Add external-truth Yogi finance hook

2026-06-28 10:00:54 +08:00
parent e18cb268a6
commit 0bee529579
3 changed files with 221 additions and 1 deletions
@@ -0,0 +1,83 @@
+# Wealth Adjudicator Fourth Pass Yogi Hook Audit (2026-06-28)
+
+## Scope
+
+This pass closes the smallest safe version of the external-truth Yogi hook for the finance adjudicator.
+
+Guardrail kept intact:
+
+- Do not calculate `yogi_planet` inside this repository.
+- Only accept `external_truth.yogi_planet` from upstream truth sources.
+- Only allow the hook to influence finance promise folding when the supplied Yogi planet passes a lightweight D1 placement filter.
+
+## Root Cause Found
+
+The first failing test was not exposing a product bug. It was exposing a bad fixture:
+
+- the test expected a Yogi uplift,
+- but the supplied `Venus` was in house `11`,
+- while the agreed lightweight gate only allows Kendra/Trikona houses: `1, 4, 5, 7, 9, 10`.
+
+So the finance pipeline was correctly rejecting the hook.
+
+## Implemented Behavior
+
+### Positive path
+
+When all of the following are true:
+
+1. `external_truth.yogi_planet` is present
+2. matching D1 planet data is present in `modules.chart.planets`
+3. the planet is in `1/4/5/7/9/10`
+4. the status is not debilitated
+
+Then:
+
+- `present_evidence.yogi_promise` is populated
+- `wealth_promise_strength.supporting_sources` gains `"yogi"`
+- `source_diversity` increases accordingly
+- `count` increases by `1`
+- `primary_source` upgrades to `dhana_yogi_hooks` when the combined sources are `dhana + yogi`
+- `secondary_context` gains `yogi_active` when a finance dominant label already exists
+
+### Negative path
+
+If the externally supplied Yogi planet falls outside Kendra/Trikona, the hook does nothing:
+
+- `yogi_promise = None`
+- no wealth promise promotion
+- no `yogi_active` secondary context
+
+## Regression Coverage
+
+Added / confirmed:
+
+1. Yogi hook activates only when external truth is present and D1 placement passes the filter.
+2. Yogi hook does not activate for house `11`.
+3. Existing finance adjudicator regressions remain green.
+
+## Verification
+
+Command:
+
+```bash
+python3 -m pytest tests/test_mcp_strict_workflow_finance.py -q
+```
+
+Observed:
+
+- `9 passed`
+
+Manual evidence probe also confirmed:
+
+- house `10` case -> `dhana_yogi_hooks` + `yogi_active`
+- house `11` case -> unchanged `dhana_yogas` promise fold and no Yogi secondary context
+
+## Next Recommended Step
+
+Keep the same non-destructive pattern and extend only one notch:
+
+1. allow externally supplied Yogi truth to enrich `wealth_promise_strength`
+2. do not compute Yogi internally
+3. next, add a very small `Yogi` source-aware bump only through source structure, not direct verdict jumping
+4. after that, move to `dominant_label + secondary_context` refinement for finance edge cases