Upload 3 files

README.md

@@ -1,12 +1,127 @@
----
-title: JSON Logic Rule Generator
-emoji: 📈
-colorFrom: green
-colorTo: purple
-sdk: docker
-pinned: false
-license: apache-2.0
-short_description: JSON_Logic_Rule_Generator
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# JSON Logic Rule Generator (with RAG & Embeddings)
+
+An AI-powered FastAPI service that converts natural-language prompts into valid JSON Logic rules using embeddings, hybrid key mapping, and a lightweight RAG layer over domain policies. The API also returns a plain-English explanation, the keys used, and a confidence score for each generated rule.
+
+## Features
+
+- Natural language → JSON Logic rule generation
+- Uses only predefined `SAMPLE_STORE_KEYS` in `{"var": "key"}` format
+- Hybrid key mapping (embeddings + BM25 + Reciprocal Rank Fusion)
+- Lightweight RAG over hard-coded policy documents (FAISS + CRAG-style refinement)
+- Self-consistency: multiple rule variants with voting + mock-data validation
+- Confidence score based on key similarity, policy relevance, and generation quality
+- Utility endpoints to inspect keys and policies
+- Ready for local dev, UV, or Docker usage
+
+## Prerequisites
+
+- Python 3.10 or higher
+- pip or uv for dependency management
+- An OpenAI API key (for GPT-4o-mini)
+
+## Installation & Setup
+
+### 1. Clone the Repository
+
+```
+git clone https://github.com/<your-username>/<your-repo>.git
+cd jsonlogic-rag-api
+```
+
+
+### 2. Create a Python Environment
+
+```
+python -m venv venv
+```
+
+
+### 3. Activate the Environment and Install Requirements
+
+#### On Windows
+
+```
+venv\Scripts\activate
+```
+
+
+#### On macOS / Linux
+
+```
+source venv/bin/activate
+```
+
+
+#### Install Requirements
+
+```
+pip install -r requirements.txt
+```
+
+
+### 4. Configure Environment Variables
+
+Create a `.env` file in the project root:
+
+```
+OPENAI_API_KEY="sk-your-key-here"
+EMBED_MODEL="all-MiniLM-L6-v2"
+RRF_K="60"
+SIM_THRESHOLD="0.7"
+```
+
+
+### 5. Run the API with Uvicorn
+
+```
+python run.py
+```
+
+
+The API will start at:
+
+- Swagger UI: http://127.0.0.1:8000/docs  
+- Health check: http://127.0.0.1:8000/health  
+
+### Or you can use Uvicorn directly
+
+```
+uvicorn app.main:app --host 127.0.0.1 --port 8000 --reload
+```
+
+### 6. Use below endpoint to get results. Here is the example input
+
+```
+http://127.0.0.1:8000/generate-rule
+```
+
+Example input
+
+```
+{
+  "context_docs": [
+    "Custom policy: Minimum age 25"
+  ],
+  "prompt": "Approve if bureau score > 700 and business vintage at least 3 years"
+}
+```
+
+
+
+The response will include:
+
+- `json_logic`: the generated JSON Logic rule
+- `explanation`: short natural-language explanation
+- `used_keys`: which fields were used
+- `key_mappings`: how user phrases mapped to keys
+- `confidence_score`: overall confidence (0–1)
+
+
+
+### Test
+
+To test it with some examples there are tests written in "tests" folder . To test just run below command
+
+```
+python tests\test_api.py
+```

requirements.txt

@@ -0,0 +1,17 @@
+fastapi 
+uvicorn[standard] 
+sentence-transformers 
+faiss-cpu 
+openai 
+pydantic-settings 
+rank-bm25 
+scikit-learn 
+pytest 
+pytest-asyncio 
+pytest-cov 
+python-dotenv 
+numpy 
+structlog
+json-logic-qubit
+httpx
+loguru

run.py

@@ -0,0 +1,17 @@
+# run.py
+import uvicorn
+import sys
+from pathlib import Path
+
+# Add project root to path
+ROOT_DIR = Path(__file__).parent
+sys.path.insert(0, str(ROOT_DIR))
+
+if __name__ == "__main__":
+    uvicorn.run(
+        "app.main:app",
+        host="127.0.0.1",  # Changed to localhost for local testing
+        port=8000,
+        reload=True,
+        log_level="info"
+    )