Overview
With the AI search mode feature (RAG chat) in Fess, you can look for information through natural conversation, in addition to traditional keyword search. When you enter a question, the AI assistant searches for relevant documents and generates an easy-to-understand answer based on their content.
Note
AI search mode is disabled by default. To use it, a system administrator must enable the feature and configure an LLM (Large Language Model) provider. For configuration instructions, see AI Mode Configuration and AI Search (RAG) and LLM Integration Overview.
Features of AI Search Mode
Conversational Search
Instead of coming up with keywords, you can ask questions just as you would in everyday conversation.
For example:
“Tell me how to install Fess”
“How do I crawl files?”
“What should I do if no search results appear?”
Context-Aware Answers
The AI assistant analyzes the intent of your question, searches for relevant documents, and extracts the necessary information from multiple search results to provide an organized answer.
Clearly Cited Sources
At the bottom of an answer, the documents that the answer was based on are listed as “Sources”. Each source is a numbered link, and clicking it lets you view the original document directly. The answer body may also include citation numbers such as [1] and [2], which correspond to the numbers in the sources list.
Multi-Turn Conversations
You are not limited to a single question; you can keep the conversation going. The AI assistant answers additional questions while taking the context of previous questions and answers into account.
How to Use Chat Search
Opening AI Search Mode
Open the Fess search screen
Click the “AI Search” link (robot icon) in the navigation bar at the top of the screen
The AI search mode screen is displayed
Note
The “AI Search” link is displayed only when AI search mode is enabled and an LLM provider is available. If the link does not appear, see Frequently Asked Questions.
Entering a Question
Enter your question in the text box at the bottom of the screen (up to 4000 characters per question)
Click the “Send message” button (paper airplane icon), or press Enter
The AI assistant starts generating an answer
Note
To insert a line break, press Shift+Enter instead of Enter, as indicated by the hint below the text box: “Press Enter to send, Shift+Enter for new line”.
Note
Generating an answer can take anywhere from a few seconds to over ten seconds. While processing, the current phase is shown in a progress indicator (Analyze → Search → Evaluate → Retrieve → Answer), along with messages such as “Thinking…”, “Searching…”, “Reviewing search results…”, “Retrieving documents…”, and “Generating answer…”. Once the search is complete, the number of documents found is also displayed.
Reviewing the Answer
The AI assistant’s answer is displayed. The answer includes the following information:
Answer body: The answer to your question (formatted in Markdown)
Sources: A numbered list of links to the documents the answer is based on (clicking a link opens the original document in a new tab)
The answer body may include citation numbers such as [1] and [2], which correspond to the numbers in the sources list. Each answer has a copy button that lets you copy its content to the clipboard.
Note
The AI uses only the top documents it considers most relevant among the search results as the basis for its answer. As a result, the number of sources may be smaller than the number of documents that matched the search.
Narrowing Down Search Targets
Depending on the topic, you can use the “Filter” button at the top of the screen to narrow down the search target by criteria such as label, file type, last modified date, and size. This is useful when you want to ask the AI about a specific set of documents only.
Continuing the Conversation
If you have additional questions, you can simply keep asking:
“Could you explain in more detail?”
“Are there other ways to do this?”
“Tell me more about XX”
The AI assistant answers while taking the context of the previous conversation into account.
Starting a New Conversation
If you want to ask about a different topic, click the “New Chat” button (plus icon). This clears the previous conversation history so you can start a new conversation.
Tips for Effective Questions
Be Specific
Specific questions get more appropriate answers than vague ones.
| Vague Question | Specific Question |
|---|---|
| How do I configure this? | How do I change the memory settings in Fess? |
| I’m getting an error | I get an “index not found” error when searching |
| About crawling | How do I configure exclusions when crawling a website? |
Include Background Information
Including your situation and purpose leads to more appropriate answers.
Good examples:
“I’m running Fess in a Docker environment. I want to change where the logs are stored — how do I do that?”
“This is my first time using Fess. What should I do first?”
Ask Questions Step by Step
Complex issues are easier to understand when you ask about them step by step.
“Can Fess crawl file shares?”
“How do I connect using the SMB protocol?”
“What should I do for shared folders that require authentication?”
Frequently Asked Questions
Q: AI Search Mode Doesn’t Appear
A: AI search mode may not be enabled. The “AI Search” link is displayed only when the feature is enabled (rag.chat.enabled=true) and an LLM provider (such as OpenAI, Gemini, or Ollama) is configured and available. Ask your system administrator to check whether AI search mode is enabled and whether the LLM provider is configured correctly (see AI Mode Configuration for details).
Q: It Takes a Long Time for the Answer to Appear
A: Since the AI analyzes your question, searches for and evaluates documents, and then generates an answer, this can take anywhere from a few seconds to over ten seconds. While processing, the current phase is displayed (“Thinking…”, “Searching…”, “Reviewing search results…”, “Retrieving documents…”, “Generating answer…”).
Q: The Sources Seem to Be Incorrect
A: Click the source links to check the original documents. The AI generates answers based on search results, but its interpretation may occasionally be incorrect. We recommend always verifying important information against the original documents.
Q: It Seems to Have Forgotten the Previous Conversation
A: The conversation session may have timed out. If there is no activity for a certain period (30 minutes by default), the conversation history is cleared. In addition, since the conversation history is temporarily kept in the server’s memory, it is also lost if the server restarts. If this happens, please start a new conversation.
Note
The “session” mentioned here refers to the AI search mode’s conversation session, which is separate from your Fess login session.
Q: I Can’t Get an Answer to a Specific Question
A: The following are possible causes:
There is no relevant information in the documents being searched
The question is too vague for the search to work properly
A rate limit has been reached (the maximum number of requests per minute, or the maximum number of concurrent requests, has been exceeded)
Try rephrasing your question, or wait a while and try again.
Q: Is There a Limit on the Number of Characters I Can Enter?
A: You can enter up to 4000 characters per question. A character counter is displayed below the text box, and it switches to a warning display as you approach the limit. If your question is long, shorten it by focusing on the key points.
Q: Can I Ask Questions in Languages Other Than English?
A: Yes, in most cases you can also ask questions in other languages. Based on your browser’s or the screen’s display language, the AI tries to respond in the same language whenever possible. However, this is a best-effort behavior, and in some cases the response may not be in the language you intended.
Notes
About AI Responses
AI responses are generated based on search results
The accuracy of responses is not guaranteed
For important decisions, always check the original documents (sources)
For the latest information, refer to the official documentation
About Privacy
The questions you enter are used for search and for AI processing by the configured LLM provider
If the system is configured to use an external LLM service (such as OpenAI or Gemini), the content of your questions and the search results are sent to that service. If you want processing to stay entirely in-house, ask your administrator about using a locally running provider (such as Ollama)
Conversation history is temporarily kept in the server’s memory and is deleted when the session times out, when you click “New Chat”, or when the server restarts
As with regular search, role (permission) and label-based access control applies, so documents you are not permitted to view are not used as a basis for answers
Depending on the system configuration, logs may be recorded
References
AI Mode Configuration - Configuring the AI search mode feature (for administrators)
AI Search (RAG) and LLM Integration Overview - Configuring LLM providers
Chat API - Chat API (for programmatic use)
AND Search - How to Use AND Search
NOT Search - How to Use NOT Search
Field-Specified Search - How to Use Field Search
Advanced Search - Advanced Search Features