[Docs/Feedback] Missing environment variables for Ollama configuration causing httpx.ConnectError

[leforense]

Hello team! I wanted to share a troubleshooting experience and suggest a small update to the LLM Provider documentation to save other self-hosting users some debugging pain.

The Problem

I was configuring a local Ollama instance as the AI agent. I followed the official documentation and added the following to my app.env:

AI_AGENT_MODEL="ollama:qwen2.5:7b"
OLLAMA_API_KEY="ollama"
OLLAMA_API_URL="http://ollama:11434" # (or the specific IP of the Ollama container)

However, whenever I tried to use the AI chat in the SysReptor UI, I got a 500 Internal Error. The container logs showed:

sysreptor-app    | 2026-05-20 23:59:38,126 [ERROR] root: All connection attempts failed
sysreptor-app    | httpx.ConnectError: All connection attempts failed
...
sysreptor-app    |   File "/usr/local/lib/python3.13/site-packages/langchain_ollama/chat_models.py", line 1102...
sysreptor-app    |   File "/usr/local/lib/python3.13/site-packages/ollama/_client.py", line 777, in inner

The Debugging Pain and root cause

Initially, I thought this was a Docker network or HTTP Proxy issue. I even ran a wget command from inside the sysreptor-app container pointing to the Ollama container IP, and it worked perfectly. The container had network access, but the Python backend kept failing.

The issue lies in how the underlying langchain_ollama and the official ollama Python client handle environment variables. The Python library does not read OLLAMA_API_URL.

Because the specific variable the library expects was missing, the client silently fell back to its default behavior: attempting to connect to http://localhost:11434. Since SysReptor runs in its own isolated container, it was trying to find Ollama inside the sysreptor-app container (localhost) instead of routing the traffic to my actual Ollama container/host. This resulted in the connection being immediately refused.

The Solution

To fix this, you must pass the exact variables expected by the LangChain/Ollama Python client. Adding OLLAMA_BASE_URL (and OLLAMA_HOST just to be completely safe) to app.env instantly resolved the issue:

AI_AGENT_MODEL="ollama:qwen2.5:7b"
OLLAMA_API_KEY="ollama"
OLLAMA_API_URL="http://ollama:11434"
# Add these so the Python client routes correctly:
OLLAMA_BASE_URL="http://ollama:11434"
OLLAMA_HOST="http://ollama:11434"

(Also, if using corporate proxies, making sure the Ollama hostname is added to NO_PROXY and SYSREPTOR_NO_PROXY is highly recommended).

Suggestion for the Docs

I recommend updating the [LLM Provider documentation] https://docs.sysreptor.com/setup/configuration/#llm-provider for Ollama to include this:

OLLAMA_BASE_URL="https://llm.example.com/"
OLLAMA_HOST="http://ollama:11434"

Thanks.... I spent almost 3 hours trying to understand the problem - lol

[MWedl]

Thanks for pointing out the documentation mistake. We will update the docs. Unfortunately, not all langchain providers support the same set of environment variables.