Extract Content

Extract entities (people, systems, concepts) and concepts (decisions, directives, proposals) from documents using LLM-based extraction.

Requirement Level

MAY use for document ingestion. For storing learnings, SHOULD use ingester_ingest directly instead (simpler path).

When to Use

MAY use when:

• •Ingesting new documents into HAIOS memory
• •Analyzing document structure and key concepts
• •Identifying entities mentioned in specifications
• •Populating the knowledge graph

Instructions

Via MCP Tool (Recommended)

Use the extract_content MCP tool:

extract_content(
    file_path="path/to/document.md",
    content="<content of the file>",
    extraction_mode="full"  # or "entities_only", "concepts_only"
)

Via Python (Direct)

Note: haios_etl is deprecated. For most use cases, the MCP tool extract_content above is recommended.

Extraction Types

Entities

• •Agent: AI agents, systems, tools
• •User: Human actors, roles
• •ADR: Architecture Decision Records
• •System: Technical systems, services

Concepts

• •Decision: Architectural or design decisions
• •Directive: Instructions, guidelines
• •Proposal: Suggested approaches
• •Critique: Objections, concerns

Example Workflow

1. •Read document content
2. •Call extract_content with the content
3. •Review extracted entities and concepts
4. •Use ingester_ingest to persist relevant items

# Step 1: Extract
result = extract_content(file_path="docs/ADR/ADR-001.md", content=doc_content)

# Step 2: Store important concepts
ingester_ingest(
    content="Use SQLite for the memory database",
    source_path="docs/ADR/ADR-001.md",
    content_type_hint="techne"
)

Related Tools

• •ingester_ingest: PRIMARY - Store extracted content with auto-classification
• •memory_search_with_experience: Search stored memories
• •marketplace_list_agents: Find agents that can help with extraction
• •memory_store: MUST NOT use - deprecated, use ingester_ingest