Skip to content

Assembler

The Knowledge Builder for Sayou Fabric.

sayou-assembler acts as the structural engineer of the pipeline. It accepts flat lists of SayouNode objects produced by the Wrapper and constructs meaningful relationships (Edges), embeddings, or specific query payloads.

While the Wrapper focuses on what the data is (Schema), the Assembler focuses on how the data connects (Topology).

1. Architecture & Role

The Assembler takes standardized Nodes and applies a Building Strategy to generate a connected graph structure or vector payloads ready for loading.

graph LR
    Nodes[Flat SayouNodes] --> Pipeline[Assembler Pipeline]

    subgraph Builders
        Graph[General Graph]
        Code[Python AST]
        Time[Timeline]
        Vector[Embeddings]
        Cypher[Cypher Query]
    end

    Pipeline -->|Config Routing| Builders
    Builders --> Payload[Connected Payload]

1.1. Core Features

  • Topology Construction: Automatically links nodes (e.g., Parent-Child, Function-Class) based on metadata.
  • Domain Specificity: Dedicated builders for Python code analysis and Video timelines.
  • Bi-directional Linking: Automatically generates reverse edges (e.g., contains <-> belongsTo) to ensure graph traversability.

2. Available Strategies

sayou-assembler provides specialized builders for different data domains.

Strategy Key Builder Class Description
graph GraphBuilder [Default] Generic builder. Links nodes based on parent_id and generic relationships.
code CodeGraphBuilder [Code] Parses Python AST metadata to link Imports, Classes, and Methods.
timeline TimelineBuilder [Media] Sequences video segments or logs chronologically (NEXT, PREV).
vector VectorBuilder [Search] Converts node content into vector embeddings.
cypher CypherBuilder [DB] Generates Neo4j MERGE statements for idempotent inserts.

3. Installation

Bash
pip install sayou-assembler

4. Usage

The AssemblerPipeline is the entry point. It transforms SayouOutput (from Wrapper) into a dictionary containing nodes and edges.

Case A: General Graph Construction

The default strategy for general documents (PDF, Markdown).

Python
from sayou.assembler import AssemblerPipeline

wrapper_output = {
    "nodes": [
        {"node_id": "doc:1", "node_class": "sayou:Document", "attributes": {...}},
        {"node_id": "doc:1:header", "attributes": {"meta:parent_id": "doc:1"}, ...}
    ]
}

graph_data = AssemblerPipeline.process(data=wrapper_output)

print(f"Nodes: {len(graph_data['nodes'])}, Edges: {len(graph_data['edges'])}")

Case B: Python Code Graph

Links functions to classes and resolves import dependencies.

Python
from sayou.assembler import AssemblerPipeline

wrapper_output = {
    "nodes": [
        {"node_id": "doc:1", "node_class": "sayou:Document", "attributes": {...}},
        {"node_id": "doc:1:header", "attributes": {"meta:parent_id": "doc:1"}, ...}
    ]
}

code_graph = AssemblerPipeline.process(data=wrapper_output)

print(f"Dependency Edges: {len(code_graph['edges'])}")

Case C: Multimedia Timeline

Sequences video segments by time.

Python
from sayou.assembler import AssemblerPipeline

wrapper_output = {
    "nodes": [
        {"node_id": "doc:1", "node_class": "sayou:Document", "attributes": {...}},
        {"node_id": "doc:1:header", "attributes": {"meta:parent_id": "doc:1"}, ...}
    ]
}

timeline_graph = AssemblerPipeline.process(data=wrapper_output)
print(f"Timeline Sequence Created.")

5. Configuration Keys

You can customize the building process via the config dictionary.

  • graph: create_reverse_edges (bool), default_edge_label.
  • python: resolve_imports (bool), include_external_libs (bool).
  • timeline: interval_threshold (float).
  • vector: embedding_model (e.g., openai, huggingface), batch_size.
  • cypher: merge_mode (bool), batch_size.

6. License

Apache 2.0 License © 2026 Sayouzone