[Bug] [agent] The LLM response content from the agent is missing.

[FOkvj]

Search before asking

• I had searched in the issues and found no similar issues.

Operating system information

Linux

Python version information

=3.11

DB-GPT version

main

Related scenes

• Chat Data
• Chat Excel
• Chat DB
• Chat Knowledge
• Model Management
• Dashboard
• Plugins

Installation Information

• Installation From Source

• Docker Installation

• Docker Compose Installation

• Cluster Installation

• AutoDL Image

• Other

Device information

Models information

deepseek v3

What happened

What you expected to happen

In the agent module, when processing the LLM's response, if the response contains the ### symbol, the content before ### will be truncated.

How to reproduce

1. execute the coede below

import asyncio

from dbgpt.agent import AgentContext, AgentMemory, LLMConfig, UserProxyAgent, ProfileConfig, ConversableAgent, \
   BlankAction
from dbgpt.model.proxy import SiliconFlowLLMClient


async def main():

   llm_client = SiliconFlowLLMClient(
       model_alias=os.getenv(
           "SILICONFLOW_MODEL_VERSION", "Qwen/Qwen2.5-Coder-32B-Instruct"
       ),
   )
   context: AgentContext = AgentContext(conv_id="test456")

   agent_memory = AgentMemory()
   agent_memory.gpts_memory.init(conv_id="test456")

   user_proxy = await UserProxyAgent().bind(agent_memory).bind(context).build()

   # 代码优化助手
   code_optimizer_profile = ProfileConfig(
       name="Jack",
       role="CodeOptimizer",
       goal="优化用户给的代码。",
       desc="专门对用户代码进行优化",
       constraints=["要给出代码优化前后的对比，用markdown格式展示代码"],
   )
   code_optimizer = (
       await ConversableAgent(profile=code_optimizer_profile)
       .bind(context)
       .bind(LLMConfig(llm_client=llm_client))
       .bind(agent_memory)
       .bind(BlankAction)
       .build()
   )

   await user_proxy.initiate_chat(
       recipient=code_optimizer,
       reviewer=user_proxy,
       message="select * from table 帮我优化代码，解释部分和代码用###分割",
   )

   ## dbgpt-vis message infos
   print(await agent_memory.gpts_memory.app_link_chat_message("test456"))


if __name__ == "__main__":
   asyncio.run(main())

model original output

### Explanation
The SQL query `SELECT * FROM table` retrieves all columns from the specified table. While this is a simple and straightforward query, it can be optimized in several ways depending on the context:

1. **Specific Columns**: Instead of selecting all columns (`*`), specify only the columns you need. This reduces the amount of data transferred and processed, which can improve performance, especially for large tables.

2. **Index Usage**: Ensure that the columns you are querying are indexed, especially if you are filtering or sorting the data. This can significantly speed up the query.

3. **Limit Clause**: If you only need a subset of the data, consider using a `LIMIT` clause to restrict the number of rows returned.

4. **Avoiding SELECT ***: Using `SELECT *` can lead to issues if the table schema changes (e.g., columns are added or removed). Explicitly listing the columns you need makes your query more robust.

### Optimized Code
Here’s an optimized version of the query, assuming you only need specific columns and a limited number of rows:

```sql
-- Original Query
SELECT * FROM table;

-- Optimized Query
SELECT column1, column2, column3 
FROM table
LIMIT 100;
```

### Explanation of Optimized Code
- **Specific Columns**: The optimized query selects only `column1`, `column2`, and `column3` instead of all columns. This reduces the amount of data processed and transferred.
- **LIMIT Clause**: The `LIMIT 100` clause restricts the result set to 100 rows, which can be useful if you only need a sample of the data or are working with a large dataset.

This optimized query is more efficient and safer to use in production environments.

processed output

 Explanation of Optimized Code
- **Specific Columns**: The optimized query selects only `column1`, `column2`, and `column3` instead of all columns. This reduces the amount of data processed and transferred.
- **LIMIT Clause**: The `LIMIT 100` clause restricts the result set to 100 rows, which can be useful if you only need a sample of the data or are working with a large dataset.

This optimized query is more efficient and safer to use in production environments.

Additional context

No response

Are you willing to submit PR?

• Yes I am willing to submit a PR!