How to use custom searx and ollama instances?

[gerroon]

HI

I set my searx instance url in .env, replacing the default, AS still uses the internal searx which I can see in the logs.

I set my ollam instance url in .env as "OLLAMA_HOST="10.10.0.1:11434" ", replacing the default but I am not sure if it is using bit because the internal searx fails because ther container can only access the internet via proxies, that is why my searx instnace is running on another PC, same as Ollama.

It looks like it is trying to download ollama models, it shoulddnt if it is using the remote instance . I also dont see any Ollama activity on the PC running ollama. I dont even know where it is downloading to.

backend   | INFO:     172.29.0.1:37700 - "GET /latest_answer HTTP/1.1" 200 OK
backend   | INFO:     172.29.0.1:37716 - "GET /screenshots/updated_screen.png?timestamp=1765395668372 HTTP/1.1" 200 OK
▉▉▉▉▅▁ Downloading deepseek-r1:14b...INFO:     172.29.0.1:37726 - "GET /health HTTP/1.1" 200 OK
backend   | INFO:     172.29.0.1:37700 - "GET /latest_answer HTTP/1.1" 200 OK
backend   | INFO:     172.29.0.1:37716 - "GET /screenshots/updated_screen.png?timestamp=1765395671375 HTTP/1.1" 200 OK
▉▅▃▁▃▅ Downloading deepseek-r1:14b...INFO:     172.29.0.1:37726 - "GET /health HTTP/1.1" 200 OK
backend   | INFO:     172.29.0.1:37700 - "GET /latest_answer HTTP/1.1" 200 OK
backend   | INFO:     172.29.0.1:37716 - "GET /screenshots/updated_screen.png?timestamp=1765395674389 HTTP/1.1" 200 OK
▉▇▃▁▁▁ Downloading deepseek-r1:14b...INFO:     172.29.0.1:37726 - "GET /health HTTP/1.1" 200 OK
backend   | INFO:     172.29.0.1:37700 - "GET /latest_answer HTTP/1.1" 200 OK
backend   | INFO:     172.29.0.1:37716 - "GET /screenshots/updated_screen.png?timestamp=1765395677401 HTTP/1.1" 200 OK
Gracefully stopping... (press Ctrl+C again to force)

I have these errors

backend | Selected agent: Browser (roles: web) [446/5683]
backend | ERROR: Exception in ASGI application backend | Traceback (most recent call last): backend | File "/app/sources/llm_provider.py", line 75, in respond backend | thought = llm(history, verbose) backend | ^^^^^^^^^^^^^^^^^^^^^ backend | File "/app/sources/llm_provider.py", line 183, in ollama_fn backend | client.pull(self.model) backend | File "/usr/local/lib/python3.11/site-packages/ollama/_client.py", line 459, in pull backend | return self._request( backend | ^^^^^^^^^^^^^^ backend | File "/usr/local/lib/python3.11/site-packages/ollama/_client.py", line 189, in _request backend | return cls(**self._request_raw(*args, *kwargs).json()) backend | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ backend | File "/usr/local/lib/python3.11/site-packages/ollama/_client.py", line 133, in _request_raw backend | raise ResponseError(e.response.text, e.response.status_code) from None backend | ollama._types.ResponseError: backend | backend | backend | backend | <title>502 - Bad Gateway (Privoxy@debtosh)</title> backend | backend | backend | backend | backend | backend | <style type="text/css"> backend | backend | / backend | * CSS for Privoxy CGI and script output backend | / backend | backend | / backend | * General rules: Font, Color, Headings, Margins, Links backend | */ backend | body,td,th { font-family: arial, helvetica, helv, sans-serif; } backend | body { background-color: #ffffff; color: #000000; } backend | backend | h1 { font-size: 140%; margin: 0px; } backend | h2 { font-size: 120%; margin: 0px; } backend | h3 { font-size: 110%; margin: 0px; } backend | backend | p,pre { margin-left: 15px; } backend | li { margin: 2px 15px; } backend | dl { margin: 2px 15px; } backend | backend | a:link { color: #0000dd; text-decoration: none; } backend | a:visited { color: #330099; text-decoration: none; } backend | a:active { color: #3333ff; text-decoration: none; } backend |

ackend | backend | backend | backend | backend | backend | backend | backend | backend |

No such domain

backend |

Your request for http://host.docker.internal:11434/api/pull backend | could not be fulfilled, because the domain name host.docker.internal could not be resolved. backend |

backend |

This is often a temporary failure, so you might just backend | try again. backend |

backend | backend | backend | backend | backend | backend |

More Privoxy:

[eotstavnov]

I solved all the problems I had with this agent by installing Gemini CLI on my Linux server and using it as a system administrator. It's become incredibly efficient and does everything much faster than me. I recommend it, especially if you have a Gemini subscription, which gives you 1,000 requests per day for each model for free. You can find instructions on how to install Gemini CLI; it's very simple...

[Fosowl]

Hello,

Setting the ollama host in the env file will not work, that is because ollama is not aware of that particular .env file and will not try to load it. The only way would be either export the host export OLLAMA_HOST=0.0.0.0:11434 so that OLLAMA_HOST is added to your global env or to reconfigure ollama address with systemctl edit ollama.service (see ollama docs "Setting environment variables on Linux")

[aniruddhaadak80]

This looks like one of those cases where the environment variables appear correct but the runtime path is still falling back to defaults. When that happens, I usually suspect either the client constructor is not reading the intended values or the container network cannot actually reach the remote host.

I would verify those two things separately. First confirm the effective values inside the running process, then confirm plain connectivity from the container to the target service. That usually tells you quickly whether the bug is configuration parsing or networking.