Deep Dive

6 Export Formats

From readable Markdown to machine-readable RAG chunks — every format explained.

01
Markdown & JSON

Markdown for humans: grouped by type, with time context and confidence. JSON for backup: complete export with nodes, relations and history. Importable via merge.

You: "Context Memory: Export as Markdown."
→ project.knowledge.md

You: "Context Memory: Make a JSON backup."
→ project.knowledge.json
02
JSON-LD & Mermaid

JSON-LD: Knowledge graph with schema.org context. For Neo4j, Gephi. Mermaid: Visual diagram, color-coded by type. For Obsidian, GitHub, mermaid.live.

JSON-LD: @context: schema.org
  Nodes: id, label, cluster, confidence, temporal

Mermaid:
  dec_001[PostgreSQL]:::dec
  arch_001 -->|depends on| dec_001
03
RAG Chunks & Quality Score

Embedding chunks: Clean text, no Markdown or emojis. Time context prepended as natural language. JSONL format. Quality Score: 8 automatic checks, 0-100.

Chunk (JSONL):
{"text": "As of 2025-06: PostgreSQL. ACID transactions...",
 "confidence": "high", "token_estimate": 45}

Quality: 92/100 — 7/8 checks passed