org.slashlib.py.inference.ollama 0.1.4

python inference handler for agents

Bottom AI CHANGELOG LICENSE

org.slashlib.py.inference.ollama

The official Ollama inference adapter for the org.slashlib.py.agent framework.

---

1. Overview & Architecture

This project provides a specialized implementation of the InferenceAdapter interface from the core Agent Framework. By decoupling the Ollama logic from the main framework, we ensure a lightweight core and allow for independent updates to the inference logic.

Architectural Role

• Adapter Pattern: Bridges the standardized framework calls to the Ollama-specific API.
• Provider Agnostic: The framework consumes this adapter through a unified interface.
• Plugin-Based: Leverages Python entry points for seamless, zero-config integration.

---

2. Prerequisites

• Ollama Service: Must be installed and reachable (Default: http://localhost:11434).
• Python: Version 3.10 or higher.
• Base Framework: org.slashlib.py.agent must be installed in the same environment.

---

3. Installation

Production / Standard

pip install org.slashlib.py.inference.ollama

Development (Editable Mode)

If you are developing both the framework and this adapter, install both in editable mode to ensure the Entry Points are correctly registered in your current Python environment:

# 1. Install the framework
cd path/to/org.slashlib.py.agent
pip install -e .

# 2. Install this adapter
cd path/to/org.slashlib.py.inference.ollama
pip install -e .

---

4. Integration Logic (Entry Points)

This adapter is designed to be "invisible" to the end user. It registers itself via the pyproject.toml entry points:

[project.entry-points."org.slashlib.py.agent.inference"]
ollama = "org.slashlib.py.inference.ollama.adapter:OllamaInferenceAdapter"

The framework automatically scans this group and maps the name ollama to the OllamaInferenceAdapter class.

---

5. Configuration

The adapter's behavior is controlled via the framework's configuration loader (usually pyproject.json).

Key	Type	Default	Description
model	string	(required)	The name of the Ollama model (e.g., "llama3", "gemma").
think	boolean	true	Enables/Disables thinking process visibility if supported.
timeout	float	600.0	Connection timeout in seconds.

Example pyproject.json:

{
  "inference": {
    "ollama": {
      "model": "gemma4",
      "think": true,
      "timeout": 300.0
    }
  }
}

---

6. Usage Example

You do not need to import any classes from this package directly. Use the Framework's factory method:

import asyncio
from org.slashlib.py.agent.agent import Agent

async def main():
    # Load the agent - the framework resolves 'ollama' via Entry Points
    my_agent = Agent.from_plugin(
        identifier="my-local-assistant",
        plugin_name="ollama"
    )

    # All standard Agent methods are now available
    response = await my_agent.chat("How does the plugin system work?")
    print(f"Assistant: {response.content}")

if __name__ == "__main__":
    asyncio.run(main())

---

7. Development & Testing

Testing with Pytest

pytest tests/

Logging

The adapter logs under the namespace org.slashlib.py.inference.ollama.

---

8. Documentation & Obsidian

The project root is fully prepared as an Obsidian Vault.

---

License

Distributed under the MIT License. See LICENSE.md for more information.

---

© 2026 org.slashlib