior exists.
- Do not use cleanup as an opportunity to introduce new features.
## Locked Decisions
- Daily, Today, and Tomorrow remain separate report IDs, prompt IDs, schemas,
templates, and public generated-text types.
- Cleanup may reduce shared internal day-style plumbing, but must not merge the
public report types.
- Daily, Today, and Tomorrow keep separate top-level template files.
- Named template partials may be introduced for repeated Daypart Forecast and
Precipitation Timing blocks.
- Configuration ownership remains in `internal/config`; config parsing and
validation should not move into `internal/app` or `internal/briefing`.
- Scriptorium and distributor dependency details must remain behind their
adapter packages.
- Existing public CLI syntax, config fields, report IDs, prompt IDs, template
IDs, schema IDs, workspace paths, distributor paths, and generated report
paths remain stable.
- Do not introduce Cobra, a workflow engine, a plugin system,
manifest/resume/progress infrastructure, per-module packages, per-report
packages, or a global test helper package.
## Stage 1: Day-Style GeneratedText Validation Helpers
### Goal
Remove duplicated Daily/Today/Tomorrow generated-text validation while
preserving public types and JSON schema behavior.
### Implementation Guidance
- Add an unexported shared helper in `internal/generatedtext` for the common
day-style shape: `summary`, `forecast_discussion`, optional
`precipitation_timing`, and optional `confidence`.
- Keep exported `Daily`, `Today`, and `Tomorrow` structs.
- Keep exported `ValidateDaily`, `ValidateToday`, and `ValidateTomorrow`
functions.
- Preserve normalized JSON output shape and unknown-field rejection.
- Preserve current error messages except for the expected report-name
substitution.
- Do not change embedded generated-text schemas in this stage except as needed
to keep tests aligned with existing behavior.
### Acceptance Criteria
- Daily, Today, and Tomorrow validation use one shared internal validation path
for common trim, required-field, optional-field, and normalization behavior.
- Public generated-text structs and function names remain unchanged.
- Existing callers do not need to change.
- Existing schema behavior remains unchanged.
### Tests
- Add table coverage proving Daily, Today, and Tomorrow share:
- required `summary` behavior;
- required non-empty `forecast_discussion` behavior;
- trim behavior;
- optional-field omission behavior;
- normalized JSON output behavior;
- unknown-field rejection.
- Run:
```sh
go test ./internal/generatedtext
```
### Prompt Size
Small enough for one implementation prompt.
## Stage 2: Day-Style Render Context And Template Shared Blocks
### Goal
Reduce repeated Daily/Today/Tomorrow render-context and Markdown template logic
without merging the reports.
### Implementation Guidance
- Add shared unexported helpers for common day-style report context fields such
as forecast date, forecast date label, generated timestamp label, valid
period, timezone, collected facts, and derived facts.
- Add a common module snapshot extraction helper for shared day-style modules.
- Keep report-specific planning modules separate:
- Daily uses `DailyPlanning`;
- Today uses `TodayPlanning`;
- Tomorrow uses `TomorrowPlanning`.
- Keep report-specific daypart wrapper types if tests or templates benefit from
explicit names.
- Add named template partial support in `internal/reporttemplate` for repeated
Daypart Forecast and Precipitation Timing blocks.
- Keep `daily.md.tmpl`, `today.md.tmpl`, and `tomorrow.md.tmpl` as separate
top-level templates that opt into shared partials.
- Preserve current rendered Markdown behavior, including Today's omission of
elapsed or missing daypart lines.
- Do not replace editable Markdown templates with Go string builders.
### Acceptance Criteria
- Common day-style render-context setup is shared internally.
- Daily, Today, and Tomorrow still expose their own render context types.
- Shared template partials reduce repeated daypart and precipitation template
logic.
- Report templates remain separately editable.
- Current rendered output remains stable except for whitespace changes that are
covered by updated tests and intentionally accepted.
### Tests
- Update render-context tests to prove:
- common module fields are populated for Daily, Today, and Tomorrow;
- each report still exposes its correct planning module;
- Today still omits missing/elapsed dayparts where current behavior expects
omission.
- Update reporttemplate tests to prove Daily, Today, and Tomorrow output remains
behaviorally stable.
- Run:
```sh
go test ./internal/generatedtext ./internal/reporttemplate
```
### Prompt Size
Likely one implementation prompt. If template partial parsing changes and
render-context helper extraction become difficult to review together, split
this into:
1. render-context helper cleanup;
2. template partial cleanup.
## Stage 3: Report Module Config Traversal Cleanup
### Goal
Make report-module config normalization, validation, and override extraction use
one canonical traversal.
### Implementation Guidance
- Refactor `internal/config/reports.go` around one unexported helper that:
- resolves report config keys through `internal/report`;
- detects duplicate aliases;
- validates report IDs against the report registry;
- normalizes module options when requested;
- builds `module.ConfigItem` values;
- validates module composition through the module registry.
- Reuse that helper from:
- `normalizeReportModules`;
- `validateReportModules`;
- `ReportModuleOverrides`.
- Preserve current configuration precedence and YAML shape.
- Preserve current option normalization behavior.
- Preserve error context such as
`reports.<key>.deterministic_modules[...]`.
- Keep `internal/config` as the owner of config loading, normalization, and
validation.
### Acceptance Criteria
- Report-module config traversal exists in one implementation path.
- Load-time normalization, validation, and override extraction cannot drift on
report-key or module composition policy.
- Existing config files continue to load unchanged.
- Existing config error messages remain materially equivalent and actionable.
### Tests
- Keep or update tests for:
- unknown report keys;
- duplicate report aliases;
- unknown modules;
- duplicate modules;
- incompatible modules;
- invalid module options;
- valid module overrides.
- Add coverage proving loaded config and manually constructed config fail
consistently for the same invalid report/module cases.
- Run:
```sh
go test ./internal/config ./internal/app
```
### Prompt Size
Small enough for one implementation prompt.
## Stage 4: Scriptorium Run Plumbing Cleanup
### Goal
Share `Run` and `StructuredRun` execution mechanics while keeping the
Scriptorium adapter API stable.
### Implementation Guidance
- Add an unexported run-like helper in `internal/adapters/scriptorium` that:
- validates prompt ID;
- validates data package path;
- validates output path;
- executes the resolved argv;
- captures stdout and stderr;
- records truncation flags;
- records output path;
- handles nonzero exit results.
- Keep exported `RunRequest`, `StructuredRunRequest`, `RunResult`, and
`StructuredRunResult`.
- Keep `StructuredRun` using the same argv shape as `Run`; do not add schema or
format flags.
- Preserve argv order.
- Preserve stdout/stderr capture and truncation fields.
- Preserve output-path fields.
- Preserve existing nonzero-exit error wording as closely as possible.
### Acceptance Criteria
- `Run` and `StructuredRun` share validation and result construction mechanics.
- Public adapter request/result types remain stable.
- Existing app-layer Scriptorium calls do not need behavior changes.
- Existing Scriptorium tests still pass with minimal expected-output updates.
### Tests
- Add or update parity tests proving `Run` and `StructuredRun` preserve:
- argv construction;
- output path;
- captured stdout/stderr;
- truncation flags;
- nonzero exit result and error behavior.
- Run:
```sh
go test ./internal/adapters/scriptorium
```
### Prompt Size
Small enough for one implementation prompt.
## Stage 5: CLI Test Fixture Cleanup
### Goal
Reduce noisy repeated CLI integration test setup without creating a
cross-package test framework.
### Implementation Guidance
- Add package-local helpers in `internal/cli` tests for repeated:
- fake Scriptorium script setup;
- generated-text JSON responses;
- config file writing;
- Weather API test server setup;
- artifact path or glob assertions.
- Keep a few explicit CLI workflow tests readable end-to-end.
- Do not add `internal/testutil` or another global test helper package.
- Do not weaken assertions while deduplicating setup.
- Do not change production CLI behavior in this stage.
### Acceptance Criteria
- Test setup repetition is reduced in the largest CLI test file.
- Test behavior and coverage remain equivalent.
- Helpers are local to `internal/cli`.
- No production code changes are required for this stage.
### Tests
- Run:
```sh
go test ./internal/cli
```
### Prompt Size
Small enough for one implementation prompt.
## Stage 6: App Test Fixture Cleanup
### Goal
Reduce noisy repeated app orchestration test setup without creating a
cross-package test framework.
### Implementation Guidance
- Add package-local helpers in `internal/app` tests where repetition is high and
the helper improves readability.
- Good candidates include repeated fake renderer setup, generated-text JSON
responses, config file writing, Weather API test server setup, distributor
notifier setup, recording store setup, and artifact assertions.
- Keep key workflow tests readable end-to-end so generation ordering remains
clear.
- Do not add `internal/testutil` or another global test helper package.
- Do not weaken assertions while deduplicating setup.
- Do not change production app behavior in this stage.
### Acceptance Criteria
- Test setup repetition is reduced in the largest app test file.
- Test behavior and coverage remain equivalent.
- Helpers are local to `internal/app`.
- No production code changes are required for this stage.
### Tests
- Run:
```sh
go test ./internal/app
```
### Prompt Size
Small enough for one implementation prompt.
## Stage 7: State Artifact Save Helper Cleanup
### Goal
Reduce repeated filesystem artifact write boilerplate while keeping artifact
semantics visible.
### Implementation Guidance
- Add small private helpers in `internal/state/filesystem.go` for resolved JSON
and byte artifact writes.
- Keep artifact-specific validation in public save methods before calling any
helper.
- Keep each public save method explicit about which artifact path it writes.
- Leave `SaveDataPackage` with its current special behavior unless a helper
cleanly preserves `promptinput.Save`.
- Leave `SaveMetadata` with its current explicit metadata-path validation and
write behavior unless a helper cleanly preserves it.
- Do not introduce a manifest, artifact registry, resume system, or broad
artifact framework.
### Acceptance Criteria
- Repeated `Paths` plus atomic write mechanics are reduced where the semantics
are identical.
- Artifact-specific validation and path choice remain easy to see.
- Managed artifact paths do not change.
- Metadata JSON shape does not change.
### Tests
- Keep or update state save tests and metadata round-trip tests.
- Run:
```sh
go test ./internal/state ./internal/app
```
### Prompt Size
Small enough for one implementation prompt.
## Stage 8: Documentation And Final Validation
### Goal
Align implemented documentation only where cleanup changes internal contracts or
template editing guidance.
### Implementation Guidance
- Update `docs/templates.md` if named template partial support changes how
report templates should be edited.
- Update relevant `docs/internal/*` files only for implemented internal
contract changes.
- Update `docs/policy/development.md` only if contributor workflow guidance
changes.
- Keep unimplemented or deferred cleanup ideas only under `docs/roadmap/`.
- Do not document future refactors as implemented behavior.
### Acceptance Criteria
- Non-roadmap docs describe only implemented behavior.
- Template-editing guidance matches the final template partial structure, if
partials were added.
- Internal docs remain accurate for generated text, templates, config, state,
and adapters touched by cleanup.
### Validation
Run:
```sh
go test ./internal/generatedtext ./internal/reporttemplate
go test ./internal/config ./internal/app
go test ./internal/adapters/scriptorium
go test ./internal/cli ./internal/state
go test ./...
go run ./cmd/weatherreporter --help
git diff --check
```
### Prompt Size
Small enough for one implementation prompt.
## Deferred Refactors
The following refactors are out of scope for this cleanup sequence:
- Generic workflow engine.
- Cobra migration or CLI redesign.
- Plugin architecture.
- Per-module or per-report packages.
- Broad Weather API source-ingestion framework.
- Manifest, resume, or progress system.
- Global test helper package.
- Consolidating Daily, Today, and Tomorrow into one public report type.
- Replacing editable Markdown templates with Go string builders.
- Build-free module catalog split.
The build-free module catalog split may be revisited later if
config/module-boundary complexity grows enough to justify separating module
metadata from module builders.
## Global Validation Checklist
Run these checks after completing the full cleanup sequence:
```sh
go test ./...
go run ./cmd/weatherreporter --help
git diff --check
```
Also run focused checks after the relevant stages:
```sh
go test ./internal/generatedtext
go test ./internal/generatedtext ./internal/reporttemplate
go test ./internal/config ./internal/app
go test ./internal/adapters/scriptorium
go test ./internal/cli
go test ./internal/app
go test ./internal/state ./internal/app
```
Manual review checklist:
- Public CLI syntax is unchanged.
- Public config fields and defaults are unchanged.
- Report IDs, prompt IDs, template IDs, and schema IDs are unchanged.
- Managed workspace artifact paths are unchanged.
- Distributor bundle paths and notification behavior are unchanged.
- Generated report Markdown behavior is unchanged except for intentional,
test-covered whitespace differences.
- Scriptorium argv construction is unchanged.
- Non-roadmap docs do not describe unimplemented cleanup work.

View File

@@ -1,348 +0,0 @@
# Daily Report Roadmap
## Purpose
This roadmap defines the target state and policy choices for replacing the
existing `daily` report with a new generated-text-template `daily` report.
The new report is not implemented yet. Current report behavior remains
documented outside `docs/roadmap/`.
## Intent
Daily should be an independent generated-text-template report for a user-chosen
local civil day. Its rendered Markdown should initially match the Tomorrow
Report format exactly, but its identity, prompt, template, schema, valid-period
resolver, config key, and planning module should be separate from Tomorrow from
the start.
The important distinction is date selection:
- `tomorrow` always targets the next local civil day.
- `daily` targets the local civil day explicitly supplied by the user with
`--date YYYY-MM-DD`.
This is a clean breaking replacement of the current Daily implementation:
- The existing direct-Markdown `daily` report is removed.
- The legacy `daily_today` report ID is removed from active report definitions.
- `weatherreporter generate daily` remains the public command name, but it now
runs the new generated-text-template Daily report.
- `weatherreporter generate daily` requires `--date YYYY-MM-DD`.
- Existing historical `daily_today` workspace artifacts do not need migration.
## Locked Decisions
- New report ID: `daily`.
- Public command: `generate daily`.
- `generate daily` must require `--date YYYY-MM-DD`.
- The provided date is interpreted as a civil date in the effective report
timezone.
- The valid period is the selected local civil day, `[00:00, next 00:00)`.
- Daily should not be an alias for Today or Tomorrow.
- Today remains the current-day scheduled morning product.
- Tomorrow remains the next-day scheduled evening product.
- Daily is manually targeted by date and is not added to morning or evening
batch membership in this roadmap.
- Daily must have its own template, generated-text schema, prompt asset,
render-context type, and planning module.
- Daily may share private helper functions with Tomorrow where mechanics are
identical, but it must not expose Tomorrow-specific public types or stanzas.
- The initial Daily output format should match Tomorrow's rendered Markdown
format.
## Target Report Shape
Daily should render the same Markdown structure as Tomorrow:
```markdown
# Monday's Weather
**Forecast date:** Monday, June 15, 2026
**Updated:** Sunday, June 14, 2026 at 9:14 AM
<GeneratedText summary>
## Daypart Forecast
- **Morning:** <deterministic daypart line>
- **Midday:** <deterministic daypart line>
- **Afternoon:** <deterministic daypart line>
- **Evening:** <deterministic daypart line>
## Precipitation Timing
- **1:00 PM** to **5:00 PM**: Precipitation is expected during this period.
The peak precipitation chance is 59% at 2:00 PM.
- <optional GeneratedText precipitation_timing>
## Forecast Discussion
<GeneratedText forecast_discussion paragraphs>
```
The precipitation section should render only when precipitation windows exist
for the selected valid period.
The title should follow Tomorrow's day-name style, for example:
- `Monday's Weather`
- `Tuesday's Weather`
- `Sunday's Weather`
## Report Identity
Replace the existing active Daily report with:
- report ID: `daily`
- public generate command: `daily`
- prompt ID: `weather.daily_generated_text`
- generation mode: `generated_text_template`
- template ID: `daily`
- generated-text schema ID: `daily`
- artifact group: `daily`
- batch output name: `daily.md`
- prior compatibility: Daily only
- comparison strategy: same valid local date
- valid period: selected local civil day in the effective report timezone,
`[00:00, next 00:00)`
Remove the legacy active report identity:
- remove active report ID `daily_today`
- remove prompt ID `weather.daily_report` from the current Daily path
- remove direct-Markdown generation mode from the Daily report definition
- remove `daily_today` config-key support unless a separate migration roadmap
explicitly reintroduces it
Historical artifacts with `daily_today` metadata may remain on disk. Do not
migrate or rewrite old workspace files in this feature.
## CLI Behavior
`weatherreporter generate daily` should require:
```sh
weatherreporter generate daily --date YYYY-MM-DD
```
Rules:
- `--date` is required for `generate daily`.
- `--date` accepts only `YYYY-MM-DD`.
- The date is interpreted in the effective report timezone after config and
`--tz` overrides are applied.
- Omitting `--date` is an error.
- A malformed date is an error.
- The command should continue supporting the existing global generation flags:
`--config`, `--units`, `--tz`, and `--out`.
- Do not default `daily` to today or tomorrow.
## Batch Behavior
Daily should not be added to scheduled batches in this roadmap.
Current intended scheduled behavior:
- Morning batch: `today`, `three_day`, and conditional `weekend`.
- Evening batch: `tomorrow`.
Daily is a manually targeted report. A future roadmap may add scheduled Daily
behavior if a concrete operational need appears.
## GeneratedText Contract
Daily should use the same structured prose shape as Tomorrow:
```json
{
"summary": "string",
"forecast_discussion": ["string"],
"precipitation_timing": "string",
"confidence": "string"
}
```
Required:
- `summary`
- `forecast_discussion`
Optional:
- `precipitation_timing`
- `confidence`
Validation should match Tomorrow semantics:
- reject malformed JSON and unknown fields
- reject trailing JSON values
- trim `summary`, `precipitation_timing`, and `confidence`
- trim each `forecast_discussion` paragraph
- drop blank discussion paragraphs
- require at least one nonblank discussion paragraph
- return canonical normalized JSON with the same public field names
Add a dedicated prompt asset:
- `internal/reporttemplate/prompts/daily.generated_text.md`
Scriptorium registration remains out of band. Weatherreporter should invoke the
Daily prompt by prompt ID and pass the data package as it does for other
generated-text reports.
## Template Context
Add dedicated Daily types under `internal/generatedtext`, rather than reusing
Tomorrow types directly:
```go
type DailyRenderContext struct {
Report DailyReportContext
GeneratedText Daily
Modules DailyTemplateModules
Collected facts.CollectedFacts
Derived facts.DerivedFacts
}
```
`DailyReportContext` should include:
- `Title`, for example `Monday's Weather`
- `ForecastDate`
- `ForecastDateLabel`, for example `Monday, June 15, 2026`
- `ForecastDayName`, for example `Monday`
- `GeneratedAt`
- `GeneratedAtLabel`
- `ValidPeriod`
- `Timezone`
`DailyTemplateModules` should expose the same categories the Daily template
needs:
- `Metadata`
- `CurrentConditions`
- `HourlyForecast`
- `DerivedDailySummary`
- `DerivedDaypartSummaries`
- ordered daypart rows
- `PrecipTiming`
- `AlertDigest`
- `SPCConvectiveOutlooks`
- `AreaForecastDiscussion`
- `SPCConvectiveDiscussion`
- `WeatherStory`
- `OutdoorWindows`
- `DailyPlanning`
Daily may share private helper functions with Tomorrow render-context
construction when the helper represents identical mechanics. Do not expose
Tomorrow-specific types through the Daily template context.
## Module Composition
The default module composition should initially mirror Tomorrow where the same
facts are useful for a dated daily report, with a Daily-specific planning
module:
- `metadata`
- `current_conditions`
- `narrative_forecast`
- `derived_daily_summary`
- `derived_daypart_summaries`
- `precip_timing`
- `alert_digest`
- `spc_convective_outlooks`
- `area_forecast_discussion`
- `spc_convective_discussion`
- `weather_story`
- `outdoor_windows`
- `daily_planning`
- `hourly_forecast`
The module order should match the intended data-package order unless tests show
a stronger reason to mirror Tomorrow's exact current order.
## Daily Planning Module
Add a Daily-specific deterministic planning module:
- module ID: `daily_planning`
- stanza name: `daily_planning`
- options type: `DailyPlanningOptions`
- output type: `DailyPlanningModule`
- supported report: `daily`
The module should be initially equivalent to `TomorrowPlanning`, but independent
from it:
- do not reuse the public `TomorrowPlanningModule` type
- do not emit the `tomorrow_planning` stanza
- do not use `module.TomorrowPlanning` in the Daily default composition
Recommended initial fields should match Tomorrow planning:
- `morning_readiness`
- `commute_school_workday_concerns`
- `overnight_change_watch`
Private helper functions may be shared with Tomorrow planning when the
underlying logic is truly identical.
## Acceptance Criteria
The feature is complete when:
- `weatherreporter generate daily --date YYYY-MM-DD` runs the new Daily report.
- `weatherreporter generate daily` without `--date` fails with an actionable
error.
- `daily` report metadata, RunID content, artifact paths, data-package paths,
distributor template variables, generated-text assets, and rendered Markdown
all use report ID `daily`.
- The active report registry includes `daily` and does not include
`daily_today`.
- `reports.daily` is the implemented config override key for Daily.
- `reports.daily_today` is rejected rather than treated as an alias.
- Daily uses generated-text-template generation with prompt ID
`weather.daily_generated_text`.
- Daily uses dedicated schema, prompt, template, generated-text type,
render-context type, and planning-module surfaces.
- Daily rendered Markdown initially matches Tomorrow's report format.
- Daily data packages include `daily_planning`, not `tomorrow_planning`.
- Daily Recent Changes compare against prior Daily snapshots for the same valid
local date.
- Daily is not included in morning or evening scheduled batches.
- Existing Today and Tomorrow report semantics remain unchanged.
- Historical `daily_today` workspace artifacts are left untouched.
- Non-roadmap documentation is updated after implementation to describe only
implemented Daily behavior.
## Implementation Plan Reference
Use `docs/roadmap/implementation.md` for the staged implementation plan. This
feature roadmap intentionally does not define implementation stages, file-by-file
work packages, or validation commands so that implementing agents have a single
sequencing authority.
## Ambiguities Addressed
- Replacement scope: new `daily` replaces and removes old active
`daily_today`.
- Date behavior: `--date` is required; no default date is used.
- Output format: initial rendered Markdown matches Tomorrow.
- Internal separation: Daily has its own template, schema, prompt, generated
text type, render context, and planning module.
- Batch behavior: Daily is not scheduled; Today remains the morning current-day
scheduled product.
- Historical artifacts: old `daily_today` workspace files are not migrated.
## Open Decisions
No open decisions remain that block implementation.
Future decisions that should not be resolved in this roadmap:
- Whether Daily should eventually support recurring scheduled generation.
- Whether Daily should diverge from Tomorrow's template or planning logic.
- Whether old `daily_today` workspace artifacts should ever receive a migration
or inspection compatibility layer.

View File

@@ -1,361 +0,0 @@
# Data Package Export Roadmap
## Purpose
This roadmap defines the target state for cleaning up module fields exposed in
YAML data packages. The goal is to keep report templates composable while making
LLM prompt inputs concise, readable, and free of template-only helper fields.
This feature is implemented. Current data-package behavior is documented in
`docs/internal/prompt-input.md`; the rich module and template boundary is
documented in `docs/internal/module.md` and `docs/templates.md`.
## Problem
Module output structs currently serve two different consumers:
- deterministic report templates, which benefit from presentation helpers such
as lower-case text, display labels, trend phrases, and hour labels;
- Scriptorium data packages, which should expose the clearest useful weather
facts to the LLM with minimal redundancy.
Those consumers now need different surfaces. Examples include:
- `current_conditions` exposes both `condition_text` and
`condition_text_lower`, plus both abbreviated and long-form wind-direction
fields.
- `hourly_forecast.periods[]` exposes both `period_begins` and `hour_label`,
and both `text_description` and `text_description_lower`.
- `derived_daypart_summaries` exposes numerous temperature and condition phrase
fields that are useful for deterministic template wording but noisy in the
prompt data package.
The cleanup should not weaken template composability. Templates should still be
able to use rich module values and helper fields.
The cleanup applies to every report that consumes these modules, including the
generated-template `today`, `tomorrow`, and `daily` reports. The same
`derived_daypart_summaries` prompt export should serve all three reports while
their templates continue to use rich daypart helper fields.
## Intent
Data packages should be curated prompt inputs, not a raw dump of every field
available to Go templates.
The intended architecture is:
- module builders produce rich internal/template module values;
- each module may define a prompt-facing export value for data-package use;
- prompt input construction serializes the prompt-facing export value;
- template rendering continues to use the full rich module value.
The result should let `weatherreporter` optimize separately for:
- precise deterministic Markdown rendering;
- compact, readable LLM input;
- stable internal module contracts.
## Locked Decisions
- Do not make the existing module structs smaller solely to clean up data
packages.
- Do not use per-module string field allowlists as the primary mechanism.
- Do not rely on reflection-heavy field filtering for nested module shapes.
- Do not use `json:"-"` or `yaml:"-"` on rich template fields as the main
boundary.
- Keep rich module outputs available for template rendering, inspection, tests,
and internal use.
- Add an explicit prompt/data-package export layer for module outputs.
- Simple modules may use default pass-through export behavior.
- No compatibility aliases are needed for removed prompt-facing fields because
the prompt schema is still pre-release.
- Bump the data-package schema version when implementing this change.
- Compute prompt export values during module snapshot construction and store the
runtime-only prompt value on `module.Output` alongside the rich `Value`.
- Do not persist prompt export values in module snapshot JSON; module snapshots
should continue to preserve rich module values.
## Target Architecture
Each module definition should be able to declare how its output is represented
in prompt data packages.
A possible shape is:
```go
type ModuleDefinition struct {
// existing fields...
PromptExporter ModulePromptExporter
}
type ModulePromptExporter func(value any) (any, error)
```
The exact API may differ if implementation discovers a cleaner fit, but the
contract should preserve these properties:
- the exporter is owned near the module definition or module builder;
- the exporter receives the rich module value and returns a prompt-facing value;
- missing exporters default to pass-through for modules whose rich value is
already prompt-appropriate;
- exporter errors include module ID and stanza context;
- promptinput uses exported prompt values instead of rich values;
- render contexts and templates continue using rich values.
The preferred implementation should avoid making `internal/promptinput` import
`internal/briefing` directly. If prompt export needs registry knowledge, either:
- record the prompt-facing value in `module.Output` when the module snapshot is
built; or
- pass an explicit export map/registry into prompt-input construction without
creating a package cycle.
The implementation should keep package boundaries consistent with existing
architecture: module output policy belongs with module definitions, and data
package serialization belongs in `internal/promptinput`.
## Prompt Export Contract
A module prompt export should be:
- **curated:** include fields useful to the LLM, omit fields used only for
deterministic sentence construction;
- **typed:** use small prompt-facing structs for modules that need reshaping;
- **stable:** keep field names intentional and avoid duplicating equivalent
facts under multiple names;
- **readable:** prefer fields that explain themselves in YAML;
- **loss-aware:** do not omit facts that the LLM needs to reason about timing,
severity, uncertainty, or practical impact;
- **module-owned:** keep each module responsible for its own prompt-facing
contract.
Prompt-facing structs may live next to the module that owns them, for example:
```go
type CurrentConditionsPromptExport struct {
ConditionText string `json:"condition_text,omitempty"`
TemperatureF *int `json:"temperature_f,omitempty"`
ApparentTemperatureF *int `json:"apparent_temperature_f,omitempty"`
RelativeHumidityPercent *int `json:"relative_humidity_percent,omitempty"`
WindSpeedMph *int `json:"wind_speed_mph,omitempty"`
WindDirection string `json:"wind_direction,omitempty"`
}
```
The names do not need to include `PromptExport` if implementation finds a
clearer convention, but they should distinguish data-package shape from
template-rendering shape.
## Initial Cleanup Targets
### Current Conditions
Keep prompt-facing fields that express current observed conditions directly:
- `condition_text`
- `is_day`
- temperature fields
- apparent temperature fields
- dewpoint fields
- relative humidity
- wind speed
- one wind direction field
Remove prompt-facing fields that are template-only duplicates:
- `condition_text_lower`
- duplicate wind-direction text when an equivalent `wind_direction` field is
present
The template surface may keep those helper fields.
### Hourly Forecast
Keep prompt-facing period fields that carry facts:
- `period_begins`
- `period_ends`
- `name`
- `is_day`
- condition code, if useful
- `text_description`
- temperature fields
- dewpoint, apparent temperature, humidity, wind, gust, pressure, visibility,
cloud cover, precipitation probability, precipitation amount, snowfall depth,
and UV index when provided by upstream data
Remove prompt-facing fields that duplicate or encode template logic:
- `hour_label`, because `period_begins` already gives the time in a friendly
local label;
- `text_description_lower`, because the LLM can interpret
`text_description`;
- `mention_precipitation`, because it is a template threshold helper when the
underlying precipitation probability is present.
The template surface may keep these helper fields.
### Derived Daypart Summaries
Keep prompt-facing fields that describe the daypart:
- `date`
- `display_name`
- `period_begins`
- `period_ends`
- temperature range or the best single temperature phrase
- apparent temperature range when useful
- maximum precipitation probability and time
- maximum wind gust and time
- dominant condition
- temperature trend
- notable conditions
- weather indicator booleans
- relevant alert count
Remove prompt-facing fields that mainly support deterministic sentence
construction:
- duplicate lower-case/display variants of the same dominant condition;
- multiple temperature phrase fragments when a smaller set can express the same
trend;
- duplicate time labels where one friendly time field is enough.
The exact retained daypart temperature fields should be chosen during
implementation with template needs and LLM readability in mind. The prompt
export should preserve the facts needed to understand whether temperatures are
rising, falling, peaking, or steady, but it does not need every phrase fragment
used by the Markdown template.
### Other Modules
Most existing modules may initially use pass-through export unless they expose
clear template-only helpers. During implementation, review at least:
- `narrative_forecast`
- `precip_timing`
- `outdoor_windows`
- `alert_digest`
- `spc_convective_outlooks`
- `spc_convective_discussion`
- `area_forecast_discussion`
- `weather_story`
- planning modules
Do not remove fields merely because they are verbose. Remove or reshape fields
when they are redundant, template-specific, or confusing in the context of LLM
input.
## Data Package Behavior
After implementation:
- saved YAML data packages should use prompt-facing module exports;
- saved module snapshots should continue preserving rich module output values;
- generated-text render contexts should continue preserving rich module values;
- Recent Changes should continue using structured module snapshots unless a
specific comparison should intentionally move to prompt-facing fields;
- inspection commands should make clear whether they are showing rich module
snapshots or prompt data packages.
- generated-template reports, including `today`, `tomorrow`, and `daily`, should
continue rendering from rich module values.
This roadmap does not require changing source warnings, report metadata,
collected facts, derived facts, or generated report artifacts.
## Schema And Versioning
This is a prompt-input schema cleanup. Because the project is pre-release, the
implementation may make a clean break in data-package field names without
compatibility aliases.
The data-package schema version should be bumped when this feature is
implemented because persisted data-package fields will be removed or renamed.
This makes artifact shape changes explicit and helps inspection tooling
distinguish old and new data packages.
## Documentation Guidance
After implementation, update implemented documentation only:
- `docs/internal/module.md`: describe the distinction between rich module output
and prompt-facing export values.
- `docs/internal/prompt-input.md`: document that data packages use curated
prompt exports, not full template module structs.
- `docs/templates.md`: clarify that templates may have richer fields than the
data package.
- Any module field examples in implemented docs should match the new
prompt-facing data package shape.
Do not document future module fields or unimplemented exporters outside
`docs/roadmap/`.
## Acceptance Criteria
The feature is complete when:
- prompt data packages serialize curated module exports instead of blindly
serializing rich module values;
- templates still render from rich module values without losing helper fields;
- `current_conditions` no longer exposes lower-case condition text or duplicate
wind-direction fields in data packages;
- `hourly_forecast.periods[]` no longer exposes `hour_label`,
`text_description_lower`, or `mention_precipitation` in data packages;
- `derived_daypart_summaries` no longer exposes redundant condition and
temperature phrase variants in data packages;
- `today`, `tomorrow`, and `daily` rendered reports continue to have access to
rich daypart helper fields for deterministic template wording;
- simple modules that do not need cleanup still export correctly through default
pass-through behavior;
- exporter errors include module/stanza context;
- YAML category ordering remains unchanged;
- module snapshot artifacts remain rich enough for templates, inspection, and
regression diagnosis;
- tests prove that removed prompt-facing fields are absent from saved YAML data
packages and still available to templates where needed.
## Testing Expectations
Implementation should add or update focused tests for:
- module registry validation for prompt exporters, if exporters are registered
there;
- promptinput construction using exported prompt values;
- pass-through behavior for simple modules;
- custom exports for current conditions, hourly forecast, and daypart summaries;
- data-package YAML output rejecting stale fields;
- template render tests proving template-only helper fields remain available for
`today`, `tomorrow`, and `daily`;
- app workflow tests proving saved data packages use curated exports while
render contexts keep rich values.
Suggested validation after implementation:
```sh
go test ./internal/module ./internal/briefing ./internal/promptinput
go test ./internal/generatedtext ./internal/reporttemplate ./internal/app
go test ./...
go run ./cmd/weatherreporter --help
git diff --check
```
## Deferred Work
Do not include these in the initial cleanup unless implementation reveals they
are necessary:
- user-configurable data-package field selection;
- per-prompt custom field profiles;
- reflection-based generic include/exclude lists;
- automatic schema generation for data-package exports;
- changing collected facts or derived facts contracts;
- changing Scriptorium invocation behavior;
- changing generated Markdown templates beyond preserving their current output.
## Open Questions
No open questions block implementation.
The implementation plan in `docs/roadmap/implementation.md` is the sequencing
authority for this feature.

View File

@@ -1,657 +0,0 @@
# Data Package Export Implementation Roadmap
## Purpose
This roadmap defines the staged implementation plan for
`docs/roadmap/data-package-exports.md`. It is written for an LLM coding agent
that will implement the curated data-package export feature in order.
The goal is to separate rich module values used by templates from curated
prompt-facing values serialized into Scriptorium data packages. The change
should make YAML data packages smaller and clearer without reducing deterministic
template composability.
This is a planning document only. It may describe unimplemented behavior because
it lives under `docs/roadmap/`.
## Source Roadmap
Use `docs/roadmap/data-package-exports.md` as the feature roadmap and target
state authority. That document defines the user intent, module/export boundary,
initial cleanup targets, and non-goals.
If this implementation plan and the feature roadmap conflict, update the feature
roadmap first so it remains the conceptual source of truth, then update this
file.
## Locked Decisions
- Keep rich module values available for templates, module snapshots, inspection,
Recent Changes, and render contexts.
- Serialize curated prompt-facing module values into data packages.
- Store prompt-facing values on `module.Output` as runtime-only values.
- Do not persist prompt-facing values in module snapshot JSON.
- Keep `internal/promptinput` independent from `internal/briefing`.
- Add prompt export behavior to the module registry/definition path where module
output policy already lives.
- Default modules without custom exporters to pass-through prompt values.
- Bump the data-package schema version from `weatherreporter.data_package.v2` to
`weatherreporter.data_package.v3`.
- No compatibility aliases are required for removed prompt-facing fields.
- Do not add user-configurable field selection, field allowlists, reflection
filters, or prompt-specific field profiles in this implementation.
## Implementation Principles
- Preserve existing generated Markdown output.
- Preserve existing template render-context richness.
- Preserve existing module snapshot richness and JSON shape except for any
unavoidable schema-version-only change. The preferred approach is to omit
prompt values from module snapshot JSON entirely.
- Keep exporter code near the module that owns the shape.
- Use typed prompt-facing structs for modules that need reshaping.
- Keep simple modules on default pass-through behavior.
- Make exporter errors actionable and include module ID/stanza context.
- Keep YAML category ordering unchanged.
- Treat generated-template `today`, `tomorrow`, and `daily` reports as equal
consumers of rich template module values.
- Update implemented docs only after code behavior exists.
## Stage 1: Runtime Prompt Value Contract
### Goal
Add a runtime-only prompt value to module outputs while preserving rich module
snapshot persistence.
### Files To Inspect
- `internal/module/module.go`
- `internal/module/module_test.go`
- `internal/state/filesystem.go`
- `internal/state/filesystem_test.go`
- `internal/app/app.go`
- `internal/app/app_test.go`
### Implementation
- Add a runtime-only field to `module.Output`, for example:
```go
type Output struct {
ID ID `json:"id"`
StanzaName string `json:"stanzaName"`
Value any `json:"value"`
PromptValue any `json:"-"`
}
```
- Add a small helper in `internal/module`, for example:
```go
func (o Output) DataPackageValue() any
```
The helper should return `PromptValue` when non-nil and fall back to `Value`
otherwise. This fallback keeps tests and any manually built snapshots simple.
- Do not require `PromptValue` in `Snapshot.Validate`.
- Do not persist `PromptValue` to module snapshot JSON.
- Do not change `StanzaValue`; it should continue decoding rich `Value`.
### Acceptance Criteria
- Module snapshots still serialize rich module values under `value`.
- Module snapshots do not serialize `promptValue` or any equivalent field.
- Existing rich-module snapshot lookup and `StanzaValue` behavior remain
unchanged.
- Code has a single helper for choosing the data-package value from an output.
### Tests
- Add module tests proving:
- `DataPackageValue` uses `PromptValue` when set;
- `DataPackageValue` falls back to `Value`;
- marshaled snapshot JSON omits `PromptValue`;
- `StanzaValue` continues decoding rich `Value`.
Suggested focused command:
```sh
go test ./internal/module ./internal/state
```
### Prompt Size
Small enough for one implementation prompt.
## Stage 2: Module Registry Prompt Exporters
### Goal
Teach the module registry to attach prompt-facing values to outputs as modules
are built.
### Files To Inspect
- `internal/briefing/modules.go`
- `internal/briefing/modules_test.go`
- `internal/briefing/*_module.go`
- `internal/module/module.go`
### Implementation
- Add a prompt exporter type in `internal/briefing`, for example:
```go
type ModulePromptExporter func(value any) (any, error)
```
- Add `PromptExporter ModulePromptExporter` to `ModuleDefinition`.
- In `ModuleRegistry.BuildModule`, after builder output ID and stanza validation:
- if `PromptExporter` is nil, set `output.PromptValue = output.Value`;
- if `PromptExporter` is present, call it with `output.Value`;
- set `output.PromptValue` to the returned value;
- wrap exporter errors with module ID and stanza context.
- Add a typed exporter helper if it keeps module exporters concise, for example:
```go
func promptExporter[T any](fn func(T) (any, error)) ModulePromptExporter
```
The helper should convert `any` through JSON marshal/unmarshal or direct type
assertion only if it meaningfully reduces boilerplate without hiding errors.
- Do not make `internal/promptinput` import `internal/briefing`.
- Do not move module-specific export policy into `internal/promptinput`.
### Acceptance Criteria
- Every built module output has a non-nil data-package value.
- Modules without custom exporters pass through rich values.
- Custom exporter errors identify the module and stanza.
- Registry validation remains focused on definitions, duplicate IDs/stanzas,
supported reports, options, builders, and missing-data policy.
### Tests
- Add or update module registry tests for:
- default pass-through prompt values;
- custom exporter prompt values;
- exporter error wrapping;
- output ID/stanza validation still runs before or around export behavior;
- output `PromptValue` is not persisted in snapshot JSON.
Suggested focused command:
```sh
go test ./internal/briefing ./internal/module
```
### Prompt Size
Small enough for one implementation prompt.
## Stage 3: Promptinput Uses Exported Values And Schema V3
### Goal
Make data-package construction serialize prompt-facing values and bump the data
package schema version.
### Files To Inspect
- `internal/promptinput/package.go`
- `internal/promptinput/package_test.go`
- `internal/state/filesystem_test.go`
- `internal/app/app.go`
- `internal/app/app_test.go`
- `docs/roadmap/data-package-exports.md`
### Implementation
- Change `promptinput.SchemaVersion` to:
```go
const SchemaVersion = "weatherreporter.data_package.v3"
```
- Update `stanzasFromSnapshot` to use `output.DataPackageValue()` rather than
`output.Value`.
- Keep `BriefingStanzas` category grouping and ordering unchanged.
- Keep `LoadYAML` validation strict for the current schema version.
- Update tests and fixtures that assert `weatherreporter.data_package.v2`.
### Acceptance Criteria
- Saved data packages use schema version `weatherreporter.data_package.v3`.
- Data package stanzas use `PromptValue` when present.
- Data package stanzas fall back to rich `Value` when `PromptValue` is absent,
which keeps hand-built test snapshots and loaded rich snapshots usable.
- YAML category ordering remains unchanged.
- Module snapshots remain rich and unaffected by prompt export serialization.
### Tests
- Update promptinput tests for:
- schema version `v3`;
- prompt value preferred over rich value;
- fallback to rich value;
- deterministic categorized YAML output unchanged apart from stanza values and
schema version;
- unknown/uncategorized stanza behavior unchanged.
- Update state tests that load data packages.
Suggested focused command:
```sh
go test ./internal/promptinput ./internal/state
```
### Prompt Size
Small enough for one implementation prompt.
## Stage 4: Current Conditions And Hourly Forecast Exports
### Goal
Add custom prompt exports for the clearest noisy raw-data modules:
`current_conditions` and `hourly_forecast`.
### Files To Inspect
- `internal/briefing/current_conditions_module.go`
- `internal/briefing/hourly_forecast_module.go`
- `internal/briefing/base_modules_test.go`
- `internal/generatedtext/render_context.go`
- `internal/reporttemplate/templates/*.md.tmpl`
- `internal/app/app_test.go`
### Current Conditions Export Shape
The prompt-facing `current_conditions` export should keep:
- `condition_text`
- `is_day`
- `temperature_c`
- `temperature_f`
- `apparent_temperature_c`
- `apparent_temperature_f`
- `dewpoint_c`
- `dewpoint_f`
- `relative_humidity_percent`
- `wind_speed_kmh`
- `wind_speed_mph`
- `wind_direction`
It should remove:
- `condition_text_lower`
- `wind_direction_text`
The rich `CurrentConditionsModule` should keep those helper fields for
templates.
### Hourly Forecast Export Shape
The prompt-facing `hourly_forecast` export should keep module-level fields:
- `product`
- `issued_at`
- `updated_at`
- `source_location`
- `source_location_id`
- `periods`
Each prompt-facing hourly period should keep:
- `period_begins`
- `period_ends`
- `name`
- `is_day`
- `condition_code`
- `text_description`
- temperature fields
- dewpoint fields
- wind speed and gust fields
- `wind_direction`
- pressure fields
- visibility fields
- apparent temperature fields
- `cloud_cover_percent`
- `probability_of_precipitation_percent`
- precipitation amount fields
- snowfall depth fields
- `uv_index`
- `relative_humidity_percent`
Each prompt-facing hourly period should remove:
- `hour_label`
- `text_description_lower`
- `mention_precipitation`
The rich `HourlyForecastPeriod` should keep those helper fields for templates.
### Implementation
- Add prompt export structs near each module.
- Add exporter functions near each module.
- Register exporters in the default module definitions.
- Prefer straightforward field copying over reflection.
- Preserve `omitempty` behavior.
### Acceptance Criteria
- Saved data packages no longer include removed fields for current conditions or
hourly forecast.
- Rich module snapshots and render contexts still include template helper
fields.
- Existing hourly, today, tomorrow, and daily template output remains unchanged
wherever those templates consume current conditions or hourly forecast values.
### Tests
- Update module tests to prove prompt exports omit stale fields and keep
expected factual fields.
- Update template/render-context tests to prove helper fields remain available
to templates.
- Update app workflow tests to inspect saved YAML and reject:
- `condition_text_lower`;
- `wind_direction_text`;
- `hour_label`;
- `text_description_lower`;
- `mention_precipitation`.
Suggested focused command:
```sh
go test ./internal/briefing ./internal/generatedtext ./internal/reporttemplate ./internal/app
```
### Prompt Size
Medium. Suitable for one implementation prompt.
## Stage 5: Derived Daypart Summary Export
### Goal
Add a curated prompt export for `derived_daypart_summaries` while preserving the
rich daypart fields used by Daily, Today, and Tomorrow templates.
### Files To Inspect
- `internal/briefing/derived_daypart_summaries_module.go`
- `internal/briefing/derived_modules_test.go`
- `internal/generatedtext/render_context.go`
- `internal/reporttemplate/templates/daily.md.tmpl`
- `internal/reporttemplate/templates/today.md.tmpl`
- `internal/reporttemplate/templates/tomorrow.md.tmpl`
- `internal/reporttemplate/reporttemplate_test.go`
- `internal/app/app_test.go`
### Export Shape
For each daypart, the prompt-facing export should keep:
- `date`
- `display_name`
- `period_begins`
- `period_ends`
- `temp_range_f`
- `apparent_temp_range_f`
- `max_pop_percent`
- `max_pop_time`
- `mention_precipitation`
- `max_wind_gust_mph`
- `max_wind_gust_time`
- `dominant_condition`
- `temperature_trend`
- `temperature_start_phrase_f`
- `temperature_end_phrase_f`
- `temperature_peak_phrase_f`
- `temperature_steady_phrase_f`
- `notable_conditions`
- `snow`
- `ice`
- `fog`
- `heat`
- `cold`
- `wind`
- `relevant_alert_count`
The prompt-facing export should remove:
- `temperature_phrase_f`
- `dominant_condition_lower`
- `dominant_condition_display`
- `max_pop_time_label`
For `max_pop_time`, use the most readable existing time label. Prefer
`MaxPopTimeLabel` when present, falling back to `MaxPopTime`, while keeping the
prompt-facing field name `max_pop_time`.
Keep the temperature trend phrase fields even though only some are populated for
each trend. They are not duplicates when used according to the trend:
- rising/falling use start and end phrases;
- peaking uses peak phrase;
- steady uses steady phrase.
### Implementation
- Add prompt export structs near the daypart module.
- Add an exporter for the `map[string]DerivedDaypartSummaryModule` value.
- Preserve map keys and values for all emitted dayparts.
- Register the exporter in the default module definitions.
- Do not alter rich `DerivedDaypartSummaryModule` fields used by templates.
### Acceptance Criteria
- Saved data packages no longer include the removed daypart fields.
- The prompt-facing daypart export still gives the LLM enough information to
understand condition, temperature trend, precipitation, wind, notable
conditions, and alert relevance.
- Daily, Today, and Tomorrow template output remains unchanged.
- Render contexts still expose rich daypart helper fields.
### Tests
- Update daypart module tests for:
- prompt export field presence;
- removed field absence;
- rising, falling, peaking, and steady trend values;
- max PoP time using the friendly label under the stable `max_pop_time` key.
- Update template tests to prove rich helper fields still render Daily, Today,
and Tomorrow daypart wording.
- Update app data-package tests to reject stale daypart fields.
Suggested focused command:
```sh
go test ./internal/briefing ./internal/generatedtext ./internal/reporttemplate ./internal/app
```
### Prompt Size
Medium. Suitable for one implementation prompt.
## Stage 6: Pass-Through Review And Workflow Regression Tests
### Goal
Confirm all other modules export correctly through pass-through behavior and add
workflow-level coverage proving the new boundary.
### Files To Inspect
- `internal/briefing/*_module.go`
- `internal/briefing/modules.go`
- `internal/promptinput/package_test.go`
- `internal/app/app_test.go`
- `internal/generatedtext/render_context_test.go`
- `internal/reporttemplate/reporttemplate_test.go`
### Implementation
- Review remaining modules:
- `metadata`
- `narrative_forecast`
- `derived_daily_summary`
- `precip_timing`
- `outdoor_windows`
- `alert_digest`
- `spc_convective_outlooks`
- `spc_convective_discussion`
- `area_forecast_discussion`
- `weather_story`
- planning modules
- Keep pass-through behavior for modules that are already prompt-appropriate.
- Add custom exporters only if a module contains clear template-only helpers or
confusing duplicate fields.
- Do not broaden this stage into a general prompt-schema redesign.
- Add workflow assertions that:
- module snapshots contain rich values;
- render contexts contain rich values;
- data packages contain curated values;
- Scriptorium receives the curated data package path exactly as before.
- Include workflow coverage for generated-template `today`, `tomorrow`, and
`daily` reports when asserting daypart and render-context behavior.
### Acceptance Criteria
- Every default module either has a custom exporter or intentionally uses
pass-through.
- App workflow tests prove saved data packages omit the cleaned fields.
- App workflow tests prove rich helper fields remain available where templates
use them.
- App workflow or render-context tests include `daily` alongside `today` and
`tomorrow` for daypart helper coverage.
- Existing report generation behavior remains unchanged except for data-package
YAML content and schema version.
### Tests
Suggested focused command:
```sh
go test ./internal/briefing ./internal/promptinput ./internal/generatedtext ./internal/reporttemplate ./internal/app
```
### Prompt Size
Small to medium. Suitable for one implementation prompt.
## Stage 7: Documentation And Final Validation
### Goal
Update implemented documentation after the code exists and run full validation.
### Files To Inspect
- `docs/internal/module.md`
- `docs/internal/prompt-input.md`
- `docs/templates.md`
- `docs/operations.md`
- `docs/troubleshooting.md`
- `docs/roadmap/data-package-exports.md`
- tests that assert documentation examples or data-package snippets
### Implementation
- Update `docs/internal/module.md` to describe:
- rich module output;
- runtime prompt export values;
- pass-through exporters;
- module-owned prompt export policy.
- Update `docs/internal/prompt-input.md` to document:
- schema version `weatherreporter.data_package.v3`;
- curated module export behavior;
- category ordering unchanged;
- data packages are not full template render contexts.
- Update `docs/templates.md` to clarify that templates can use richer fields
than the data package exposes.
- Update any Daily template-variable documentation alongside Today and Tomorrow
references where the same module fields are discussed.
- Update any implemented docs containing stale examples of removed data-package
fields.
- Do not document deferred configurable field profiles or reflection filters as
implemented behavior.
### Acceptance Criteria
- Non-roadmap docs describe only implemented data-package export behavior.
- Docs do not imply data packages contain template-only helper fields.
- Examples and snippets use schema version `v3` when they show data packages.
### Validation Commands
```sh
go test ./internal/module ./internal/briefing ./internal/promptinput
go test ./internal/generatedtext ./internal/reporttemplate ./internal/app
go test ./...
go run ./cmd/weatherreporter --help
git diff --check
```
### Stale Field Greps
Run focused checks against docs and expected saved data-package fixtures/snippets:
```sh
rg -n "condition_text_lower|wind_direction_text|hour_label|text_description_lower|mention_precipitation|dominant_condition_lower|dominant_condition_display|max_pop_time_label|temperature_phrase_f" docs internal/*/*_test.go
```
Review matches manually. Some matches should remain in rich module structs,
template tests, and render-context tests; they should not remain as expected
data-package fields.
### Prompt Size
Medium. Suitable for one implementation prompt.
## Deferred Work
Do not include these in this implementation sequence:
- user-configurable data-package field selection;
- per-prompt custom field profiles;
- reflection-based generic include/exclude lists;
- automatic schema generation for data-package exports;
- changing collected facts or derived facts contracts;
- changing Scriptorium invocation behavior;
- changing generated Markdown templates beyond preserving their output;
- migrating old workspace data-package artifacts.
## Open Questions
No open questions block implementation.
The implementation plan intentionally locks in the recommended choices from the
feature roadmap: schema version bump to `weatherreporter.data_package.v3`, and
runtime-only prompt export values stored on `module.Output`.
## Global Validation Checklist
- `go test ./...` passes.
- `go run ./cmd/weatherreporter --help` is accurate.
- `git diff --check` passes.
- Saved data packages use schema version `weatherreporter.data_package.v3`.
- Saved data packages serialize prompt-facing module exports.
- Saved module snapshots serialize rich module values and do not persist
prompt-facing values.
- Render contexts continue to expose rich module values.
- Hourly, Today, Tomorrow, and Daily templates continue to render the same
Markdown output.
- YAML briefing category order remains unchanged.
- `current_conditions` data packages omit `condition_text_lower` and
`wind_direction_text`.
- `hourly_forecast.periods[]` data packages omit `hour_label`,
`text_description_lower`, and `mention_precipitation`.
- `derived_daypart_summaries` data packages omit `temperature_phrase_f`,
`dominant_condition_lower`, `dominant_condition_display`, and
`max_pop_time_label`.
- Non-roadmap docs describe only implemented behavior.