RSSFlow Reader Help Center

Configure AI in Settings → AI Settings → AI Summary Function Enable. After turning on the switch, configure the corresponding settings in model configuration. Currently supports the following four types:

1. OpenAI Compatible Format: Supports OpenAI and LLMs that support OpenAI format. Fill in the corresponding key, api host, and model name. If you use a proxy (such as gemini to OpenAI), also choose this type.
   If you don't have an account, visit the OpenAI official website https://openai.com/ to register. After successful registration, log in to OpenAI, go to the API keys management page, and click "Create new secret key" to create a new API Key.
   In addition, similar models like DeepSeek and kimi that support OpenAI format can be configured in this option.
2. DeepSeek: Log in to the DeepSeek open platform and create a key. If you don't have an account, register one on the official website: https://platform.deepseek.com/
3. Google Gemini: Recommended choice. For the basic model for generating summaries, choose gemini-1.5-flash-8b or above to meet the requirements, then set advanced LLM in custom settings for the model in chat mode.
   If you don't have an account, go directly to Google's official website to register. Address: https://aistudio.google.com/app/apikey After registration, create an API key.
4. SiliconFlow: Directly fill in the api key. If the dropdown options don't have the model you're using, manually input the model name. If you don't have an account, register directly. Registration address: https://account.siliconflow.cn/

Chat mode and podcast features support a 5-day trial period. Activation codes are free to obtain.

After successful AI configuration, as long as AI summaries have data, they will be displayed in DayFlow or TagFlow views. There are currently two ways to trigger AI summary generation:

1. You can trigger AI summary generation by manually adding to Marked
2. You can enable auto-summary function in subscription sources. Whenever there's new content in the subscription source, it will trigger automatic AI summary generation

Another situation: In general settings, if the query date range is set too short, historical AI summary data beyond this date range will not be displayed in the Flow tag view.

Chat mode and voice podcast are both based on AI and AI summary data. When you successfully configure AI and have successfully generated AI summaries, you will see the corresponding chat and podcast entries in the Flow view.

Mark as read method: Double-click the left mouse button to mark as read. The core reading operation of the entire extension is double-clicking the left mouse button.

Double-click to mark as read works in both title mode and full-text reading (immersive reading).

About read content: Marked-as-read content cannot be found again. Currently only unread articles are displayed. If an article is great, you can bookmark it. Bookmarked articles will always be there and won't be affected by RSS content retention duration and query date range restrictions.

Keyboard shortcuts only apply to full-text reading (immersive reading). Immersive reading is also a feature of RSSFlow.

Through the above shortcuts, you can complete full-text reading entirely through keyboard operations.

Only articles with more than 200 words will trigger summary generation. If articles are too short, like one-sentence news, there's no need to generate summaries.

Yes, you can customize them. In Settings → AI Settings → Prompt Settings, there's a "Tag Range" option, which corresponds to user-defined tags.

RSSFlow currently supports 12 languages and can subscribe to RSS content from any country in any language, then respond in one of these 12 languages.

If an article doesn't have clear viewpoints or conclusions, it's essentially a wishy-washy article that doesn't need detailed reading. Just looking at the Opinions can meet basic needs.

If the viewpoints are clear, you can click to view detailed content.

Yes, you can customize them, but it's not recommended.

Custom quick commands are not currently supported. If needed, you can directly input your commands in the chat window.

Yes, RSSFlow is a browser extension that depends on the browser running. You just need to keep the browser open, and it will automatically get updates and process data in the background.

RSSFlow is actually an AI Agent, but this AI Agent is fully integrated into RSS reading scenarios with high integration, making it look more like a reader.

Differences from workflows:

• Workflows require complex processes, various external content scraping service support, and complex connections
• Limited data storage, inflexible data operations, and restricted computing resources
• RSSFlow uses local storage with efficient database query performance
• As long as local storage and device performance are sufficient, it can store massive data
• Can flexibly assemble, filter, and analyze data
• Far superior to workflows in terms of stability, real-time performance, and efficiency

RSSFlow is based on RSS subscriptions. As long as your subscribed RSS sources have accumulated 1 year of data, you just need to:

1. In general settings, set RSS content retention duration to 365 days
2. Also set the query date range to 365 days
3. Then you can query 1 year of data

The method is to streamline the context through data filtering functions, based on specified tags and keywords to complete data selection and filtering.

The effectiveness of quick commands depends on the data you want to analyze and the LLM:

1. Data relevance: The data you analyze needs to match your command scenario. For example, if you execute a tech news command but your content has no tech-related content, the results will definitely be poor.
2. LLM capability: Quick commands describe some analysis and business methods, requiring strong reasoning or logical capabilities. Therefore, the stronger the LLM, the better the results.
3. Parameter adjustment: LLM parameters also affect final results. You can appropriately adjust Temperature and Top P, making some fine-tuning in content writing or areas requiring divergent thinking to achieve optimal results.

Currently supported. In chat mode, there's an AI settings function that you can configure directly.

Possible causes:

• RSS feed URL format error or invalid
• Website doesn't support RSS/Atom format
• Network connection issues
• Firewall or proxy settings blocking access

Possible causes:

• API key configuration error
• Insufficient API account balance
• Inappropriate model selection
• Network connection issues
• API service temporarily unavailable

Possible causes:

• Browser cache issues
• Extension accidentally disabled
• Browser version too old
• Conflicts with other extensions
• Extension files corrupted

Possible causes:

• AI not properly configured
• No AI summaries generated
• Activation code not entered or expired
• Functions hidden or disabled

Possible causes:

• Browser data was cleared
• Extension was reinstalled
• Browser profile corrupted
• Insufficient storage space