Wrapper

The Ontology Mapper for Sayou Fabric.

sayou-wrapper takes the fragmented chunks produced by sayou-chunking and wraps them into a standardized graph structure (SayouNode). This is the final preparation step before data is assembled into a Knowledge Graph or loaded into a Vector DB.

It applies the Sayou Ontology Schema (Namespace -> Class -> Predicate) to raw data, turning simple text into semantically rich entities.

---

1. Architecture & Role

The Wrapper sits between Chunking and Assembly. It transforms raw dictionaries (Chunks) into standardized Objects (Nodes) with globally unique URIs.

graph LR
    Chunks[Raw Chunks] --> Pipeline[Wrapper Pipeline]

    subgraph Logic
        Schema[Schema Mapping]
        URI[URI Generation]
        Attr[Attribute Norm]
    end

    Pipeline --> Logic
    Logic --> Nodes[SayouNodes]

1.1. Core Features

• Ontology Enforcement: Assigns strict classes (e.g., sayou:Topic, sayou:Function) based on chunk metadata.
• URI Normalization: Generates idempotent IDs (e.g., sayou:doc:hash_123) to prevent duplication.
• Metadata Preservation: Carries over source coordinates (page numbers, line numbers) into node attributes.

---

2. Available Strategies

sayou-wrapper applies different mapping rules based on the source domain.

Strategy Key	Target Domain	Description
document_chunk	PDF, Markdown	Maps headers to sayou:Topic and text to sayou:TextFragment. Preserves hierarchy metadata.
code_chunk	Python, Java	Maps AST objects to sayou:Class, sayou:Function, or sayou:Module.
video_chunk	YouTube, MP4	Maps timestamps to sayou:VideoSegment. Preserves start/end times.
general	Plain Text	[Default] Treats everything as generic sayou:Unstructured nodes.

---

3. Installation

Bash

pip install sayou-wrapper

---

4. Usage

The WrapperPipeline converts a list of dictionaries (Chunks) into a WrapperOutput object containing nodes.

Case A: Document Processing (Default)

Converts hierarchical chunks into Topic and Fragment nodes.

Python

from sayou.wrapper import WrapperPipeline

chunks = [
    {
        "content": "# Introduction",
        "metadata": {
            "chunk_id": "h_1", 
            "semantic_type": "heading", 
            "depth": 1
        }
    },
    {
        "content": "Sayou is great.",
        "metadata": {
            "chunk_id": "p_1", 
            "parent_id": "h_1",
            "semantic_type": "text"
        }
    }
]

result = WrapperPipeline.process(
    data=chunks,
    strategy="document_chunk"
)

for node in result.nodes:
    print(f"[{node.node_class}] {node.node_id}")

Case B: Code Processing

Converts AST-based chunks into structural code nodes.

Python

from sayou.wrapper import WrapperPipeline

code_chunks = [
    {
        "content": "def my_func(): pass",
        "metadata": {"type": "function", "name": "my_func"}
    }
]

result = WrapperPipeline.process(
    data=code_chunks,
    strategy="code_chunk"
)

print(f"Node Class: {result.nodes[0].node_class}")

---

5. Configuration Keys

The config dictionary controls how IDs are generated and which schema strictness to apply.

• namespace: The prefix for URIs (default: sayou:).
• id_strategy: hash (deterministic) or uuid (random).
• strict_mode: If True, raises error for unknown semantic types.
• attribute_filter: List of metadata keys to exclude from the final Node attributes.

---

6. License

Apache 2.0 License © 2026 Sayouzone