Ingesting Data

Engram supports multiple data sources for ingestion.

Supported Sources

Source	Command	Description
Claude	engram ingest claude	Claude conversation exports
ChatGPT	engram ingest chatgpt	ChatGPT data exports
Craft	engram ingest craft	Craft MCP API

Claude Conversations

Export your Claude conversations and ingest them:

engram ingest claude ./data/claude/conversations.json

ChatGPT Conversations

Export your ChatGPT data from OpenAI and ingest:

engram ingest chatgpt ./data/chatgpt/conversations.json

Craft Documents

Ingest from a running Craft MCP server:

engram ingest craft http://localhost:8080

What Happens During Ingestion

1. Parsing: Documents are parsed from the source format
2. Chunking: Long documents are split into semantic chunks (500 tokens, 50 token overlap)
3. Embedding: Each chunk is embedded using Azure OpenAI
4. Storage: Chunks are stored in PostgreSQL with pgvector

Checking Ingestion Status

engram stats

See Also

• Search Guide - Search your ingested data
• CLI Reference: ingest - Full command reference