[BUG] Retrieve(LLM-based) with qwen-max model not worked

[sugopher]

Description

memU version: v1.1.2

llm profile config in test_postgres.py:

llm_profiles={
            "default": {
                "base_url": "https://dashscope.aliyuncs.com/compatible-mode/v1",
                "api_key": "xxxxxx",
                "chat_model": "qwen-max",
                "embed_model": "text-embedding-v3",
                "client_backend": "sdk"
            }
        },

run : python test_postgres.py

[POSTGRES] Starting test...
[POSTGRES] DSN: postgresql+psycopg://postgres:postgres@localhost:5432/memu
============================================================

[POSTGRES] Memorizing...
  - personal_info: ...
  - preferences: # preferences
## Supervision and Habits
- The user prefers to have supervision t...
  - relationships: # relationships
## Ideal Partner
- The user discussed their ideal type for a par...
  - activities: # activities
## Exercise
- The user has not exercised much recently due to rainy...
  - goals: # goals
## Health and Wellness
- The user is trying to maintain healthy sleep ha...
  - experiences: # experiences
## Daily Routine
- The user went to bed at 1:30 AM last night due ...
  - knowledge: ...
  - opinions: # opinions
## Core Traits
- The user admires people who can handle real-life rel...
  - habits: # habits
## Daily Routines
- The user usually goes to bed around 12:30 AM
- The ...
  - work_life: ...

[POSTGRES] RETRIEVED - RAG
  Categories:
    - preferences: # preferences
## Supervision and Habits
- The user prefers to have supervision t...
    - experiences: # experiences
## Daily Routine
- The user went to bed at 1:30 AM last night due ...
    - goals: # goals
## Health and Wellness
- The user is trying to maintain healthy sleep ha...
  Items:
    - [profile] The user prefers to have supervision to build and maintain exercise habits...
    - [profile] The user is diligent with games and does not want to miss out on in-game resources...
    - [profile] The user likes sesame noodles, yogurt parfait, and spicy hotpot...
  Resources:
    - [conversation] example_conversation_#segment_1.json...
    - [conversation] example_conversation_#segment_3.json...
    - [conversation] example_conversation_#segment_1.json...

[POSTGRES] RETRIEVED - LLM
Failed to parse LLM item ranking response: No JSON object found
  Categories:
    - preferences: # preferences
## Supervision and Habits
- The user prefers to have supervision t...
    - activities: # activities
## Exercise
- The user has not exercised much recently due to rainy...
    - goals: # goals
## Health and Wellness
- The user is trying to maintain healthy sleep ha...
  Items:

[POSTGRES] Test completed!

Environment

macOS amd64

Steps to reproduce

python test_inmemory.py python test_postgres.py

Expected behavior

LLM-based retrieve should not output result like:

Failed to parse LLM item ranking response: No JSON object found

Version

v1.1.2

Severity

Minor

Additional Information

No response