imports

The imports collection will represent raw, unprocessed user input/data/memories that will later be ingested into the ContextCore memory system

Categories of Data:

• Core Data: This like the raw_content of the user input or data and the source_provider of the data

{
  "raw_content": { ... },   // original user input (JSON object or array)
  "source_provider": "openai",  // e.g., openai, anthropic, obsidian_plugin
}

• Import Classification: import_type, import_trigger, import_session

{
  "import_type": "provider",      // "provider", "integration", "manual_upload", "file_upload"
  "import_trigger": "manual",     // "manual", "scheduled", "auto_response"
  "import_session": "sess_abc123" // all imports part of one UX event share this
}

• User + Traceability

{
  "user_id": ObjectId,
  "source_prompt_id": ObjectId,    // (if applicable — which prompt was used to request export)
  "num_memories": 15
}

• Status + Workflow: Needed for ingestion and data pipelines

{
  "status": "pending",             // "pending", "processed", "errored", "duplicate"
  "processing_notes": null,        // or a string with error/debug message
  "is_duplicate": false            // based on hash match
}

• Timestamps

Schema

{
  "_id": ObjectId,
  "user_id": ObjectId,
  "raw_content": [...],                  // the unprocessed JSON array of memories
  "source_provider": "openai",           // origin of export
  "import_type": "provider",             // provider | integration | manual_upload
  "import_trigger": "manual",            // manual | auto_response | scheduled
  "import_session": "sess_abc123",       // all imports from same UX flow share this
  "source_prompt_id": ObjectId,          // optional
  "num_memories": 15,                    // count of raw memory objects
 
  "status": "pending",                   // pending | processed | errored
  "processing_notes": null,             // useful for debugging errors
  "is_duplicate": false,                 // hash match of content (optional)
 
  "created_at": ISODate(),
  "processed_at": null
}

Ideas for later:

Field	Purpose
client_metadata	JSON blob (user-agent, app version, etc.) for telemetry/debugging
import_hash	Hash of full payload — helps with deduplication at batch level
memory_fingerprints	Optional array of memory hashes extracted from payload (if pre-processed)