Ollama users keep filing bugs

Provide API helping at checking for updates and applying them

I think Ollama should gain an API to let applications check if it is up-to-date (by checking the installed version against the latest one by probing the official channels, e.g: github releases, ollama site) and also an API to permit the update of the existing installation....

feat: Support MinerU 4.1 Tasks API in the built-in MinerU document loader

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue(a11y): pressing Tab while recording a shortcut binds it to Tab, and Tab then stops moving focus anywhere in the app

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

Add support for Unix domain sockets (IPC) alongside TCP/IP

Problem Currently, Ollama listens exclusively on TCP/IP addresses (such as 127.0.0.1:11434 or custom IPv4/IPv6 bindings). While this ensures broad cross-platform compatibility, it introduces unnecessary overhead and lacks local IPC isolation for setups running entirely on...

docs: document `ollama cp` and `ollama push` in CLI reference

## Problem The CLI Reference covers several model-management commands but does not document the supported `ollama cp SOURCE DESTINATION` or `ollama push MODEL` commands. Users may not discover how to duplicate a local model or push one to a registry. ## Proposed change Add...

No support for EAI email addresses

### What is the issue? When an account is created - whether it is the admin account at the outset or a user account added later - it is currently not possible to use non-ASCII characters in the email address. This has a significant impact on the local part (before the @ symbol),...

I subscribed to Ollama's cloud/paid tier to access cloud models via API, but I am unable to access the service that I paid for.

...- No response after multiple attempts - Tried Discord - Unable to get help there - Tried Twitter/X - No resolution - Checked accounts page - Payment went through but API access is non-functional - There is no contact form - ollama.com/contact returns 404 Expected I need either:...

[BUG] Prices from MLFLOW_MODEL_CATALOG_URI are ignored when litellm is installed

### Issues Policy acknowledgement - [x] I have read and agree to submit bug reports in accordance with the [issues policy](https://www.github.com/mlflow/mlflow/blob/master/ISSUE_POLICY.md) ### Where did you encounter this bug? Other ### MLflow version 3.17.2 ### System...

bug: [arm64] Please cap torch for aarch64 via environment marker

...Summary When installing open-webui[all] on a lxc arm64 from a orange pi 5 plus, it will fail to run because it pulls in torch==2.15.0+cu132. However, this cannot run on this CPU. LLM detailed a fix here: ### Expected Behavior see above ### Actual Behavior see above ### Steps to...

issue(a11y): Valves fields and the Chat Controls Valves selects have no names

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue(a11y): a group's checkbox in Add Access has no name and sits inside the group's button

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue(a11y): a filter's clear (x) button has no name and sits inside the filter's button

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue(a11y): fields on the model editor, knowledge, notes and playground pages have no name besides their placeholder

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue(a11y): after scrolling the model list with the mouse, the arrow keys move to a row that isn't there

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue(a11y): fields in many windows have no name for screen readers, or only their placeholder

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue(a11y): on the calendar, "and N more" can't be reached with the keyboard, and events sit inside the day's button

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue(a11y): a notebook's cells can only be opened for editing with a double-click

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue(a11y): the sidebar can only be resized with the mouse

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue(a11y): in a terminal's Ports list, Refresh sits inside the Ports button and "Open in new tab" can't be reached with the keyboard

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue(a11y): the emoji in "Update your status" can't be reached with the keyboard

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue(a11y): a multi-select valve's options can't be reached with the keyboard, and Escape closes the whole Valves window

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue(a11y): the Analytics tables and a group's member list can't be sorted from the keyboard

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue(a11y): most windows have no name, so screen readers announce just "dialog"

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue(a11y): in Admin Analytics, a model's details can only be opened with the mouse

...in the Models table opens a window with that model's details. From the keyboard, there's no way to open it. Tab never stops on a model, because the whole row is clickable but nothing in it can take focus. ### Expected Behavior Each model's name can be reached with Tab, and Enter...

Please make Ollama model library on ollama.com filterable for MLX yes/no

It's not the first time but I encountered this again with the new lfm3 models: No hint on the model page that these require MLX. I wish it would be possible to filter such models out (I don't have Apple silicon).

[Feature]: Support Ollama decision models through the Decisions API

...'ollama_chat'`. The same happens with `ollama/d3:4b` and on `POST /v1/decisions` 4. The workaround is to configure the Ollama server a second time as `hosted_vllm/d3:4b` with `api_base` set to ` /v1`. This works, but the deployment is labeled as a vLLM server in the proxy and in...

Gpt oss

[Feature Request] Add Korean language support for the new TTS (phonon) model

I am very excited about the recent addition of the Text-to-Speech (TTS) feature using the phonon model in Ollama. It’s an incredible update! Currently, the phonon:40m-multilingual model officially supports 5 languages (English, French, German, Spanish, and Portuguese). I would...

Ollama does not work even after a fresh install

...it. I tried uninstall and reinstall of ollama even cleaning all models. but still it does not work even when the download and manifest shows success after verification, the model is not listed in ### Relevant log output ### OS Windows ### GPU Nvidia ### CPU Intel ### Ollama...

Proposal: add OpenSeal to `ollama launch`

OpenSeal (https://github.com/axiom-studio/openseal) is an open-source (Apache-2.0) runtime for autonomous agents and teams. It calls its model through an OpenAI-compatible Chat Completions endpoint with tool calling, which Ollama provides, but users currently have to configure...

`stop` is ignored by models on the MLX engine

### What is the issue? On the MLX engine, `options.stop` has no effect and generation runs until EOS or `num_predict`. The same request stops at the stop sequence on a model that runs on llama.cpp. | model | engine | `message.content` | |---|---|---| | `gemma4:26b-nvfp4` | MLX |...

Does OLLAMA_NUM_PARALLEL=1 do anything?

### What is the issue? I ran container with ` -e OLLAMA_NUM_PARALLEL=1 `. But I have now two goose sessions processing, outputting simultaneously to the chat. I can't have two sessions run at once, because I have limited memory, everything needs to run in sequence. newest...

issue: # knowledge attach in chat does not reach the model, while model-level meta.knowledge works

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue(a11y): sidebar section titles and chat date groups aren't headings for screen readers

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

install.sh calling sudo immediately after download forces user to be present at download completion

### What is the issue? `install.sh` downloads ollamaxxx.tar.zst and uses sudo to extract the file. If user is not present the sudo password prompt will timeout. On slower connections it's practical to leave the install unattended while the download completes. However if sudo...

issue(a11y): The Archived Chats and Keyboard lists in Settings shrink to almost nothing at 400% zoom

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue: Release workflow runs after unrelated main pushes

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue: Compose launcher help examples fail when pasted into zsh

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue: chat hover preview squeezes multi-model replies into unreadable slivers

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue: the model picker scrolls away and stacks menus as you click its "..." buttons

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

mlx: create refuses MLX checkpoints whose config.json gives only model_type

### What is the issue? Some MLX checkpoints' config.json gives a `model_type` but no `architectures` list. `ollama create` reads only `architectures[0]`, so it refuses them even when the `model_type` names an architecture the MLX runner supports, and the runner's model loader...

[ollama.com] MLX filter to Models list

Hi, Would it be possible to add an "MLX" filter to the Models page of ollama.com, please? Thanks.

issue: Admin viewing another user's completed chat repeatedly jumps to the bottom

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

Auto-Start of ollama without permisson

### What is the issue? Ollama auto starts, even if: - auto update is disabled - its removed from the auto start via the windows taskmanger/autostart - and no copilot or other service that would launch it is running Please fix, or I'll have no choice but to report it as malware,...

issue: scrolling a chat right after clicking Notes throws a TypeError in the console

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

QOL Request - folders for chats

Can we get folders added to the chat section? im dealing with lots of projects it would be nice if i could group chats that are connected to a project, as i spend ages going through each chat to see whats been done. - thanks

Error running rnj-1: llama-server process has terminated: GGML_ASSERT(hparams.is_swa_any()) failed: GGML_ASSERT(hparams.is_swa_any()) failed

### What is the issue? Ever since updating my Ollama version above 0.3x ish, `ollama run rnj-1` fails with the following error: I tried removing and redownloading multiple times, but it still fails. ### Relevant log output journalctl -u ollama --no-pager --follow --pager-end Oct...

Option to disable automatic model upgrade

Would it be possible to **disable** the new "automatic model upgrade" introduced with release v0.40.2? I always have very limited capacity left on my SSD. An automatic download could cause issues or even a "no space left on device" error. So it would be nice if this could be...

Gemma4:12b throws: "Gemma4Assistant requires ctx_other to be set"

### What is the issue? When I'm calling the gemma4 12b model I receive the error below. Gemma4:latest runs without issues though. ### Relevant log output ### OS Windows ### GPU Nvidia ### CPU Intel ### Ollama version 0.40.2

Error: image generation models are not currently supported

### What is the issue? $ ollama list NAME ID SIZE MODIFIED x/z-image-turbo:latest 77b78ce4e883 12 GB 12 minutes ago $ollama run x/z-image-turbo Error: image generation models are not currently supported ### Relevant log output ### OS macOS ### GPU Apple ### CPU Apple ### Ollama...

Probabilities output is messed up

### What is the issue? As shown in the picture, when I swap the items in questions.criteria in the request body, the output results are inconsistent. api: http://ip:11434/v1/systemone version: 0.40.1 ### Relevant log output ### OS Windows ### GPU Intel ### CPU Intel ### Ollama...

ffn_down_exps.weight size overflows

### What is the issue? run gpt-oss:latest in ollama version 0.40.1 ### Relevant log output ### OS Linux ### GPU Nvidia ### CPU Intel ### Ollama version 0.40.1

manifests-v2

### What is the issue? What is manifests-v2? Why is it creating strange aliases? What is the impact for ollama servers of different versions sharing storage? Why is a what seems to be a significant structural change not announced? ### Relevant log output ### OS Linux ### GPU...

Error when trying to use Clef-flash

### What is the issue? I installed the Ollama app at this path: D:\OllamaApp and I put the models location at this path: D:\Ollama\models https://ollama.com/library/clef-flash **When I installed clef-flash and try the example request on the site with curl, i get this error...

ollama: `ChatOllama` sends tool results without `tool_name`, so models see them as an unknown tool

### Submission checklist - [x] This is a bug, not a usage question. - [x] I added a clear and descriptive title that summarizes this issue. - [x] I used the GitHub search to find a similar question and didn't find it. - [x] I am sure that this is a bug in LangChain rather than...

Ollama home page issue

### What is the issue? The page got issue. I try clear cache and cookies but it same. ### Relevant log output ### OS _No response_ ### GPU _No response_ ### CPU _No response_ ### Ollama version _No response_

issue: the model selector's filter list opens off the bottom of the window

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

device index is incremented for filtered pseudo-devices

### What is the issue? In discover/llama_server.go, the device index was incremented even when a device was discarded because it reported zero memory (e.g. the BLAS pseudo-device). This causes subsequent real devices to receive incorrect device indices. The index must only be...

extend `ollama launch` to browsers (Chrome, Edge, Firefox, Opera, Brave)

`ollama launch` currently supports CLI/agent integrations and VS Code. I'd like to propose extending it to **web browsers** - Chrome, Edge, Firefox, Opera, Brave - so an Ollama model (local at `127.0.0.1:11434` or Ollama Cloud) can power the browser's AI assistant through a...

Is 0.35.0 a pre-release?

### What is the issue? 0.35.0 is marked as a pre-release at https://github.com/ollama/ollama/releases/tag/v0.35.0 but doesn't contain an `-rc` suffix like other release candidates. Is this intended?

issue: imported chats all show as unread in the sidebar

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue: disabled skills are still offered in the $ and / pickers, and picking one does nothing

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue: archiving a chat loses its tags' names, so they come back in lowercase with underscores

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue: saving Settings > Interface replaces the user's default model

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue: pressing Enter on a highlighted chat in the search dialog closes it without opening the chat

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue: saving a repeating calendar event opened from a later occurrence deletes every earlier occurrence

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

issue: deleting the open chat with the Delete Chat shortcut leaves it listed in its sidebar folder

### Before Submitting - [x] I searched open and closed issues and discussions for an existing report. - [x] I reproduced this bug on the latest release AND on the current `dev` branch, right before submitting this report. I did not just check an old version or rely on a check...

More granular memory split controls

For users with multiple GPU setups, it would be great to have more control over how ollama allocates memory to the cards. In particular, in my case, I have three GPUs, with the third on a slow pcie3 slot. The goal was to use this card for k/v cache overflow (some models were...

Enable to show thinking blocks in the TUI

...thought. In the desktop app you can click at it to show the content, in the tui I know no way to show the thinking text again. Please provide a option to never collapse the thinking text

Support `spark2_5` architecture (Spark-X2.5-4B / 1.7B)

...does not yet recognize the `spark2_5` architecture. The model page itself documents a workaround that requires building a forked llama.cpp. Why I think this is worth supporting: - Compact, general-purpose, bilingual model (4B and 1.7B) with a native 1M-token context and 200+...

Need to have qwen3.8-flash-next on Ollama cloud.

It is a very good model and it's capabilities should be available for everyone to use via cloud as well.