[PR #1394] [MERGED] docs: drop the nonexistent Ollama reasoning-parser flag and document the think parameter #1360

New issue

Closed

opened 2026-10-18 14:19:07 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-18 14:19:07 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/open-webui/docs/pull/1394
Author: @silentoplayz
Created: 9/18/2026
Status: ✅ Merged
Merged: 9/18/2026
Merged by: @silentoplayz

Base: main ← Head: docs/ollama-thinking

📝 Commits (1)

395981b docs: drop the nonexistent Ollama reasoning-parser flag and document the think parameter

📊 Changes

3 files changed (+36 additions, -66 deletions)

View changed files

📝 docs/features/chat-conversations/chat-features/reasoning-models.mdx (+25 -54)
📝 docs/getting-started/quick-start/connect-a-provider/starting-with-ollama.mdx (+10 -11)
📝 docs/tutorials/integrations/llm-providers/ipex_llm.md (+1 -1)

📄 Description

Summary

Two pages tell Ollama users to start the server with ollama serve --reasoning-parser deepseek_r1 so that thinking blocks display correctly. That flag does not exist in Ollama. ollama serve registers no command-line flags at all (cmd/cmd.go on main), and --reasoning-parser is a vLLM option. Since Ollama v0.9.0 (29 May 2025) thinking is separated natively: the think request parameter turns it on or off (or sets a level: low, medium, high, max), and the trace comes back in a separate thinking field, which Open WebUI already reads as a reasoning delta (utils/response.py, utils/middleware.py).

This PR:

Rewrites the "Using Reasoning / Thinking Models" section of the Ollama quick-start around the think (Ollama) advanced parameter and its four states (Default, On, Custom level, Off), matching what AdvancedParams.svelte sends.
Rewrites the "For Ollama Users" troubleshooting section of the Reasoning & Thinking Models page: no parser flag, check the Ollama version, turn think on, verify from the CLI with ollama run <model> --think. Removes the "Available Reasoning Parsers" table and the ps aux | grep ollama checklist step.
Corrects the same page's think (Ollama) table (adds the Custom level state), the Path 2 sentence, the provider-compatibility row, the non-streaming notes, and the Reasoning Effort note (it is dropped on native Ollama connections, but Ollama's OpenAI-compatible endpoint accepts reasoning_effort).
Fixes https://localhost:11434 to http:// in the IPEX-LLM tutorial.

None. Found while checking the Ollama claims across the docs.

Checklist

I have reviewed the relevant documentation and matched the existing style.
This PR meets Open WebUI's contribution standards: it is accurate, relevant to users, narrowly scoped, maintainable, and not promotional content, advertising, lead generation, SEO placement, or a request to list a product, service, provider, integration, gateway, tool, or company primarily for visibility.
I understand that PRs that do not meet these standards may be closed without review and will not be merged. Repeated, low-quality, off-topic, promotional, or intentionally misleading submissions may result in the contributor being blocked from future participation in Open WebUI repositories.

Notes for reviewers

Sources checked: Ollama cmd/cmd.go (serve has no flags, run has --think and --hidethinking), docs.ollama.com/capabilities/thinking (values, supported models, thinking on by default), docs.ollama.com/api/chat (think accepts true, false, low, medium, high, max), docs.ollama.com/openai (reasoning_effort supported), the v0.9.0 release notes, and Open WebUI utils/payload.py, utils/response.py, utils/middleware.py, AdvancedParams.svelte on dev.

Independent of #1393 (the 2048 default), which touches the context-length section of the same quick-start page in a different hunk.

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/open-webui/docs/pull/1394 **Author:** [@silentoplayz](https://github.com/silentoplayz) **Created:** 9/18/2026 **Status:** ✅ Merged **Merged:** 9/18/2026 **Merged by:** [@silentoplayz](https://github.com/silentoplayz) **Base:** `main` ← **Head:** `docs/ollama-thinking` --- ### 📝 Commits (1) - [`395981b`](https://github.com/open-webui/docs/commit/395981bfa70abdda27e5fd0638898700af5861e1) docs: drop the nonexistent Ollama reasoning-parser flag and document the think parameter ### 📊 Changes **3 files changed** (+36 additions, -66 deletions) <details> <summary>View changed files</summary> 📝 `docs/features/chat-conversations/chat-features/reasoning-models.mdx` (+25 -54) 📝 `docs/getting-started/quick-start/connect-a-provider/starting-with-ollama.mdx` (+10 -11) 📝 `docs/tutorials/integrations/llm-providers/ipex_llm.md` (+1 -1) </details> ### 📄 Description ## Summary Two pages tell Ollama users to start the server with `ollama serve --reasoning-parser deepseek_r1` so that thinking blocks display correctly. That flag does not exist in Ollama. `ollama serve` registers no command-line flags at all (cmd/cmd.go on main), and `--reasoning-parser` is a vLLM option. Since Ollama v0.9.0 (29 May 2025) thinking is separated natively: the `think` request parameter turns it on or off (or sets a level: `low`, `medium`, `high`, `max`), and the trace comes back in a separate `thinking` field, which Open WebUI already reads as a reasoning delta (`utils/response.py`, `utils/middleware.py`). This PR: - Rewrites the "Using Reasoning / Thinking Models" section of the Ollama quick-start around the `think` (Ollama) advanced parameter and its four states (Default, On, Custom level, Off), matching what `AdvancedParams.svelte` sends. - Rewrites the "For Ollama Users" troubleshooting section of the Reasoning & Thinking Models page: no parser flag, check the Ollama version, turn `think` on, verify from the CLI with `ollama run <model> --think`. Removes the "Available Reasoning Parsers" table and the `ps aux | grep ollama` checklist step. - Corrects the same page's `think (Ollama)` table (adds the Custom level state), the Path 2 sentence, the provider-compatibility row, the non-streaming notes, and the Reasoning Effort note (it is dropped on native Ollama connections, but Ollama's OpenAI-compatible endpoint accepts `reasoning_effort`). - Fixes `https://localhost:11434` to `http://` in the IPEX-LLM tutorial. ## Related issue or discussion None. Found while checking the Ollama claims across the docs. ## Checklist - [x] I have reviewed the relevant documentation and matched the existing style. - [x] This PR meets Open WebUI's contribution standards: it is accurate, relevant to users, narrowly scoped, maintainable, and not promotional content, advertising, lead generation, SEO placement, or a request to list a product, service, provider, integration, gateway, tool, or company primarily for visibility. - [x] I understand that PRs that do not meet these standards may be closed without review and will not be merged. Repeated, low-quality, off-topic, promotional, or intentionally misleading submissions may result in the contributor being blocked from future participation in Open WebUI repositories. ## Notes for reviewers Sources checked: Ollama `cmd/cmd.go` (serve has no flags, run has `--think` and `--hidethinking`), docs.ollama.com/capabilities/thinking (values, supported models, thinking on by default), docs.ollama.com/api/chat (`think` accepts `true`, `false`, `low`, `medium`, `high`, `max`), docs.ollama.com/openai (`reasoning_effort` supported), the v0.9.0 release notes, and Open WebUI `utils/payload.py`, `utils/response.py`, `utils/middleware.py`, `AdvancedParams.svelte` on `dev`. Independent of #1393 (the 2048 default), which touches the context-length section of the same quick-start page in a different hunk. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-18 14:19:07 +00:00