[tutorial] Using Ollama with OpenCommit for Local Commit Message Generation

📌 Introduction

This article covers using Ollama with OpenCommit for generating commit messages locally. It includes an overview of running Ollama in a Docker container, instructions for using the Ollama CLI, and how to combine Ollama with OpenCommit for generating commit messages.

🚀 Quick Start

Start a container

docker run -d -v ollama:/root/.ollama -p 11434:11434 --name ollama-commit ollama/ollama:0.3.6

Enter the Docker container

docker exec -it ollama-commit bash

Pull model

ollama run gemma2:2b

Start a chat

>>> Send a message (/? for help)

Exit the chat

Type /bye to exit the chat.

Install opencommit

npm install -g opencommit

Generate commit messages with local ollama server

OCO_AI_PROVIDER='ollama/gemma2:2b' opencommit

output:

┌  open-commit
│
◇  1 staged files:
  README.md
│
◇  📝 Commit message generated
│
└  Generated commit message:
——————————————————
feat(README.md): add link to Ollama website and examples of running models


——————————————————

│
◇  Confirm the commit message?
│  No
│
◇  Do you want to regenerate the message ?
│  Yes
│
◇  📝 Commit message generated
│
└  Generated commit message:
——————————————————
feat(README.md): add link to Ollama website and examples of model usage in README.md 

Error code 127

Error: llama runner process has terminated: exit status 127

>>> [GIN] 2024/08/28 - 18:43:24 | 200 |      68.455µs |       127.0.0.1 | HEAD     "/"
>>> [GIN] 2024/08/28 - 18:43:24 | 200 |    7.845273ms |       127.0.0.1 | POST     "/api/show"
>>> time=2024-08-28T18:43:24.392Z level=INFO source=memory.go:309 msg="offload to cpu" layers.requested=-1 layers.model=33 layers.offload=0 layers.split="" memory.available="[7.3 GiB]" memory.required.full="5.5 GiB" memory.required.partial="0 B" memory.required.kv="1.0 GiB" memory.required.allocations="[5.5 GiB]" memory.weights.total="4.7 GiB" memory.weights.repeating="4.6 GiB" memory.weights.nonrepeating="105.0 MiB" memory.graph.full="560.0 MiB" memory.graph.partial="585.0 MiB"
>>> time=2024-08-28T18:43:24.392Z level=INFO source=server.go:391 msg="starting llama server" cmd="/tmp/ollama731275887/runners/cpu/ollama_llama_server --model /root/.ollama/models/blobs/sha256-ff82381e2bea77d91c1b824c7afb83f6fb73e9f7de9dda631bcdbca564aa5435 --ctx-size 8192 --batch-size 512 --embedding --log-disable --no-mmap --parallel 4 --port 33357"
>>> time=2024-08-28T18:43:24.392Z level=INFO source=sched.go:450 msg="loaded runners" count=1
>>> time=2024-08-28T18:43:24.393Z level=INFO source=server.go:591 msg="waiting for llama runner to start responding"
>>> time=2024-08-28T18:43:24.393Z level=INFO source=server.go:625 msg="waiting for server to become available" status="llm server error"
>>> /tmp/ollama731275887/runners/cpu/ollama_llama_server: error while loading shared libraries: libllama.so: cannot open shared object file: No such file or directory
>>> time=2024-08-28T18:43:24.644Z level=ERROR source=sched.go:456 msg="error loading llama server" error="llama runner process has terminated: exit status 127"
[GIN] 2024/08/28 - 18:43:24 | 500 |  266.021797ms |       127.0.0.1 | POST     "/api/chat"

The error code occurs when the Docker image version is greater than 0.3.6. Therefore, you need to pull the ollama image with version 0.3.6 and run the container. Click here to view the discussion.

🔁 Recap

• Ollama allows for the generation of commit messages using AI models.
• The article details setting up Ollama in a Docker environment.
• OpenCommit is integrated to simplify the process of generating commit messages using an AI model.
• Users can interact with the AI model through a chat interface.

🔗 References

• https://ollama.com/models
• https://github.com/ollama/ollama/issues/6541