p59f
This commit is contained in:
team 1
@@ -0,0 +1,58 @@
|
||||
# RetrieX Patch 59E - Genre Validation Warning Cleanup
|
||||
|
||||
## Purpose
|
||||
|
||||
Patch 59E removes validation false positives that appeared after the genre.yaml source-of-truth cleanup.
|
||||
|
||||
The affected warnings were caused by comparing legacy/raw YAML inventory values against effective runtime values that are now intentionally resolved through genre.yaml or the effective vocabulary facade.
|
||||
|
||||
## Changes
|
||||
|
||||
- `RetriexEffectiveConfigProvider::validateRetrieval()` now skips legacy inventory drift warnings for retrieval exact-selection keys when those keys are backed by `genre.configuration_values.retrieval_and_language.exact_selection`.
|
||||
- `RetriexEffectiveConfigProvider::validateVocabulary()` now checks protected short-model tokens against the active vocabulary view, not only the raw legacy `vocabulary.yaml` add-list.
|
||||
|
||||
## Scope
|
||||
|
||||
This is a validation/audit cleanup only.
|
||||
|
||||
No changes were made to:
|
||||
|
||||
- retrieval ranking
|
||||
- vector/keyword retrieval
|
||||
- prompt building
|
||||
- shop query generation
|
||||
- intent detection
|
||||
- search repair
|
||||
- response generation
|
||||
- genre.yaml values
|
||||
|
||||
## Why this is safe
|
||||
|
||||
The exact-selection warning was a false positive: the active retriever config reads the exact-selection lists from `genre.yaml`, while `retrieval.inventory` still represents the legacy/raw parameter snapshot.
|
||||
|
||||
The protected-short-model warning was also a false positive: `vocabulary.views.retrieval.important_short_model_tokens.add` is intentionally empty after the genre migration, while the active `DomainVocabularyConfig` view resolves the required tokens from `genre.yaml`.
|
||||
|
||||
## Expected result
|
||||
|
||||
After this patch, the following previously reported warnings should disappear:
|
||||
|
||||
- `retrieval.inventory.exact_selection_* differs from active retriever config`
|
||||
- `vocabulary.views.retrieval.important_short_model_tokens should contain protected token ...`
|
||||
|
||||
## Required checks
|
||||
|
||||
Run:
|
||||
|
||||
```bash
|
||||
bin/console mto:agent:config:validate
|
||||
bin/console mto:agent:regression:test
|
||||
bin/console mto:agent:config:audit-source --details
|
||||
bin/console mto:agent:config:audit-patterns --details
|
||||
```
|
||||
|
||||
Expected:
|
||||
|
||||
- config validation OK without the p59D false-positive warnings
|
||||
- regression baseline OK
|
||||
- source audit OK
|
||||
- pattern audit OK
|
||||
@@ -0,0 +1,78 @@
|
||||
# RetrieX Patch 59F - Genre Review Group Metadata
|
||||
|
||||
## Purpose
|
||||
|
||||
Patch 59F continues the p59 cleanup after the p59A-p59E genre migration work.
|
||||
It does not remove remaining legacy/runtime paths yet. Instead, it makes every
|
||||
remaining `genre.yaml` review group machine-readable so the next cleanup patches
|
||||
can decide per category whether a path should be rewired, kept as technical
|
||||
configuration, kept as runtime-resolved configuration, or kept as governance.
|
||||
|
||||
## Changes
|
||||
|
||||
- Adds machine-readable metadata to every `genre.adaptation_surface.*.review_path_groups.*` group:
|
||||
- `classification`
|
||||
- `source_state`
|
||||
- `cleanup_action`
|
||||
- Extends `validateGenre()` so declared review groups must now include this metadata.
|
||||
- Keeps all existing `paths` entries unchanged.
|
||||
- Keeps all governance hashes unchanged.
|
||||
|
||||
## Review group classifications
|
||||
|
||||
Current classifications:
|
||||
|
||||
- `legacy_compatibility_view`
|
||||
- `legacy_runtime_helper`
|
||||
- `technical_regex_template`
|
||||
- `legacy_signal_fallback`
|
||||
- `legacy_context_fallback`
|
||||
- `technical_cleanup_profile`
|
||||
- `technical_shop_mapping`
|
||||
- `runtime_resolved_connection`
|
||||
- `governance_guardrail`
|
||||
|
||||
Current source states:
|
||||
|
||||
- `legacy_frozen_non_empty`
|
||||
- `legacy_runtime_resolved_allowed`
|
||||
|
||||
## Runtime impact
|
||||
|
||||
No runtime behavior change is intended.
|
||||
|
||||
This patch does not change:
|
||||
|
||||
- retrieval ranking
|
||||
- prompt building
|
||||
- shop query logic
|
||||
- search repair logic
|
||||
- intent routing
|
||||
- scoring
|
||||
- genre values
|
||||
- governance hash values
|
||||
|
||||
## Local checks
|
||||
|
||||
Run in a dependency-complete project checkout:
|
||||
|
||||
```bash
|
||||
bin/console mto:agent:config:validate
|
||||
bin/console mto:agent:regression:test
|
||||
bin/console mto:agent:config:audit-source --details
|
||||
bin/console mto:agent:config:audit-patterns --details
|
||||
```
|
||||
|
||||
In the ZIP-only local workspace, `vendor/` was missing, so Symfony console checks
|
||||
could not be executed here. Local structural checks were run instead:
|
||||
|
||||
- PHP lint for `src/Config/RetriexEffectiveConfigProvider.php`
|
||||
- YAML parse for `config/retriex/genre.yaml`
|
||||
- Review group metadata completeness simulation
|
||||
|
||||
## Next step
|
||||
|
||||
A later p59G can start removing or rewiring one classification group at a time.
|
||||
Good candidates are the low-risk legacy compatibility/signal fallback groups,
|
||||
while technical regex templates, shop mapping, runtime connection values and
|
||||
governance guardrails should remain separate until dedicated cleanup patches.
|
||||
Reference in New Issue
Block a user