[PR #15122] runner: Replace llama CGO runner with llama-server #20310

Open
opened 2026-04-16 07:32:56 -05:00 by GiteaMirror · 0 comments
Owner

📋 Pull Request Information

Original PR: https://github.com/ollama/ollama/pull/15122
Author: @dhiltgen
Created: 3/28/2026
Status: 🔄 Open

Base: mainHead: llama-runner


📝 Commits (2)

  • e65ccf4 runner: Replace llama runner with llama-server
  • d84623f fix - cross-compile bug

📊 Changes

69 files changed (+46405 additions, -2191 deletions)

View changed files

📝 .github/workflows/release.yaml (+8 -0)
📝 .github/workflows/test.yaml (+12 -0)
📝 CMakeLists.txt (+178 -6)
📝 CMakePresets.json (+22 -0)
📝 Dockerfile (+19 -1)
📝 llama/llama.cpp/.rsync-filter (+19 -6)
llama/llama.cpp/common/arg.cpp (+3415 -0)
llama/llama.cpp/common/arg.h (+129 -0)
llama/llama.cpp/common/chat-parser-xml-toolcall.cpp (+879 -0)
llama/llama.cpp/common/chat-parser-xml-toolcall.h (+45 -0)
llama/llama.cpp/common/chat-parser.cpp (+1543 -0)
llama/llama.cpp/common/chat-parser.h (+133 -0)
llama/llama.cpp/common/chat-peg-parser.cpp (+124 -0)
llama/llama.cpp/common/chat-peg-parser.h (+105 -0)
llama/llama.cpp/common/chat.cpp (+2870 -0)
llama/llama.cpp/common/chat.h (+233 -0)
llama/llama.cpp/common/console.cpp (+1137 -0)
llama/llama.cpp/common/console.h (+41 -0)
llama/llama.cpp/common/download.cpp (+1126 -0)
llama/llama.cpp/common/download.h (+57 -0)

...and 49 more files

📄 Description

This change replumbs the llama runner subprocess to use the llama-server directly instead of the CGO based llama runner. This allows us to take better advantage of new capabilities from llama.cpp.


🔄 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/ollama/ollama/pull/15122 **Author:** [@dhiltgen](https://github.com/dhiltgen) **Created:** 3/28/2026 **Status:** 🔄 Open **Base:** `main` ← **Head:** `llama-runner` --- ### 📝 Commits (2) - [`e65ccf4`](https://github.com/ollama/ollama/commit/e65ccf4840912d81fdfa4b7847cfb02dbe9fc7b6) runner: Replace llama runner with llama-server - [`d84623f`](https://github.com/ollama/ollama/commit/d84623fcadb7bc813c7b0211529d498956427643) fix - cross-compile bug ### 📊 Changes **69 files changed** (+46405 additions, -2191 deletions) <details> <summary>View changed files</summary> 📝 `.github/workflows/release.yaml` (+8 -0) 📝 `.github/workflows/test.yaml` (+12 -0) 📝 `CMakeLists.txt` (+178 -6) 📝 `CMakePresets.json` (+22 -0) 📝 `Dockerfile` (+19 -1) 📝 `llama/llama.cpp/.rsync-filter` (+19 -6) ➕ `llama/llama.cpp/common/arg.cpp` (+3415 -0) ➕ `llama/llama.cpp/common/arg.h` (+129 -0) ➕ `llama/llama.cpp/common/chat-parser-xml-toolcall.cpp` (+879 -0) ➕ `llama/llama.cpp/common/chat-parser-xml-toolcall.h` (+45 -0) ➕ `llama/llama.cpp/common/chat-parser.cpp` (+1543 -0) ➕ `llama/llama.cpp/common/chat-parser.h` (+133 -0) ➕ `llama/llama.cpp/common/chat-peg-parser.cpp` (+124 -0) ➕ `llama/llama.cpp/common/chat-peg-parser.h` (+105 -0) ➕ `llama/llama.cpp/common/chat.cpp` (+2870 -0) ➕ `llama/llama.cpp/common/chat.h` (+233 -0) ➕ `llama/llama.cpp/common/console.cpp` (+1137 -0) ➕ `llama/llama.cpp/common/console.h` (+41 -0) ➕ `llama/llama.cpp/common/download.cpp` (+1126 -0) ➕ `llama/llama.cpp/common/download.h` (+57 -0) _...and 49 more files_ </details> ### 📄 Description This change replumbs the llama runner subprocess to use the llama-server directly instead of the CGO based llama runner. This allows us to take better advantage of new capabilities from llama.cpp. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
GiteaMirror added the pull-request label 2026-04-16 07:32:56 -05:00
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#20310