feat: add generated scene skill platform hardening
This commit is contained in:
@@ -0,0 +1,71 @@
|
||||
# G1-E Second Sample Reuse Report
|
||||
|
||||
> Date: 2026-04-18
|
||||
> Scope: `G1-E` second-sample reuse validation
|
||||
|
||||
## Summary
|
||||
|
||||
`G1-E` has moved from a single P0 anchor to a reusable family baseline.
|
||||
|
||||
This round added a second repo-local `G1-E` expansion fixture and promoted the result into the P1 family assets. The expansion fixture keeps the same family contract as the P0 sample:
|
||||
|
||||
1. one resolved main request
|
||||
2. lightweight enrichment requests
|
||||
3. a merge plan with join keys, aggregate rules, and output columns
|
||||
|
||||
The result validates that `G1-E` should remain distinct from plain `single_request_table`, while still staying below the `G6/G7/G8` boundary.
|
||||
|
||||
## Assets Updated
|
||||
|
||||
1. `tests/fixtures/generated_scene/g1e_light_enrichment_expansion/index.html`
|
||||
2. `tests/fixtures/generated_scene/p1_family_manifest.json`
|
||||
3. `tests/fixtures/generated_scene/p1_family_results.json`
|
||||
4. `tests/scene_generator_test.rs`
|
||||
5. `tests/scene_generator_p1_family_test.rs`
|
||||
6. `tests/scene_generator_family_policy_test.rs`
|
||||
|
||||
## Representative And Expansion Baseline
|
||||
|
||||
| Role | Fixture | Expected Archetype |
|
||||
| --- | --- | --- |
|
||||
| Representative | `tests/fixtures/generated_scene/g1e_light_enrichment` | `single_request_enrichment` |
|
||||
| Expansion | `tests/fixtures/generated_scene/g1e_light_enrichment_expansion` | `single_request_enrichment` |
|
||||
|
||||
The expansion baseline asserts:
|
||||
|
||||
1. main request contains `getWkorderAll`
|
||||
2. enrichment request contains `queryMeterInfo`
|
||||
3. merge join key contains `wkOrderNo`
|
||||
4. merge aggregate rules contain `group_by:countyCodeName`
|
||||
5. output columns contain `meterCapacityThisMonth`
|
||||
|
||||
## Validation
|
||||
|
||||
Executed:
|
||||
|
||||
```powershell
|
||||
cargo test --test scene_generator_test -- --nocapture
|
||||
cargo test --test scene_generator_p1_family_test -- --nocapture
|
||||
cargo test --test scene_generator_family_policy_test -- --nocapture
|
||||
cargo test --test scene_generator_canonical_test -- --nocapture
|
||||
```
|
||||
|
||||
All four regression suites passed.
|
||||
|
||||
## Result
|
||||
|
||||
`G1-E` reuse is now validated at family level:
|
||||
|
||||
1. `representativeRuns = 1`
|
||||
2. `expansionRuns = 1`
|
||||
3. `passedRuns = 2`
|
||||
4. `failedRuns = 0`
|
||||
5. `successRate = 1.0`
|
||||
|
||||
## Next Step
|
||||
|
||||
Per the frozen roadmap order, the next implementation line is `G6`:
|
||||
|
||||
1. start independent `G6` spec / plan
|
||||
2. use `电能表现场检验完成率指标报表` as the first boundary sample
|
||||
3. build fail-closed classification for host-bridge multi-step workflow before attempting runnable generation
|
||||
Reference in New Issue
Block a user