miniflux-ai — AI 翻译工具中文文档

Requirements

• Python 3.11+
• Dependencies: Install via pip install -r requirements.txt
• Miniflux API Key
• API Key compatible with OpenAI-compatible LLMs (e.g., Ollama for LLaMA 3.1)

Docker Setup

The project includes a docker-compose.yml file for easy deployment:

If using webhook or AI news, it is recommended to use the same docker-compose.yml with miniflux and access it via container name.

services:
    miniflux_ai:
        container_name: miniflux_ai
        image: ghcr.io/qetesh/miniflux-ai:latest
        restart: always
        environment:
            TZ: Asia/Shanghai
        volumes:
            - ./config.yml:/app/config.yml
            # - ./entries.json:/app/entries.json # Provide persistent for AI news

docker-compose up -d

Usage

1. Ensure config.yml is properly configured.
2. Run the script: python main.py
3. The script will fetch unread RSS entries, process them with the LLM, and update the content in Miniflux.

Configuration

The repository includes a template configuration file: config.sample.yml. Modify the config.yml to set up:

If using a webhook, enter the URL in Settings > Integrations > Webhook > Webhook URL. If deploying in a container alongside Miniflux, use the following URL: http://miniflux_ai/api/miniflux-ai.

• Miniflux: Base URL and API key.
• LLM: Model settings, API key, and endpoint.Add timeout, max_workers parameters due to multithreading
• AI News: Schedule and prompts for daily news generation
• Agents: Define each agent's prompt, allow_list/deny_list filters, and output style（style_block parameter controls whether the output is formatted as a code block in Markdown）.

FAQ

<details> <summary> If the formatting of summary content is incorrect, add the following code in Settings > Custom CSS: </summary>

pre code {
    white-space: pre-wrap;
    word-wrap: break-word;
}

<details> <summary> fetcher: refusing to access private network host "xx.xx.xx.xx" </summary>

Starting with Miniflux 2.2.18, FETCHER_ALLOW_PRIVATE_NETWORKS=1 must now be enabled to access feeds hosted on a local network. See the official release notes for details.

</details>

<details> <summary> level=WARN msg="Unable to send new entries to Webhook" ... client: connection to private network is blocked: host "miniflux-ai" resolves to a non-public IP address" </summary> Starting with Miniflux 2.2.18, INTEGRATION_ALLOW_PRIVATE_NETWORKS=1 must now be enabled to access third-party integration services hosted on a local network. See the official release notes for details.

</details>