Skip to content

run_memorize_usage.md

Latest commit

 

History

History
300 lines (226 loc) · 9.21 KB

run_memorize_usage.md

File metadata and controls

300 lines (226 loc) · 9.21 KB

run_memorize.py Usage Documentation

Overview

run_memorize.py is a group chat memory storage script that reads JSON files conforming to the GroupChatFormat format and stores them item by item into the memory system via HTTP API.

Features

  • ✅ Read and validate JSON files in GroupChatFormat format
  • ✅ Support both assistant and companion scenarios
  • ✅ Automatically save conversation metadata (conversation-meta)
  • ✅ Call memorize interface item by item to process messages
  • ✅ Provide format validation mode
  • ✅ Detailed logging output

Usage

1. Basic Usage

Store memories via HTTP API (must specify scene):

python src/bootstrap.py src/run_memorize.py \
  --input data/group_chat.json \
  --api-url http://localhost:1995/api/v1/memories \
  --scene assistant

2. Using companion Scenario

python src/bootstrap.py src/run_memorize.py \
  --input data/group_chat.json \
  --api-url http://localhost:1995/api/v1/memories \
  --scene companion

3. Format Validation Only

Validate whether the input file format is correct without performing storage (no API address needed):

python src/bootstrap.py src/run_memorize.py \
  --input data/group_chat.json \
  --scene assistant \
  --validate-only

Command-Line Arguments

Argument Required Description
--input Yes Input group chat JSON file path (GroupChatFormat format)
--scene Yes Memory extraction scenario, only supports assistant or companion
--api-url No* memorize API address (required for non-validation mode)
--validate-only No Only validate input file format, do not perform storage

*Note: When using --validate-only, no need to provide --api-url, otherwise it's required.

Input File Format

The input file must conform to the GroupChatFormat specification, see data_format/group_chat/group_chat_format.py.

Format Example

{
  "version": "1.0.0",
  "conversation_meta": {
    "name": "Smart Sales Assistant Project Team",
    "description": "Development discussion group for Smart Sales Assistant project",
    "group_id": "group_sales_ai_2025",
    "created_at": "2025-02-01T01:00:00Z",
    "default_timezone": "UTC",
    "user_details": {
      "user_101": {
        "full_name": "Alex",
        "role": "Tech Lead"
      },
      "user_102": {
        "full_name": "Betty",
        "role": "Product Manager"
      }
    },
    "tags": ["AI", "Sales", "Project Development"]
  },
  "conversation_list": [
    {
      "message_id": "msg_001",
      "create_time": "2025-02-01T02:00:00Z",
      "sender": "user_101",
      "sender_name": "Alex",
      "type": "text",
      "content": "Good morning everyone, let's discuss project progress today",
      "refer_list": []
    }
  ]
}

Processing Flow

The script executes the following steps:

  1. Format Validation

    • Read input JSON file
    • Validate whether it conforms to GroupChatFormat specification
    • Output data statistics

  2. Save Conversation Metadata

    • Call conversation-meta interface
    • Save metadata such as scene, group information, user details
    • API address: {base_url}/api/v1/conversation-meta

  3. Process Messages Item by Item

    • Call memorize interface sequentially for each message
    • Each message includes: message_id, create_time, sender, content, etc.
    • Automatically add group_id, group_name, scene information
    • API address: {api_url} (specified by --api-url argument)

  4. Output Results

    • Display number of successfully processed messages
    • Display total number of saved memories

Output Example

Successful Output

🚀 Group Chat Memory Storage Script
======================================================================
📄 Input File: /path/to/group_chat.json
🔍 Validation Mode: No
🌐 API Address: http://localhost:1995/api/v1/memories
======================================================================
======================================================================
Validating Input File Format
======================================================================
Reading file: /path/to/group_chat.json
Validating GroupChatFormat format...
✓ Format validation passed!

=== Data Statistics ===
Format Version: 1.0.0
Group Chat Name: Smart Sales Assistant Project Team
Group Chat ID: group_sales_ai_2025
Number of Users: 5
Number of Messages: 8
Time Range: 2025-02-01T02:00:00Z ~ 2025-02-01T02:05:00Z

======================================================================
Reading Group Chat Data
======================================================================
Reading file: /path/to/group_chat.json
Using simple direct single message format, processing item by item

======================================================================
Starting to Call memorize API Item by Item
======================================================================
Group Name: Smart Sales Assistant Project Team
Group ID: group_sales_ai_2025
Number of Messages: 8
API Address: http://localhost:1995/api/v1/memories

--- Saving Conversation Metadata (conversation-meta) ---
Saving conversation metadata to: http://localhost:1995/api/v1/conversation-meta
Scene: assistant, Group ID: group_sales_ai_2025
  ✓ Conversation metadata saved successfully
  Scene: assistant

--- Processing Message 1/8 ---
  ✓ Successfully saved 1 memory

--- Processing Message 2/8 ---
  ⏳ Waiting for episode boundary

--- Processing Message 3/8 ---
  ✓ Successfully saved 2 memories

--- Processing Message 4/8 ---
  ⏳ Waiting for episode boundary

--- Processing Message 5/8 ---
  ⏳ Waiting for episode boundary

--- Processing Message 6/8 ---
  ✓ Successfully saved 1 memory

--- Processing Message 7/8 ---
  ⏳ Waiting for episode boundary

--- Processing Message 8/8 ---
  ✓ Successfully saved 2 memories

======================================================================
Processing Complete
======================================================================
✓ Successfully Processed: 8/8 messages
✓ Total Saved: 6 memories

======================================================================
✓ Processing Complete!
======================================================================

Error Handling

File Does Not Exist

Error: Input file does not exist: /path/to/file.json

Format Validation Failed

✗ Format validation failed!
Please ensure input file conforms to GroupChatFormat specification

JSON Parsing Error

✗ JSON parsing failed: Expecting value: line 1 column 1 (char 0)

Development Notes

Core Dependencies

  • infra_layer.adapters.input.api.mapper.group_chat_converter: Format validation
  • httpx: HTTP client (async requests)
  • core.observation.logger: Logging utilities

API Endpoints

The script calls two API endpoints:

  1. conversation-meta: Save conversation metadata

    • Path: {base_url}/api/v1/conversation-meta
    • Method: POST
    • Data: Contains metadata such as scene, group_id, user_details

  2. memorize: Store single message memory

    • Path: {api_url} (specified by --api-url argument)
    • Method: POST
    • Data: Contains message_id, sender, content, scene, etc.

Extension Suggestions

  1. Batch Processing: Support processing multiple files in a directory
  2. Progress Display: Add progress bar to show processing status
  3. Error Retry: Add failure retry mechanism
  4. Concurrent Processing: Support batch concurrent API calls (note: maintain message order)
  5. Result Export: Export storage results as JSON file

Common Questions

Q1: Why is it recommended to start with bootstrap.py?

A: bootstrap.py automatically handles:

  • Python path setup
  • Environment variable loading
  • Dependency injection container initialization
  • Mock mode support

This ensures the script runs in a complete application context.

Q2: What's the difference between assistant and companion scenarios?

A:

  • assistant: Assistant scenario, suitable for AI assistant and user conversations
  • companion: Companion scenario, suitable for AI companion interactive conversations

Different scenarios affect memory extraction strategies and storage methods. Choose based on actual application scenario.

Q3: Why does message processing show "Waiting for episode boundary"?

A: The memory system uses "Episode Boundary" to determine when to form complete memory fragments.

  • Not every message immediately generates a memory
  • The system waits for a complete conversation episode to end before extracting memories
  • This is normal processing behavior, not a failure

Q4: Can I not provide an API address?

A: No. The current version only supports calling via HTTP API, you must provide the --api-url argument (unless using --validate-only for format validation only).

Q5: What to do if API call fails?

A: Check the following:

  1. Ensure memory service is running
  2. Confirm API address is correct (including port number)
  3. View server logs to understand detailed error information
  4. Confirm input data format is correct

References