When you have existing conversation history to import, batch processing lets you efficiently ingest large volumes of messages into EverOS. This guide covers data formats, the v1 SDK for batch imports, and monitoring.Documentation Index
Fetch the complete documentation index at: https://docs.evermind.ai/llms.txt
Use this file to discover all available pages before exploring further.
Prerequisites
Install the EverOS SDK:Data Format
EverOS accepts two message formats depending on whether you are importing group conversations (multi-participant) or personal conversations (single user with an assistant).Group Conversation Format
Each message must includesender_id and sender_name to identify the participant:
Personal Conversation Format
For one-on-one conversations between a user and an assistant:Field Reference
Group messages:| Field | Required | Description |
|---|---|---|
role | Yes | "user" for all group participants |
sender_id | Yes | Unique identifier for the sender |
sender_name | Yes | Display name of the sender |
timestamp | Yes | Unix timestamp in milliseconds |
content | Yes | Message text content |
message_id | No | Unique message identifier |
| Field | Required | Description |
|---|---|---|
role | Yes | "user" or "assistant" |
timestamp | Yes | Unix timestamp in milliseconds |
content | Yes | Message text content |
Converting Your Data
Convert your existing chat data to the EverOS format before importing:Batch Import Script
Here is a complete batch import script using the EverOS SDK with async mode for efficient processing:Monitoring Progress
Add progress tracking for large imports:Handling Large Conversations
For very large conversations, process in smaller chunks to manage memory and avoid timeouts:Error Recovery
Handle partial failures with checkpointing:Best Practices
Ordering
Ordering
Always sort messages chronologically before import. EverOS uses message timestamps for boundary detection.
Chunk Size
Chunk Size
The SDK sends messages in batches. Adjust
chunk_size based on your use case:Concurrency
Concurrency
Balance concurrency to avoid overwhelming the server. Higher concurrency speeds up imports but may trigger rate limits.
Memory Management
Memory Management
For large imports, process files one at a time rather than loading all into memory.
Flush After Import
Flush After Import
Always call flush after importing a conversation to trigger memory extraction immediately. Without flushing, EverOS waits for the boundary detection timeout before processing.
Next Steps
Python Integration
Production-ready client patterns
Team Collaboration
Use imported group chat memories