Pywayne Chat Ollama Gradio

This module provides a Gradio-based web chat interface for Ollama models with multi-session support.

Quick Start

from pywayne.llm.chat_ollama_gradio import OllamaChatGradio

# Create and launch chat interface
app = OllamaChatGradio(
    base_url="http://localhost:11434/v1",
    server_port=7870
)
app.launch()

Configuration

Parameter	Default	Description
base_url	"http://localhost:11434/v1"	Ollama API base URL
server_name	"0.0.0.0"	Server host name
server_port	7870	Server port
root_path	""	Root path for reverse proxy
api_key	"ollama"	API key (for Ollama compatibility)

Model Discovery

Automatically discovers available Ollama models by running ollama list:

• Excludes models with 'embed' in the name
• Falls back to qwen2.5:0.5b if no models found

Session Management

Creating New Sessions

# UI method: Click "新建会话" button
new_chat_id, new_history, new_choices = app.create_new_chat()

Switching Sessions

# UI method: Select from "历史会话" radio list
history = app.switch_chat(selected_chat_id)

API Reference

OllamaChatGradio

Method	Description
get_ollama_models()	Get list of available Ollama models
init_chat_manager()	Initialize ChatManager instance
create_new_chat()	Create new chat session, returns (chat_id, history, radio_update)
switch_chat(chat_id)	Switch to specified chat session
format_history(history)	Format history for display
chat(message, history, model_name)	Process chat message with streaming
create_demo()	Create Gradio interface
launch()	Launch Gradio server

UI Components

Component	Description
chatbot	Main chat display area
msg	Message input textbox
model_dropdown	Model selection dropdown
chat_id_text	Current session ID (read-only)
new_chat_btn	Button to create new session
chat_history_list	Radio list for session switching

Requirements

• gradio - Web UI framework
• pywayne.llm.chat_bot - ChatManager and LLMConfig
• ollama CLI - Must be installed and accessible

Notes

• Uses ChatManager for multi-session support
• Streaming responses update UI in real-time
• Session history persists in memory (not persistent)
• Requires Ollama to be running before launching