Error: something went wrong, please see the ollama server logs for details from ollama list, ollama ps or ollama show appeared in this test when OLLAMA_HOST carried a path, http://127.0.0.1:15960/api. Remove the path, keep only scheme, host and port, and the same commands work.

The server log, as the message suggests, shows the reason: every command sent HEAD "/api/" and got a 404.

The error

PS> $env:OLLAMA_HOST = "http://127.0.0.1:15960/api"
PS> ollama list
Error: something went wrong, please see the ollama server logs for details
exit=1
PS> ollama ps
Error: something went wrong, please see the ollama server logs for details
exit=1
PS> ollama show qwen2.5:0.5b
Error: something went wrong, please see the ollama server logs for details
exit=1

Why it happens

The value looks reasonable, because Ollama's endpoints do live under /api. But the server log of the test server shows what the client asked for with that value: 

PS> docker logs --since 1m fix9-olc-srv 2>&1 | ForEach-Object { "$_" } | Select-String '/api' | ForEach-Object { "$_" }
...
[GIN] 2026/11/10 - 03:26:30 | 404 |      16.877µs |      172.17.0.1 | HEAD     "/api/"
[GIN] 2026/11/10 - 03:26:30 | 404 |      16.358µs |      172.17.0.1 | HEAD     "/api/"
[GIN] 2026/11/10 - 03:26:30 | 404 |      16.279µs |      172.17.0.1 | HEAD     "/api/"

One HEAD "/api/" per command, each answered 404, and the client turned that into the generic message. A minute earlier the same server had logged GET "/api/tags" with 200 for an ollama list run without the path.

Each log line reads from left to right as date and time, HTTP status, time taken, the client address, the method and the path. When the CLI says "something went wrong", those last columns are where to look: in this test a 404 on a path nobody meant to call pointed at the address in OLLAMA_HOST, not at the model or the server. Only the container's log was read here; the Windows app and a Linux service keep theirs in other places. A request to the doubled path answers 404 as well: 

PS> curl.exe -s -i http://127.0.0.1:15960/api/api/tags
HTTP/1.1 404 Not Found
Content-Type: text/plain
...
404 page not found

The fix

PS> $env:OLLAMA_HOST = "http://127.0.0.1:15960"
PS> ollama list
NAME    ID    SIZE    MODIFIED
exit=0
PS> ollama ps
NAME    ID    SIZE    PROCESSOR    CONTEXT    RUNNER    UNTIL
exit=0

Both tables are empty because the test server had no models and nothing loaded; the headers show that the server answered. In the test OLLAMA_HOST worked with the address of the server and failed with the address of its API. A trailing slash with no path, http://127.0.0.1:15960/, also worked in the same session, and so did the plain 127.0.0.1:15960 without a scheme, so the only part to remove is the path.

How it was reproduced

Windows client ollama version is 0.40.14 in Windows PowerShell 5.1, with OLLAMA_HOST set in the script's own process only. The server was docker run -d --name fix9-olc-srv -p 127.0.0.1:15960:11434 ollama/ollama (server 0.40.8), the test's own container, removed afterwards. Its log was read with docker logs --since 1m right after the three failing commands.

Frequently asked

What does Ollama something went wrong, please see the ollama server logs mean?
In this test the server answered HEAD /api/ with 404 and the client printed only this generic message; the cause was a path in OLLAMA_HOST.
Should OLLAMA_HOST include /api?
No. In the test http://127.0.0.1:15960/api gave something went wrong for list, ps and show, and http://127.0.0.1:15960 worked.
Where are the Ollama server logs?
For the Docker container in this test, docker logs with the container name printed them. The log lines show each request path and its status code.

More decoded errors in the Fixes category. Other OLLAMA_HOST typos and their messages are in invalid host: ParseAddr unexpected character.