[Bug] [Module Name] No model service available

[Retire-At35]

Search before asking

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

Operating system information

Linux

Python version information

=3.11

What happened

When I run the AI-SRE feature, it works properly if I use the derisk-proxy-deepseek.toml configuration file, but only with the deepseek-r1 model. If I use configuration files for other large models(such as openai、tongyi), the frontend reports an error: “Failed to allocate model service, Failed to allocate model service, No model service available!!!”

What you expected to happen

I have tested that direct conversations with the model work fine. The issue only occurs with the AI-SRE feature. During runtime, the terminal outputs the following logs, which caught my attention:

2025-12-27 17:54:41 A10-1 derisk.agent.core.base_agent[3177376] INFO Agent[Kevin:ai-sre] Final build:
Resources: None
Models: ['Kimi-K2-Instruct', 'deepseek-r1']
Parameters: None
Memory: None
2025-12-27 17:54:41 A10-1 derisk_serve.agent.agents.chat.agent_chat[3177376] INFO _build_agent_by_gpts: ai-reporter, Reporter(Kevin), end: 2025-12-27 17:54:41.428727
2025-12-27 17:54:41 A10-1 derisk_serve.agent.agents.chat.agent_chat[3177376] INFO _build_employees loop: <derisk.util.configure.base.ConfigInfo object at 0x7b261005c810>, Reporter(Kevin). Now you have decided to complete the reasoning process. You should now provide the final answer to the question. Possible root cause components and cause candidates have been provided. You must choose the root cause component and cause from the given candidates., 135404767932448, end: 2025-12-27 17:54:41.428833
2025-12-27 17:54:41 A10-1 derisk.agent.core.base_agent[3177376] INFO register_variables Coder
2025-12-27 17:54:41 A10-1 derisk.agent.core.base_agent[3177376] INFO register_variables end Coder
2025-12-27 17:54:43 A10-1 derisk_serve.rag.service.service[3177376] INFO run_refresh_task start 2025-12-27 17:54:43.549862
2025-12-27 17:54:43 A10-1 derisk_serve.rag.service.service[3177376] INFO run_refresh_task start query db: refresh hour is 0
2025-12-27 17:54:43 A10-1 derisk_serve.rag.service.service[3177376] INFO auto_refresh_periodic wait hours is 6.087901560555556h, next_run_time is 2025-12-28 00:00:00
2025-12-27 17:54:43 A10-1 derisk.agent.core.memory.gpts.gpts_memory[3177376] INFO memory object released!
2025-12-27 17:54:43 A10-1 derisk.agent.core.base_agent[3177376] INFO Agent[Magic:ai-sre] Final build:
Resources: None
Models: ['deepseek-v3', 'deepseek-r1']
Parameters: None
Memory: None
2025-12-27 17:54:43 A10-1 derisk_serve.agent.agents.chat.agent_chat[3177376] INFO _build_agent_by_gpts: ai-coder, Coder(Magic), end: 2025-12-27 17:54:43.749826
2025-12-27 17:54:43 A10-1 derisk_serve.agent.agents.chat.agent_chat[3177376] INFO _build_employees loop: <derisk.util.configure.base.ConfigInfo object at 0x7b26100a7890>, Coder(Magic). Your goal is to write Python code to answer DevOps questions. For each question, you need to write Python code that solves the problem by retrieving and processing telemetry data from the target system. The Python code you generate will be automatically submitted to the IPython kernel. The execution output from the IPython kernel will be used as the answer., 135407534443728, end: 2025-12-27 17:54:43.749979
2025-12-27 17:54:43 A10-1 derisk.agent.core.base_agent[3177376] INFO register_variables DevOps
2025-12-27 17:54:43 A10-1 derisk.agent.core.base_agent[3177376] INFO register_variables end DevOps
2025-12-27 17:54:43 A10-1 derisk.agent.core.memory.gpts.gpts_memory[3177376] INFO memory object released!
2025-12-27 17:54:43 A10-1 derisk.agent.core.base_agent[3177376] INFO Agent[Devid:ai-sre] Final build:
Resources: None
Models: ['deepseek-r1']
Parameters: None
Memory: None
2025-12-27 17:54:43 A10-1 derisk_serve.agent.agents.chat.agent_chat[3177376] INFO _build_agent_by_gpts: sre-planner, DevOps(Devid), end: 2025-12-27 17:54:43.751941
2025-12-27 17:54:43 A10-1 derisk_serve.agent.agents.chat.agent_chat[3177376] INFO _build_employees loop: <derisk.util.configure.base.ConfigInfo object at 0x7b2603f022d0>, DevOps(Devid). To solve each given problem, you need to iteratively instruct the agents to work and perform data analysis on the telemetry files of the target system. By analyzing the execution results, you need to gradually converge on the answer., 135405131399888, end: 2025-12-27 17:54:43.752063
2025-12-27 17:54:43 A10-1 derisk.util.api_utils[3177376] WARNING No healthy urls found, selecting randomly
2025-12-27 17:54:43 A10-1 derisk.util.api_utils[3177376] WARNING No healthy urls found, selecting randomly
2025-12-27 17:54:43 A10-1 derisk.util.api_utils[3177376] WARNING No healthy urls found, selecting randomly
2025-12-27 17:54:43 A10-1 derisk.util.api_utils[3177376] WARNING No healthy urls found, selecting randomly
2025-12-27 17:54:43 A10-1 derisk_serve.agent.agents.chat.agent_chat[3177376] INFO _build_employees return: ['Reporter(Kevin)', 'Coder(Magic)', 'DevOps(Devid)'], end: 2025-12-27 17:54:43.792474
2025-12-27 17:54:43 A10-1 derisk.agent.core.memory.gpts.gpts_memory[3177376] INFO memory object released!
2025-12-27 17:54:43 A10-1 derisk.agent.core.base_agent[3177376] INFO register_variables AI-SRE(DeRisk)
2025-12-27 17:54:43 A10-1 derisk.agent.core.base_agent[3177376] INFO register_variables end AI-SRE(DeRisk)
2025-12-27 17:54:43 A10-1 derisk.agent.core.memory.gpts.gpts_memory[3177376] INFO memory object released!
2025-12-27 17:54:43 A10-1 derisk.agent.core.base_agent[3177376] INFO Agent[AI-SRE(DeRisk): ai-sre] Final build:
Resources: <class 'derisk.agent.resource.app.AppResource.resource_parameters_class.._DynAppResourceParameters'>; <class 'derisk.agent.resource.app.AppResource.resource_parameters_class.._DynAppResourceParameters'>; <class 'derisk.agent.resource.app.AppResource.resource_parameters_class.._DynAppResourceParameters'>; <class 'derisk_ext.agent.agents.open_rca.resource.open_rca_resource.OpenRcaSceneResource.resource_parameters_class.._DynDBParameters'>
Models: ['deepseek-r1']
Parameters: None
Memory: None
2025-12-27 17:54:43 A10-1 derisk_serve.agent.agents.chat.agent_chat[3177376] INFO _build_agent_by_gpts return: AI-SRE(DeRisk), AI-driven root cause analysis agent that cuts through complex system noise to precisely pinpoint the source of problems., 135404753345568
2025-12-27 17:54:43 A10-1 derisk_serve.agent.agents.chat.agent_chat[3177376] INFO _build_agent_by_gpts: ai-sre, AI-SRE(DeRisk), end: 2025-12-27 17:54:43.846787
2025-12-27 17:54:43 A10-1 derisk.agent.core.base_agent[3177376] INFO register_variables Human
2025-12-27 17:54:43 A10-1 derisk.agent.core.base_agent[3177376] INFO register_variables end Human
2025-12-27 17:54:43 A10-1 derisk.agent.core.memory.gpts.gpts_memory[3177376] INFO memory object released!
2025-12-27 17:54:43 A10-1 derisk.agent.core.base_agent[3177376] INFO Agent[User: ai-sre] Final build:
Resources: None
Models: None
Parameters: None
Memory: None
2025-12-27 17:54:43 A10-1 context[3177376] INFO context window, enter --> [1]

This output appears regardless of whether I use OpenAI, Tongyi, or other models. This leads me to suspect that the AI-SRE feature may be hard-coded to only work with deepseek-r1 (or a small subset of models). However, the project’s file structure is quite complex, and I haven’t been able to locate where the supported models for the AI-SRE feature are defined.

How to reproduce

I simply cloned the project and followed the README step by step. After starting the service using commands such as
uv run python packages/derisk-app/src/derisk_app/derisk_server.py --config configs/derisk-proxy-tongyi.toml,
then the issue occurs.

Additional context

No response

Are you willing to submit PR?

• Yes I am willing to submit a PR!

[yhjun1026]

tongyi model config such as ：
[[models.llms]]
name = "qwen-vl-max"
provider = "proxy/tongyi"
api_base = "https://dashscope.aliyuncs.com/compatible-mode/v1"
api_key = "sk-xxxx"
backend = "qwen-vl-max"

openai model：
[models]
[[models.llms]]
name = "${env:LLM_MODEL_NAME:-gpt-4o}"
provider = "${env:LLM_MODEL_PROVIDER:-proxy/openai}"
api_base = "${env:OPENAI_API_BASE:-https://api.openai.com/v1}"
api_key = "${env:OPENAI_API_KEY}"

Use the provider configuration for each model.

[Retire-At35]

It still does not work. I suspect that the model configuration might be hard-coded somewhere in the codebase, so the configuration file I specified is not actually being used.

Additionally, I did not see any option in the frontend to select which LLM the AI-SRE feature should use. This makes me wonder where in the project code the set of LLMs available to the AI-SRE functionality is defined.

[Retire-At35]

I later modified part of the code and solved the problem. Thanks!