rti-ethics-engine 2.0.0__tar.gz → 2.0.1__tar.gz

rti_ethics_engine-2.0.1/INPUTS_OUTPUTS.md

@@ -0,0 +1,62 @@
+# RTI Ethics Engine: Input & Output Specification
+
+Detailed API reference for the package components.
+
+## 1. Input: Metrics Dictionary Shape
+
+The primary input to `evaluate(metrics)` is a nested dictionary structured by logical dimensions.
+
+### Dimensions & Metric Keys
+
+| Dimension          | Key              | L1 Metrics (Gates)        | L2 Metrics (Risk)      | L3 Metrics (Forensic) |
+| ------------------ | ---------------- | ------------------------- | ---------------------- | --------------------- |
+| **Fairness**       | `fairness`       | `spd`, `di`, `eop`, `aod` | `fnrd`, `refusal_disp` | `mutual_info`         |
+| **Drift**          | `drift`          | `psi`, `ks`               | `pred_drift`           | `embed_drift`         |
+| **Explainability** | `explainability` | `confidence`              | `support_ratio`        | `stability`           |
+| **Legal**          | `legal`          | `lawfulness`, `cvr`       | `proportionality`      | `explanation`         |
+| **Performance**    | `performance`    | `accuracy`, `recall`      | `f1`                   | `group_f1`            |
+| **UX**             | `ux`             | `tcr`, `sla`              | `cognitive_load`       | `dtc`                 |
+| **Security**       | `security`       | `adv_acc`                 | `pisr`                 | `mir`                 |
+| **Monitoring**     | `monitoring`     | `alc`                     | `pvr`                  | `irf`                 |
+| **Environment**    | `environmental`  | `inf_cost`                | `gpu_hours`            | `compute_intensity`   |
+
+---
+
+## 2. Dynamic Computation: `build_metrics_from_raw_data`
+
+Used to convert raw arrays and runtime data into the metrics dictionary above.
+
+### Parameters
+
+| Name                   | Type         | Description                                 |
+| ---------------------- | ------------ | ------------------------------------------- |
+| `query`                | `str`        | User input string                           |
+| `response`             | `str`        | AI response string                          |
+| `predictions`          | `np.ndarray` | Binary classification array (0/1)           |
+| `labels`               | `np.ndarray` | Ground truth array (0/1)                    |
+| `sensitive_attr`       | `np.ndarray` | Sensitive group indicator (0/1)             |
+| `latency_ms`           | `float`      | Response time in milliseconds               |
+| `output_tokens`        | `int`        | Token count of AI response                  |
+| `grounding_score`      | `float`      | Score from grounding judge (0-1)            |
+| `baseline_predictions` | `np.ndarray` | (Optional) Predictions for drift comparison |
+
+---
+
+## 3. Output: `EthicsResult` Schema
+
+The object returned by evaluation gives detailed access to the 3-layer audit results.
+
+### Summary Fields
+
+- `final_label`: `ETHICAL` (Pass), `CONDITIONALLY_ETHICAL` (Review), `UNETHICAL` (Block)
+- `s_llm`: Balanced audit score [0 to 1]
+- `risk_band`: `LOW`, `MEDIUM`, `HIGH`, `CRITICAL`
+- `layer1_status`: `PASS` (All gates green) or `FAIL`
+
+### Accessor Methods
+
+- `get_l1_metric(dim, key)`: Returns raw float value (e.g., `0.021`)
+- `get_l2_risk(dim)`: Returns normalized risk [0 to 1]
+- `get_l3_evidence(dim)`: Returns forensic explanation string or dictionary
+- `get_failed_dims()`: List of dimension names that failed L1 gates
+- `to_json()`: Serializes the entire audit trail to JSON

{rti_ethics_engine-2.0.0 → rti_ethics_engine-2.0.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: rti-ethics-engine
-Version: 2.0.0
+Version: 2.0.1
 Summary: India's first granular AI ethics framework for RTI applications (36 metrics, 9 dimensions)
 Project-URL: Homepage, https://github.com/rti-ethics/rti-ethics-engine
 Project-URL: Documentation, https://rti-ethics-engine.readthedocs.io
@@ -144,8 +144,8 @@ audit_json = result.to_json()
 
 For detailed input/output schemas and advanced configuration, see:
 
-- [INPUTS_OUTPUTS.md](INPUTS_OUTPUTS.md) - API Reference
-- [FORUMULAS.md](rti_ethics_engine/calculators.py) - Mathematical Manifest
+- [API Reference (INPUTS_OUTPUTS.md)](https://github.com/Krishil-Parikh/bazinga_tsec_hacks/blob/krishil/rti_ethics_engine/INPUTS_OUTPUTS.md)
+- [Mathematical Manifest (calculators.py)](https://github.com/Krishil-Parikh/bazinga_tsec_hacks/blob/krishil/rti_ethics_engine/rti_ethics_engine/calculators.py)
 
 ## 🧪 Testing
 

{rti_ethics_engine-2.0.0 → rti_ethics_engine-2.0.1}/README.md

@@ -86,8 +86,8 @@ audit_json = result.to_json()
 
 For detailed input/output schemas and advanced configuration, see:
 
-- [INPUTS_OUTPUTS.md](INPUTS_OUTPUTS.md) - API Reference
-- [FORUMULAS.md](rti_ethics_engine/calculators.py) - Mathematical Manifest
+- [API Reference (INPUTS_OUTPUTS.md)](https://github.com/Krishil-Parikh/bazinga_tsec_hacks/blob/krishil/rti_ethics_engine/INPUTS_OUTPUTS.md)
+- [Mathematical Manifest (calculators.py)](https://github.com/Krishil-Parikh/bazinga_tsec_hacks/blob/krishil/rti_ethics_engine/rti_ethics_engine/calculators.py)
 
 ## 🧪 Testing
 

{rti_ethics_engine-2.0.0 → rti_ethics_engine-2.0.1}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "rti-ethics-engine"
-version = "2.0.0"
+version = "2.0.1"
 description = "India's first granular AI ethics framework for RTI applications (36 metrics, 9 dimensions)"
 readme = "README.md"
 license = {file = "LICENSE"}
@@ -62,12 +62,14 @@ Issues = "https://github.com/rti-ethics/rti-ethics-engine/issues"
 
 [tool.hatch.build.targets.wheel]
 packages = ["rti_ethics_engine"]
+include = ["INPUTS_OUTPUTS.md"]
 
 [tool.hatch.build.targets.sdist]
 include = [
     "/rti_ethics_engine",
     "/tests",
     "/README.md",
+    "/INPUTS_OUTPUTS.md",
     "/LICENSE"
 ]
 

{rti_ethics_engine-2.0.0 → rti_ethics_engine-2.0.1}/rti_ethics_engine/__init__.py

@@ -81,7 +81,7 @@ from .calculators import (
     MetricCalculator
 )
 
-__version__ = "2.0.0"  # Major version bump for dynamic computation
+__version__ = "2.0.1"  # Major version bump for dynamic computation
 __author__ = "RTI Ethics Engine Team"
 __license__ = "MIT"