rnsr 0.1.0__py3-none-any.whl

rnsr-0.1.0.dist-info/METADATA

@@ -0,0 +1,592 @@
+Metadata-Version: 2.4
+Name: rnsr
+Version: 0.1.0
+Summary: Recursive Neural-Symbolic Retriever - Hierarchical document retrieval with font-based structure analysis
+Author: RNSR Contributors
+License-Expression: MIT
+Project-URL: Homepage, https://github.com/theeufj/RNSR
+Project-URL: Documentation, https://github.com/theeufj/RNSR#readme
+Project-URL: Repository, https://github.com/theeufj/RNSR.git
+Project-URL: Issues, https://github.com/theeufj/RNSR/issues
+Keywords: rag,retrieval,document-processing,llm,hierarchical-indexing,pdf-parsing,neural-symbolic
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: Intended Audience :: Science/Research
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Programming Language :: Python :: 3.14
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Topic :: Text Processing :: Indexing
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: pymupdf>=1.23.0
+Requires-Dist: pdfplumber>=0.10.0
+Requires-Dist: numpy>=1.24.0
+Requires-Dist: scipy>=1.10.0
+Requires-Dist: structlog>=23.0.0
+Requires-Dist: python-dotenv>=1.0.0
+Requires-Dist: llama-index>=0.10.0
+Requires-Dist: langgraph>=0.2.0
+Requires-Dist: langchain-core>=0.2.0
+Requires-Dist: transformers>=4.35.0
+Requires-Dist: torch>=2.0.0
+Requires-Dist: torchvision>=0.15.0
+Requires-Dist: pillow>=10.0.0
+Requires-Dist: pdf2image>=1.16.0
+Requires-Dist: scikit-learn>=1.3.0
+Requires-Dist: pymupdf>=1.23.0
+Provides-Extra: openai
+Requires-Dist: openai>=1.0.0; extra == "openai"
+Requires-Dist: llama-index-llms-openai>=0.1.0; extra == "openai"
+Requires-Dist: llama-index-embeddings-openai>=0.1.0; extra == "openai"
+Provides-Extra: anthropic
+Requires-Dist: anthropic>=0.20.0; extra == "anthropic"
+Requires-Dist: llama-index-llms-anthropic>=0.1.0; extra == "anthropic"
+Provides-Extra: gemini
+Requires-Dist: google-genai>=1.0.0; extra == "gemini"
+Provides-Extra: all
+Requires-Dist: rnsr[anthropic,gemini,openai]; extra == "all"
+Provides-Extra: benchmarks
+Requires-Dist: datasets>=2.0.0; extra == "benchmarks"
+Requires-Dist: ragas>=0.1.0; extra == "benchmarks"
+Provides-Extra: demo
+Requires-Dist: gradio>=4.0.0; extra == "demo"
+Provides-Extra: dev
+Requires-Dist: pytest>=7.0.0; extra == "dev"
+Requires-Dist: pytest-cov>=4.0.0; extra == "dev"
+Requires-Dist: pytest-asyncio>=0.21.0; extra == "dev"
+Requires-Dist: mypy>=1.0.0; extra == "dev"
+Requires-Dist: ruff>=0.1.0; extra == "dev"
+Dynamic: license-file
+
+# RNSR - Recursive Neural-Symbolic Retriever
+
+A state-of-the-art document retrieval system that preserves hierarchical structure for superior RAG performance. Combines PageIndex, Recursive Language Models (RLM), Knowledge Graphs, and Tree of Thoughts navigation.
+
+## Overview
+
+RNSR combines neural and symbolic approaches to achieve accurate document understanding:
+
+- **Font Histogram Algorithm** - Automatically detects document hierarchy from font sizes (no training required)
+- **Skeleton Index Pattern** - Lightweight summaries with KV store for efficient retrieval
+- **Tree-of-Thoughts Navigation** - LLM reasons about document structure to find answers
+- **RLM Unified Extraction** - LLM writes extraction code, grounded in actual text
+- **Knowledge Graph** - Entity and relationship storage for cross-document linking
+- **Self-Reflection Loop** - Iterative answer improvement through self-critique
+- **Adaptive Learning** - System learns from your document workload over time
+
+## Key Features
+
+| Feature | Description |
+|---------|-------------|
+| **Hierarchical Extraction** | Preserves document structure (sections, subsections, paragraphs) |
+| **RLM Unified Extractor** | LLM writes extraction code + ToT validation (grounded, no hallucination) |
+| **Provenance System** | Every answer traces back to exact document citations |
+| **LLM Response Cache** | Semantic-aware caching for 10x cost/speed improvement |
+| **Self-Reflection** | Iterative self-correction improves answer quality |
+| **Reasoning Memory** | Learns successful query patterns for faster future queries |
+| **Query Clarification** | Detects ambiguous queries and asks clarifying questions |
+| **Table/Chart Parsing** | SQL-like queries over tables, chart trend analysis |
+| **Adaptive Learning** | 6 registries that learn from usage and persist to disk |
+| **Multi-Document Detection** | Automatically splits bundled PDFs |
+| **Vision Mode** | OCR-free analysis for scanned documents and charts |
+
+## Installation
+
+```bash
+# Clone the repository
+git clone https://github.com/theeufj/RNSR.git
+cd RNSR
+
+# Create virtual environment
+python -m venv .venv
+source .venv/bin/activate  # On Windows: .venv\Scripts\activate
+
+# Install with all LLM providers
+pip install -e ".[all]"
+
+# Or install with specific provider
+pip install -e ".[openai]"      # OpenAI only
+pip install -e ".[anthropic]"   # Anthropic only
+pip install -e ".[gemini]"      # Google Gemini only
+```
+
+## Quick Start
+
+### 1. Set up API keys
+
+Create a `.env` file:
+
+```bash
+cp .env.example .env
+# Edit .env with your API keys
+```
+
+```env
+# Choose your preferred LLM provider
+OPENAI_API_KEY=sk-...
+# or
+ANTHROPIC_API_KEY=sk-ant-...
+# or
+GOOGLE_API_KEY=AI...
+
+# Optional: Override default models
+LLM_PROVIDER=anthropic
+SUMMARY_MODEL=claude-sonnet-4-5
+```
+
+### 2. Use the Python API
+
+```python
+from rnsr import RNSRClient
+
+# Simple one-line Q&A
+client = RNSRClient()
+answer = client.ask("contract.pdf", "What are the payment terms?")
+print(answer)
+
+# Advanced navigation with verification and self-reflection
+result = client.ask_advanced(
+    "complex_report.pdf",
+    "Compare liability clauses in sections 5 and 8",
+    enable_verification=True,
+    enable_self_reflection=True,
+    max_recursion_depth=3,
+)
+```
+
+### 3. Run the Demo UI
+
+```bash
+python demo.py
+# Open http://localhost:7860 in your browser
+```
+
+## New Features
+
+### Provenance System
+
+Every answer includes traceable citations:
+
+```python
+from rnsr.agent import ProvenanceTracker, format_citations_for_display
+
+tracker = ProvenanceTracker(kv_store=kv_store, skeleton=skeleton)
+record = tracker.create_provenance_record(
+    answer="The payment terms are net 30.",
+    question="What are the payment terms?",
+    variables=navigation_variables,
+)
+
+print(f"Confidence: {record.aggregate_confidence:.0%}")
+print(format_citations_for_display(record.citations))
+# Output:
+# **Sources:**
+# 1. [contract.pdf] Section: Payment Terms, Page 5: "Payment shall be due within 30 days..."
+```
+
+### LLM Response Caching
+
+Automatic caching reduces costs and latency:
+
+```python
+from rnsr.agent import wrap_llm_with_cache, get_global_cache
+
+# Wrap any LLM function with caching
+cached_llm = wrap_llm_with_cache(llm.complete, ttl_seconds=3600)
+
+# Use cached LLM - repeated prompts hit cache
+response = cached_llm("What is 2+2?")  # Calls LLM
+response = cached_llm("What is 2+2?")  # Returns cached (instant)
+
+# Check cache stats
+print(get_global_cache().get_stats())
+# {'entries': 150, 'hits': 89, 'hit_rate': 0.59}
+```
+
+### Self-Reflection Loop
+
+Answers are automatically critiqued and improved:
+
+```python
+from rnsr.agent import SelfReflectionEngine, reflect_on_answer
+
+# Quick one-liner
+result = reflect_on_answer(
+    answer="The contract expires in 2024.",
+    question="When does the contract expire?",
+    evidence="Contract dated 2023, 2-year term...",
+)
+
+print(f"Improved: {result.improved}")
+print(f"Final answer: {result.final_answer}")
+print(f"Iterations: {result.total_iterations}")
+```
+
+### Reasoning Chain Memory
+
+The system learns from successful queries:
+
+```python
+from rnsr.agent import get_reasoning_memory, find_similar_chains
+
+# Find similar past queries
+matches = find_similar_chains("What is the liability cap?")
+for match in matches:
+    print(f"Similar query: {match.chain.query}")
+    print(f"Similarity: {match.similarity:.0%}")
+    print(f"Past answer: {match.chain.answer}")
+```
+
+### Table Parsing
+
+Extract and query tables from documents:
+
+```python
+from rnsr.ingestion import TableParser, TableQueryEngine
+
+parser = TableParser()
+tables = parser.parse_from_text(document_text)
+
+# SQL-like queries
+engine = TableQueryEngine(tables[0])
+results = engine.select(
+    columns=["Name", "Amount"],
+    where={"Status": "Active"},
+    order_by="Amount",
+)
+
+# Aggregations
+total = engine.aggregate("Amount", "sum")
+```
+
+### Query Clarification
+
+Handle ambiguous queries gracefully:
+
+```python
+from rnsr.agent import QueryClarifier, needs_clarification
+
+# Check if query needs clarification
+is_ambiguous, analysis = needs_clarification(
+    "What does it say about the clause?"
+)
+
+if is_ambiguous:
+    print(f"Ambiguity: {analysis.ambiguity_type}")
+    print(f"Clarifying question: {analysis.suggested_clarification}")
+    # "What does 'it' refer to in your question?"
+```
+
+## Adaptive Learning
+
+RNSR learns from your document workload. All learned data persists in `~/.rnsr/`:
+
+```
+~/.rnsr/
+├── learned_entity_types.json       # New entity types discovered
+├── learned_relationship_types.json # New relationship types
+├── learned_normalization.json      # Title/suffix patterns
+├── learned_stop_words.json         # Domain-specific stop words
+├── learned_header_thresholds.json  # Document-type font thresholds
+├── learned_query_patterns.json     # Successful query patterns
+├── reasoning_chains.json           # Successful reasoning chains
+└── llm_cache.db                    # LLM response cache
+```
+
+The more you use RNSR, the better it gets at understanding your domain.
+
+## How It Works
+
+### Document Ingestion Pipeline
+
+```
+PDF → Font Analysis → Header Classification → Tree Building → Skeleton Index
+         ↓                    ↓                    ↓              ↓
+   Detect font sizes   Classify H1/H2/H3    Build hierarchy   Create summaries
+                                                  ↓
+                                        Multi-doc detection
+                                        (page number resets)
+```
+
+### Query Processing
+
+```
+Question → Clarify → Pre-Filter → Tree Navigation → Answer → Self-Reflect → Verify
+              ↓           ↓              ↓             ↓           ↓           ↓
+        Ask if ambig  Keyword scan  ToT reasoning  Synthesize  Critique   Fact-check
+                                         ↓                        ↓
+                                  Sub-LLM recursion        Improve answer
+                                  (complex queries)        (if issues)
+```
+
+### Entity Extraction (RLM Unified)
+
+```
+Document → LLM writes code → Execute on DOC_VAR → ToT validation → Cross-validate
+              ↓                     ↓                   ↓               ↓
+     Generates regex/Python   Grounded results   Probability scores  Entity↔Relationship
+                                    ↓
+                            All tied to exact text spans
+```
+
+## Architecture
+
+```
+rnsr/
+├── agent/                   # Query processing
+│   ├── rlm_navigator.py     # Main navigation agent
+│   ├── provenance.py        # Citation tracking (NEW)
+│   ├── llm_cache.py         # Response caching (NEW)
+│   ├── self_reflection.py   # Answer improvement (NEW)
+│   ├── reasoning_memory.py  # Chain memory (NEW)
+│   ├── query_clarifier.py   # Ambiguity handling (NEW)
+│   ├── graph.py             # LangGraph workflow
+│   └── variable_store.py    # Context management
+├── extraction/              # Entity/relationship extraction
+│   ├── rlm_unified_extractor.py  # Best extractor (NEW)
+│   ├── learned_types.py     # Adaptive type learning
+│   ├── entity_linker.py     # Cross-document linking
+│   └── models.py            # Entity/Relationship models
+├── indexing/                # Index construction
+│   ├── skeleton_index.py    # Summary generation
+│   ├── knowledge_graph.py   # Entity/relationship storage
+│   ├── kv_store.py          # SQLite/in-memory storage
+│   └── semantic_search.py   # Optional vector search
+├── ingestion/               # Document processing
+│   ├── pipeline.py          # Main ingestion orchestrator
+│   ├── font_histogram.py    # Font-based structure detection
+│   ├── header_classifier.py # H1/H2/H3 classification
+│   ├── table_parser.py      # Table extraction (NEW)
+│   ├── chart_parser.py      # Chart interpretation (NEW)
+│   └── tree_builder.py      # Hierarchical tree construction
+├── llm.py                   # Multi-provider LLM abstraction
+├── client.py                # High-level API
+└── models.py                # Data structures
+```
+
+## API Reference
+
+### High-Level API
+
+```python
+from rnsr import RNSRClient
+
+client = RNSRClient(
+    llm_provider="anthropic",  # or "openai", "gemini"
+    llm_model="claude-sonnet-4-5"
+)
+
+# Simple query
+answer = client.ask("document.pdf", "What is the main topic?")
+
+# Vision mode (for scanned docs)
+answer = client.ask_vision("scanned.pdf", "What does the chart show?")
+```
+
+### Low-Level API
+
+```python
+from rnsr import (
+    ingest_document,
+    build_skeleton_index,
+    run_rlm_navigator,
+    SQLiteKVStore
+)
+from rnsr.extraction import RLMUnifiedExtractor
+from rnsr.agent import ProvenanceTracker, SelfReflectionEngine
+
+# Step 1: Ingest document
+result = ingest_document("document.pdf")
+print(f"Extracted {result.tree.total_nodes} nodes")
+
+# Step 2: Build index
+kv_store = SQLiteKVStore("./data/index.db")
+skeleton = build_skeleton_index(result.tree, kv_store)
+
+# Step 3: Extract entities (grounded, no hallucination)
+extractor = RLMUnifiedExtractor()
+extraction = extractor.extract(
+    node_id="section_1",
+    doc_id="document",
+    header="Introduction",
+    content="..."
+)
+
+# Step 4: Query with provenance
+answer = run_rlm_navigator(
+    question="What are the key findings?",
+    skeleton=skeleton,
+    kv_store=kv_store
+)
+
+# Step 5: Get citations
+tracker = ProvenanceTracker(kv_store=kv_store)
+record = tracker.create_provenance_record(answer, question, variables)
+```
+
+## Configuration
+
+### Environment Variables
+
+| Variable | Description | Default |
+|----------|-------------|---------|
+| `LLM_PROVIDER` | Primary LLM provider | `auto` (detect from keys) |
+| `SUMMARY_MODEL` | Model for summarization | Provider default |
+| `AGENT_MODEL` | Model for navigation | Provider default |
+| `EMBEDDING_MODEL` | Embedding model | `text-embedding-3-small` |
+| `KV_STORE_PATH` | SQLite database path | `./data/kv_store.db` |
+| `LOG_LEVEL` | Logging verbosity | `INFO` |
+| `RNSR_LLM_CACHE_PATH` | Custom cache location | `~/.rnsr/llm_cache.db` |
+| `RNSR_REASONING_MEMORY_PATH` | Custom memory location | `~/.rnsr/reasoning_chains.json` |
+
+### Supported Models
+
+| Provider | Models |
+|----------|--------|
+| **OpenAI** | `gpt-5.2`, `gpt-5-mini`, `gpt-5-nano`, `gpt-4.1`, `gpt-4o-mini` |
+| **Anthropic** | `claude-opus-4-5`, `claude-sonnet-4-5`, `claude-haiku-4-5` |
+| **Gemini** | `gemini-3-pro-preview`, `gemini-3-flash-preview`, `gemini-2.5-pro`, `gemini-2.5-flash` |
+
+## Benchmarks
+
+RNSR is designed for complex document understanding tasks:
+
+- **Multi-document PDFs** - Automatically detects and separates bundled documents
+- **Hierarchical queries** - "Compare section 3.2 with section 5.1"
+- **Cross-reference questions** - "What does the appendix say about the claim in section 2?"
+- **Entity extraction** - Grounded extraction with ToT validation (no hallucination)
+- **Table queries** - "What is the total for Q4 2024?"
+
+## Sample Documents
+
+RNSR includes sample documents for testing and demonstration:
+
+### Synthetic Documents (`samples/`)
+
+| File | Type | Features Demonstrated |
+|------|------|----------------------|
+| `sample_contract.md` | Legal Contract | Entities (people, orgs), relationships, payment tables, legal terms |
+| `sample_financial_report.md` | Financial Report | Financial tables, metrics, executive names, quarterly data |
+| `sample_research_paper.md` | Academic Paper | Citations, hierarchical sections, technical content, tables |
+
+### Real Test Documents (`rnsr/test-documents/`)
+
+Legal documents from the Djokovic visa case (public court records) for testing with actual PDFs:
+- Affidavits and court applications
+- Legal submissions and orders
+- Interview transcripts
+
+### Using Sample Documents
+
+```python
+from pathlib import Path
+from rnsr.ingestion import TableParser
+from rnsr.extraction import CandidateExtractor
+
+# Parse a sample document
+sample = Path("samples/sample_contract.md").read_text()
+
+# Extract tables
+parser = TableParser()
+tables = parser.parse_from_text(sample)
+print(f"Found {len(tables)} tables")
+
+# Extract entities
+extractor = CandidateExtractor()
+candidates = extractor.extract_candidates(sample)
+print(f"Found {len(candidates)} entity candidates")
+```
+
+## Testing
+
+### Test Suite Overview
+
+RNSR has comprehensive test coverage with **281+ tests**:
+
+```bash
+# Run all tests
+pytest tests/ -v
+
+# Run specific feature tests
+pytest tests/test_provenance.py tests/test_llm_cache.py -v
+
+# Run end-to-end workflow tests
+pytest tests/test_e2e_workflow.py -v
+
+# Run with coverage
+pytest tests/ --cov=rnsr --cov-report=html
+```
+
+### Test Categories
+
+| Test File | Tests | Coverage |
+|-----------|-------|----------|
+| `test_e2e_workflow.py` | 18 | Full pipeline: ingestion → extraction → KG → query → provenance |
+| `test_provenance.py` | 17 | Citations, contradictions, provenance records |
+| `test_llm_cache.py` | 17 | Cache get/set, TTL, persistence |
+| `test_self_reflection.py` | 13 | Critique, refinement, iteration limits |
+| `test_reasoning_memory.py` | 15 | Chain storage, similarity matching |
+| `test_query_clarifier.py` | 19 | Ambiguity detection, clarification |
+| `test_table_parser.py` | 26 | Markdown/ASCII tables, SQL-like queries |
+| `test_chart_parser.py` | 16 | Chart detection, trend analysis |
+| `test_rlm_unified.py` | 13 | REPL execution, code cleaning |
+| `test_learned_types.py` | 13 | Adaptive learning registries |
+
+### End-to-End Workflow Tests
+
+The `test_e2e_workflow.py` demonstrates the complete pipeline:
+
+```python
+# Tests cover:
+# 1. Document Ingestion - Parse structure and tables
+# 2. Entity Extraction - Pattern-based grounded extraction  
+# 3. Knowledge Graph - Store entities and relationships
+# 4. Query Processing - Ambiguity detection, table queries
+# 5. Provenance - Citations and evidence tracking
+# 6. Self-Reflection - Answer improvement loop
+# 7. Reasoning Memory - Learn from successful queries
+# 8. LLM Cache - Response caching
+# 9. Adaptive Learning - Type discovery
+# 10. Full Workflow - Contract and financial analysis
+```
+
+## Development
+
+```bash
+# Install dev dependencies
+pip install -e ".[dev]"
+
+# Run linting
+ruff check .
+
+# Type checking
+mypy rnsr/
+```
+
+## Requirements
+
+- Python 3.10+
+- At least one LLM API key (OpenAI, Anthropic, or Gemini)
+
+## License
+
+MIT License - see [LICENSE](LICENSE) for details.
+
+## Contributing
+
+See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.
+
+## Research
+
+RNSR is inspired by:
+- [Hybrid Document Retrieval System Design](Research/Hybrid%20Document%20Retrieval%20System%20Design.pdf) - Core architecture and design principles
+- [PageIndex (VectifyAI)](https://github.com/VectifyAI/PageIndex) - Vectorless reasoning-based tree search
+- [Recursive Language Models](https://arxiv.org/html/2512.24601v1) - REPL environment with recursive sub-LLM calls
+- Tree of Thoughts - LLM-based decision making with probabilities
+- Self-Refine / Reflexion - Iterative self-correction patterns

rnsr-0.1.0.dist-info/RECORD

@@ -0,0 +1,72 @@
+rnsr/__init__.py,sha256=KOphXkwumziihpMKN0R1hipflpKoKD2BJ2vZHhVgBAw,3588
+rnsr/__main__.py,sha256=bupD9Fx6wJuDvmUMWQ4JWkOsk0nrhevmXuhqvolsaz4,6901
+rnsr/client.py,sha256=23JKz_5xUnp-QDq8FysqMMQxTtDoJ28Zo8QQU5tby6g,19041
+rnsr/document_store.py,sha256=NsYhkMAp8uB1DjkoZ_hKZiGQ9ut9ws2GCwA_3njtQh0,11168
+rnsr/exceptions.py,sha256=ukfZG6lv6L1G1o0_2nu2hMdcHDoaHEtx7rA-qo1coi0,1209
+rnsr/llm.py,sha256=Tyd5jF1NLKWOKoLPkMaCVlbtkTdQVU1Bq1lHIMCE2EY,26394
+rnsr/models.py,sha256=XTxyb5o8LFpyummhLfsPhuLzHnHcFj6_3TXcaeVuvZA,4288
+rnsr/py.typed,sha256=ixa8YukDZ3kLo0WsFJRGohLMyHzbMur1ALmmASML2cs,64
+rnsr/agent/__init__.py,sha256=UjSESsMqu84EgEWV2zMzRM0CmVAtZmGafFhVf8gA5sI,5622
+rnsr/agent/cross_doc_navigator.py,sha256=QD8m0EFy3P7Klp68hq57VK7QviIV5TFan_kZvu0ljnM,24118
+rnsr/agent/graph.py,sha256=khxfdwHU9s2Fi4tWkShA_6_pJ4UxklgFtgP8DHUonPI,52516
+rnsr/agent/llm_cache.py,sha256=mz4_IblfaTwhQQz58snn5Imc9HkfzW3DpMdmrRghsKY,17467
+rnsr/agent/navigator_api.py,sha256=cvYpqPjDUeSsP-UflPN_E6a6X44qpkAwM-N5jXDg-k0,15747
+rnsr/agent/provenance.py,sha256=pFYAVec-kbluhm2ibVAeQZuADQkexdHDWnLP7-Ycauk,25084
+rnsr/agent/query_clarifier.py,sha256=t3BvmtxQp37U5Xx4qsxxCE2X-Bzegxm7BgQsujn5YWg,21071
+rnsr/agent/reasoning_memory.py,sha256=d_JWt1rbWgXAPfcNh3_gi_75_JYFhCxIiurvqBij7pY,23984
+rnsr/agent/repl_env.py,sha256=2lurSa0C9VGGwm0SJxMzZjA1S-sjKmGlPhcDvuWrM8s,25346
+rnsr/agent/rlm_navigator.py,sha256=TSa3x85MywwERuaLfNfNN9GESQlnAX7b2ycpLwFuXPM,72070
+rnsr/agent/self_reflection.py,sha256=XnQ6D9dVCNlr1zWGSxwWRgAkDWOYvSbtjeYbUcCq4jQ,19231
+rnsr/agent/variable_store.py,sha256=NfKx2JzzRtP7WjPG0Zzr_3IFjac9BZl6jBb4KNNJ0-k,9187
+rnsr/benchmarks/__init__.py,sha256=A2WNIIzsAdRFg7WYtcG3f6kemQFiE3m4_zpq4ldS7cI,3085
+rnsr/benchmarks/comprehensive_benchmark.py,sha256=HqZNQ9DZiZ0LkgeEVCv3uhiATbSAYm73PookujXqe_c,24085
+rnsr/benchmarks/evaluation_suite.py,sha256=gXmkXyAXQMtxLXGGveBdZYfuY10-9ma6u0yboIhNpU4,44904
+rnsr/benchmarks/finance_bench.py,sha256=bSKsBwMtuabcWyKQEYJFYhnnJ2jRw0khzE9NBsTiJRY,5330
+rnsr/benchmarks/pdf_merger.py,sha256=UNQHefghWTh7eSkWE2qD_1Rhi97gHcfWs0jXY8F9MDM,6323
+rnsr/benchmarks/performance.py,sha256=bZnR-4xZtII8legY1c9b9ingtML9aSiHkHvQbc4oRbE,9728
+rnsr/benchmarks/quality.py,sha256=Mog6nThF-5N9A0kFkFnXlfl1eOcLoKMUbHI_EK-adko,10054
+rnsr/benchmarks/runner.py,sha256=hzbJZBAk1MgGDAPBC0bE3Gd_KKCobhSpdWiZOhh0cfg,9740
+rnsr/benchmarks/standard_benchmarks.py,sha256=1xSLZH6jPYNwhZpdVTHp4vckoAm-P9UmUru9gpb47P0,34793
+rnsr/extraction/__init__.py,sha256=4EBU5M2cIf3gqR26n7EuALO96VcZ8tfJUJhxGbmUr9A,4528
+rnsr/extraction/candidate_extractor.py,sha256=9pmuxG3PhoRvDrRvyr0gF9ZntUMkS7iTDapc28Bhy7I,13457
+rnsr/extraction/entity_extractor.py,sha256=DRH-tdr0jc7ujPGa4VuvG5xVDfkSavf4OhU7LVQhmnc,19496
+rnsr/extraction/entity_linker.py,sha256=peqzkqvKkrWo8ZJ3MQF1qBVK9BoXY9tNmtPRIIZE9YA,28279
+rnsr/extraction/grounded_extractor.py,sha256=aBiB7ckkFWnQiUXk9IrRYzgDN2PXujzdR6dqAGsYrJo,24512
+rnsr/extraction/learned_types.py,sha256=XdAmwQkaQ2hMlakYYvM1sw_9TK1RLAbKoePhbKuDJBc,19117
+rnsr/extraction/models.py,sha256=GwoGkczyp6hwnYrzUGR9mevATlEkxPRzOMEeLrwYo98,8921
+rnsr/extraction/relationship_extractor.py,sha256=9a5U3iZECC-b5mwazFP2N3_fiwEz99heONq0PeIA-Ek,20080
+rnsr/extraction/relationship_patterns.py,sha256=ByFAOHXqINqVsD-LlisYPS92lZNUKG60rWkAP6WFPyQ,18910
+rnsr/extraction/relationship_validator.py,sha256=56aPDpKIx_EXRbPvdjBUc7fqBZU4mV8RehPNuh19pbg,13483
+rnsr/extraction/rlm_extractor.py,sha256=UsSyMPPqcyPJ02jzkttjn4CzEK7u8TDYF0ciAFpnRLo,20366
+rnsr/extraction/rlm_unified_extractor.py,sha256=JPHyOW04AamYs6RT4h1IEbFg3OBHSW8h7bAa2gjeXp8,36762
+rnsr/extraction/tot_validator.py,sha256=mmM6TtEoyTV-8pCUucBy8TcJ3hPOpABa9yeFgLRDGfI,21586
+rnsr/extraction/unified_extractor.py,sha256=AuRhPJ0Zna1_7vTpZ2fnl0-s_7xjQfGHQ6oNKGJzDH8,12540
+rnsr/indexing/__init__.py,sha256=on-bohmE6qoLIxMQNBHQCXSRKr98aAC0RzVOkrVlIVA,1417
+rnsr/indexing/knowledge_graph.py,sha256=xuceER-kevyZJHGNlCcIHxnoFifZWxmTvjnqPgxFQa8,38592
+rnsr/indexing/kv_store.py,sha256=iXNwGGIKev-AhMDx-dinYqqhUE3UT20TGM2uTOSHQRM,9438
+rnsr/indexing/persistence.py,sha256=GhE8VWCsB0SKAiN_SQiw5fLsYAXm849SqRdhUb9BMDE,9427
+rnsr/indexing/semantic_retriever.py,sha256=ypNl0_gi6ySPBDhgYwML0_OkS5bgbOg5H1yV0xy8tQU,7365
+rnsr/indexing/semantic_search.py,sha256=Qv_25cr3D-uX7Duk2clt_SJR6wATU4UdQogzeTeVYfw,10395
+rnsr/indexing/skeleton_index.py,sha256=jNC6U0JdT-7zoQ-Z85mpB0xk_mgmTlEUfDUVAb6cFvk,11717
+rnsr/ingestion/__init__.py,sha256=eWEWlWp7qmD_RkAjjNwUnWUHfSOnyNnPQdprochhgZM,4583
+rnsr/ingestion/chart_parser.py,sha256=O5y7FfgeCoHL8W3DY0AQSPPlOAMV4JhvkbpxJnPlmlw,18672
+rnsr/ingestion/document_boundary.py,sha256=aGsFxa7haXSCFuefxbTWiWVtXmHKmsTjsZI1y0GhAgc,25447
+rnsr/ingestion/font_histogram.py,sha256=B8PqlJjjEASyYkd97JPPFPtNN6zO7uVBi4p6XqVFaTc,11370
+rnsr/ingestion/header_classifier.py,sha256=9AT7BHNB4yXHyPpxROb0oydmcHUYJyMtshB6UTqnT_o,19535
+rnsr/ingestion/hierarchical_cluster.py,sha256=-A77Kfvzwl6y6XfKjt-kF_Xk7tOy5aPc-K0QBIP2bOQ,17343
+rnsr/ingestion/layout_detector.py,sha256=kHarT41pbETINewhI_MU-5bmB5OmU2LiJFeHHOzbDZo,10693
+rnsr/ingestion/layout_model.py,sha256=41XtQsjxIny0Z6WfeuclXNACh6ayVYjlXILmQzjHljw,11629
+rnsr/ingestion/ocr_fallback.py,sha256=WaUiY6YnoNYtD-smIogSltRKYwQ4MNDNId2mOT5fsD8,4615
+rnsr/ingestion/pipeline.py,sha256=ciGeq51mzWPNViNUxdEccJ9gsKVp5nCa68xF9451zh8,30449
+rnsr/ingestion/semantic_fallback.py,sha256=LhnY4qfqRDhNgutBVJqW_6I_uKyX0IvxM6tT6YhYwSo,13207
+rnsr/ingestion/table_parser.py,sha256=-QzXsFLyZm4nsic6zLKBQWFaB8xlSYXg9QtpneojJ0o,24699
+rnsr/ingestion/text_builder.py,sha256=NGc80EfehQCNQdf0do4U1lnhKpiObu6VsV0u9EepvSI,14700
+rnsr/ingestion/tree_builder.py,sha256=2XrpJDhk1rPQNTR4FQKELKPOXFa5TDDmDGCXYebYmXY,12561
+rnsr/ingestion/vision_retrieval.py,sha256=an5IOIQlWavZ1rgNHz8pdP_QVTLaBsBfWnxqj7gSoxI,31000
+rnsr/ingestion/xy_cut.py,sha256=2KvC9RFtWZ0c2JhPkKtT5saOOuLaWAGgCMh40Fx1_7Q,18025
+rnsr-0.1.0.dist-info/licenses/LICENSE,sha256=VplyxAvQdVnyS4R-X2Iakr_ffeFSgRAiIfEU7kZgxf8,1074
+rnsr-0.1.0.dist-info/METADATA,sha256=P2lAd_Y1WkACXjbOIib1zTlHF5n-0eA4BqOrut35OeQ,19867
+rnsr-0.1.0.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
+rnsr-0.1.0.dist-info/entry_points.txt,sha256=WFGf4rTMJbDEOt_PvjaHHtHRmJHNk0qfaJpghduMaM8,44
+rnsr-0.1.0.dist-info/top_level.txt,sha256=hylIhN9Hbr5V92x_rNExhkodqj8VvlbhZ8Evog443dY,5
+rnsr-0.1.0.dist-info/RECORD,,

rnsr-0.1.0.dist-info/WHEEL

@@ -0,0 +1,5 @@
+Wheel-Version: 1.0
+Generator: setuptools (80.10.2)
+Root-Is-Purelib: true
+Tag: py3-none-any
+

rnsr-0.1.0.dist-info/entry_points.txt

@@ -0,0 +1,2 @@
+[console_scripts]
+rnsr = rnsr.__main__:main

rnsr-0.1.0.dist-info/licenses/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 RNSR Contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.