Using custom LLM : can't reach litellm api server

[kiruon019-rgb]

Affected Component
AI Agents (Researcher/Developer/...)

Describe the bug
Configuring the custom LLMs (litellm) section to include Qwen/Qwen3.6-27B doesn't work. When I want to create a flow, pentagi can't reach the api server.

Steps to Reproduce
Go to the .env file and in the custom LLMs section fill out the following details -
#Custom LLM provider
LLM_SERVER_URL=https://litellm.xxx.xxx.xxx.net
LLM_SERVER_KEY=sk-secret
LLM_SERVER_MODEL=Qwen/Qwen3.6-27B
LLM_SERVER_PROVIDER=
LLM_SERVER_CONFIG_PATH=
LLM_SERVER_LEGACY_REASONING=
LLM_SERVER_PRESERVE_REASONING=

Run PentAGI with the following changes.

System Configuration
Deployment type: Custom Deployment.
LLM Provider: LiteLLM

Logs and Artifacts
[GIN] 2026/06/25 - 11:12:39 | 200 | 130.114746ms | 10.2.0.1 | GET "/api/v1/info?refresh_cookie=true" time="2026-06-25T11:12:44Z" level=info msg="flow created in DB" flow_id=104 provider_name=custom provider_type=custom user_id=1 time="2026-06-25T11:12:44Z" level=error msg="failed to get flow provider" error="failed to get primary docker image: network error: failed to reach API server" time="2026-06-25T11:12:44Z" level=error msg="graphql request handled with errors" component=pentagi-gql duration=312.207875ms gql.errors="input: createAssistant failed to create flow worker: failed to get flow provider: failed to get primary docker image: network error: failed to reach API server\n" operation_name=createAssistant operation_type=mutation

Screenshots or Recordings

Verification

I have checked that this issue hasn't been already reported

I have provided all relevant configuration files (with sensitive data removed)

I have included relevant logs and error messages

I am running the latest version of PentAGI

[vchaplygim]

@kiruon019-rgb

Hi, thanks for the issue.

Could you confirm whether this is still reproducible on your current setup?

If yes, please check reachability of LiteLLM from inside the PentAGI container (not only from the host):

docker exec <pentagi_container> sh -c 'wget -S -O- --header "Authorization: Bearer $LLM_SERVER_KEY" "$LLM_SERVER_URL/v1/models"'

Share the full output — that will show whether it's DNS, connectivity, TLS, or an HTTP/API path issue.

If the request works from the container:

1. Confirm LLM_SERVER_URL includes the OpenAI-compatible base path your LiteLLM expects (often /v1)
2. Share LiteLLM server-side logs for a fresh failing attempt (if you have access), plus a fresh PentAGI error log