ollama-coder 0.2.0

Agentic coding assistant for Ollama - Claude Code alternative

OllamaCoder 🦙💻

An agentic coding assistant for Ollama - like Claude Code, but running locally!

OllamaCoder transforms your local Ollama models into a powerful autonomous coding agent with tool use, multi-step task execution, and project-aware context.

✨ Features

• 🤖 Agentic Architecture: Autonomous multi-step task execution with planning, execution, and verification
• 🔧 Full Tool System: bash, file operations, git, code search, web search, and more
• 💭 Think Tool: Structured reasoning for complex problems (like extended thinking)
• 📝 Multi-Edit: Batch file edits in a single operation
• 🔄 Autonomous Mode: Let the AI work through complex tasks independently
• ⚙️ Hierarchical Config: User and project-level settings with OLLAMA.md context files
• 🌐 Remote Ollama: Connect to remote Ollama servers via OLLAMA_HOST
• 💾 Persistent History: Conversation history saved across sessions
• 🎨 Rich Output: Beautiful terminal output with syntax highlighting (optional)

📦 Install

From PyPI:

```bash pip install ollama-coder ```

For enhanced terminal output (recommended):

```bash pip install ollama-coder rich ```

For local development:

```bash git clone https://github.com/lalomorales22/ollama-coder.git cd ollama-coder pip install -e . ```

🚀 Quick Start

```bash

Start interactive mode

ollama-coder

Start with autonomous mode enabled

ollama-coder --auto

Run a single command (headless mode)

ollama-coder -p "fix the bug in app.py"

Use a specific model

ollama-coder --model codellama:13b

Work in a specific directory

ollama-coder --dir /path/to/project ```

🛠️ Available Tools

Tool	Description
`think`	Structured reasoning for complex problems
`bash`	Execute shell commands
`read_file`	Read file contents with optional line ranges
`write_file`	Create or overwrite files
`edit_file`	Make surgical edits using find/replace
`multi_edit`	Batch multiple edits in one operation
`list_directory`	Explore project structure
`search_code`	Search for patterns using grep/ripgrep
`git`	Version control operations
`web_search`	Search the web (when configured)

⌨️ Commands

Command	Description
`/auto`	Toggle autonomous mode
`/model`	Show or set the active model
`/models`	List installed Ollama models
`/host`	Show or set the Ollama host
`/config`	Show current configuration
`/clear`	Clear conversation history
`/help`	Show available commands
`/quit`	Exit OllamaCoder

⚙️ Configuration

User config: `~/.ollamacode/settings.json`

```json { "model": "llama3.3:latest", "max_iterations": 25, "max_tool_rounds": 8, "temperature": 0.7, "ollama": { "host": "http://127.0.0.1:11434", "timeout_sec": 60 }, "web_search": { "enabled": false, "provider": "custom", "endpoint": "", "api_key": "" } } ```

Project Context

Create `OLLAMA.md` files to provide project-specific context:

• `~/.ollamacode/OLLAMA.md` - User-level context (applies to all projects)
• `.ollamacode/OLLAMA.md` - Project-level context (in your project root)

🔌 Remote Ollama

Connect to remote Ollama servers:

```bash

Via environment variable

export OLLAMA_HOST=http://your-server:11434 ollama-coder

Or use the /host command

ollama-coder

/host http://your-server:11434 ```

📋 Requirements

• Python 3.9+
• Ollama server running locally or accessible remotely
• Optional: `rich` package for enhanced terminal output

🔄 Comparison with Claude Code

Feature	OllamaCoder	Claude Code
Local/Private	✅	❌
Free	✅	❌
Tool Use	✅	✅
Autonomous Mode	✅	✅
Thinking Tool	✅	✅
Multi-Edit	✅	✅
Web Search	✅	✅
Project Context	✅	✅
Image Analysis	❌	✅
MCP Support	🔜	✅

📝 License

MIT

🔗 Links

• GitHub: https://github.com/lalomorales22/ollama-coder
• PyPI: https://pypi.org/project/ollama-coder/
• Ollama: https://ollama.ai/

📤 Publishing

Version bumping and publishing is automated via GitHub Actions. Just:

1. Bump version in `pyproject.toml` and `ollama_coder/init.py`
2. Commit and push
3. Create and push a tag: ```bash git tag v0.1.3 git push origin v0.1.3 ```