AskHandle

Document Search

Answer from your documents, not from the open web

Document Search lets an AskHandle agent answer questions using only the content in the documents you connect to the node.

Example: hotel information

For example, a document might say:

Guests may bring one domestic animal weighing no more than 25 pounds.

The user asks:

Are small pets allowed?
Semantic searchDocument Search

Upload the files that contain the information your customers, employees, or partners need. When a question reaches Document Search, AskHandle searches those files, finds the most relevant information, and generates an answer grounded in that content.

The node is designed for situations where accuracy, source control, and consistency matter more than open-ended general knowledge.

Add the documents the node should search

Inside the Document Search settings, customers choose which documents belong to that node.

You can upload new files or select documents already available in AskHandle.

A single node can use up to 10 documents.

This lets you keep knowledge organized by purpose.

Document Search3 / 10
Example: hotel information
PDF · DOCX · TXT · Markdown

What Document Search does

Document Search gives an AI agent a controlled knowledge source.

The customer selects the documents the node should use. AskHandle then searches those documents when a user asks a question and responds based on the information it finds.

Supported document types include:

  • PDF
  • DOCX
  • TXT
  • Markdown

Each Document Search node can use up to 10 documents.

Unlike AI Answer with live web browsing enabled, Document Search is not intended to search the public internet or answer from general public information. Its job is to stay grounded in the documents connected to the node.

That makes it useful for information such as:
  • policies
  • manuals
  • procedures
  • FAQs
  • service information
  • property information
  • employee guides
  • membership information
  • training materials
  • product documentation
  • contracts and reference documents
  • operational instructions
Customer policies · Product documentation · Employee support

For example, instead of putting every company document into one place, an agent could use separate Document Search nodes for:

Customer policies

  • returns
  • refunds
  • delivery
  • warranties

Product documentation

  • manuals
  • specifications
  • setup guides

Employee support

  • HR policies
  • onboarding
  • internal processes

When combined with Router, different requests can be sent to the most relevant document collection.

Organize documents around the job

The best document collection is not necessarily the largest one.

A focused node is easier to manage and gives the search process a clearer set of sources.

For example, a hotel agent may use one Document Search node containing:

  • property information
  • amenities
  • policies
  • arrival instructions
  • transportation information

A separate internal agent might use completely different documents for employee procedures.

Keep together the documents that serve the same conversational purpose.

When the jobs are substantially different, separate nodes can make the agent easier to understand and maintain.

Search the meaning and the exact wording

Document Search uses hybrid search, combining lexical and semantic retrieval.

That means it can look for both:

  • exact words and phrases
  • information that matches the meaning of the question

The combination improves precision without requiring the user to phrase a question exactly the same way the document is written.

Lexical search

Lexical search looks for words, phrases, names, numbers, and other direct text matches.

This is useful when the question contains specific terminology.

For example, a document might contain:

Late checkout is available until 1:00 PM for an additional fee.

A guest asks:

What is the late checkout time?

Lexical search can recognize the direct relationship between terms such as "late checkout" and the relevant passage.

It is especially useful for:

  • names
  • policy terms
  • product codes
  • exact phrases
  • numbers
  • dates
  • technical terminology

Semantic search

Semantic search looks at meaning rather than depending only on matching words.

The wording is different, but the meaning is closely related.

Semantic retrieval helps the agent find the relevant passage even when the user's language does not exactly match the source document.

Why both matter

Business documents rarely use the exact same language customers use in conversation.

Hybrid search allows Document Search to handle both cases:

Exact request

What is the cancellation fee?

and:

Natural conversational request

What happens if I need to cancel?

Both can lead to the same relevant information when the source material supports the answer.

This helps the conversation feel natural while keeping the answer grounded in the source documents.

The documents are the source of truth

Document Search is designed to answer from the connected files.

If the information is not present in those documents, the node should not invent an answer from unrelated general knowledge.

This makes Document Search useful when a business wants tighter control over what the AI can say.

The principle is the same: the connected documents define the available knowledge.

For example, a hotel may want answers about:

  • check-in and check-out
  • parking
  • pet policies
  • breakfast
  • amenities
  • cancellation terms

to come from the hotel's own information rather than from public websites or general assumptions.

An internal support agent may need to answer from:

  • employee policies
  • IT procedures
  • onboarding guides
  • internal manuals

A membership organization may use:

  • membership rules
  • benefits
  • event procedures
  • renewal information

Suppose a hotel uploads its guest information guide, property policies, and amenities document.

A guest asks:

Can I arrive before the normal check-in time?

Document Search looks across the connected files.

It finds the section explaining early arrival and check-in policy.

The agent answers based on that information.

Then the guest asks:

Is there somewhere I can leave my bags if the room isn't ready?

Document Search searches again and returns the relevant luggage-storage information if it exists in the documents.

The guest can ask naturally. They do not need to know which file contains the answer.

Configure how answers should sound

Document Search also lets the customer control the answer tone.

The tone setting changes how the answer is presented without changing the underlying information being retrieved.

Available options include:

  • Friendly
  • Formal
  • Brief
  • Custom

Useful for normal customer-facing conversations where the answer should sound natural and approachable.

Guests may bring one domestic animal weighing no more than 25 pounds.

AI Agent · Friendly
Yes, pets are welcome. You can bring one pet weighing up to 25 pounds.

Tone and knowledge are separate

This distinction is important.

The documents control what the agent knows.

The answer tone controls how that information is communicated.

For example, the same policy could be returned differently.

The underlying source information stays the same.

Let users upload documents in Chat Portal

Document Search can also allow users to upload their own documents through Chat Portal.

When this setting is enabled, a visitor can attach a document during the conversation and the search can use that uploaded file.

This can be useful for workflows where the user wants to ask questions about a document they provide.

Examples include:

  • reviewing a reference document
  • asking questions about a manual
  • exploring a report
  • working through training material
  • searching a document supplied for the conversation

This option is intended for Chat Portal use. The product interface specifically notes that user document uploads are not recommended with the widget, API, or WhatsApp.

Use Document Search with Router

Document Search becomes especially useful when Router sends different questions to different knowledge sources.

For example:

Router

Property information
→ Document Search: Property guide

Policies
→ Document Search: Policies

Availability
→ Data Search

Book appointment
→ AI Answer with Scheduling enabled

Talk to staff
→ Human Handoff

The customer can keep each knowledge source focused while still giving users one continuous conversational experience.

When to use Document Search

Document Search is a strong choice when:

  • the answer should come from business-owned documents
  • customers ask many questions about policies or procedures
  • important information is buried in long files
  • users phrase questions differently from the source text
  • the business wants answers grounded in controlled knowledge
  • employees need conversational access to internal documentation
  • the agent should search several related documents at once

It is particularly useful when the source material already exists and the goal is to make it easier to access through conversation.

Document Search vs. AI Answer

Both nodes can answer questions, but they are designed for different jobs.

Use Document Search when

The answer should come from a controlled set of documents.

Examples:

  • policies
  • manuals
  • company guides
  • property information
  • internal knowledge
  • detailed reference material

Document Search searches the connected files and grounds its response in what it finds.

Use AI Answer when

The agent needs broader conversational intelligence, general information, a large directly supplied briefing, live web browsing, or action-oriented skills such as scheduling and email.

A simple distinction is:

Document Search
Search these documents and answer from them.

AI Answer
Use the configured intelligence, knowledge, web access, and skills to handle the conversation.

The two nodes can also work in the same agent when different requests require different behavior.

AI Answer →
Document Search vs. Data Search

Document Search is for unstructured written information.

Data Search is for structured records.

Use Document Search for content such as:

  • PDFs
  • DOCX files
  • TXT files
  • Markdown
  • manuals
  • policies
  • guides

Use Data Search when users need to retrieve exact records from structured datasets such as CSV or Excel files.

For example:

What is your cancellation policy?

is a natural Document Search question.

But:

Show me every vehicle under $40,000 with fewer than 30,000 miles.

is better suited to Data Search because the answer depends on filtering structured records.

Data Search →
When another node may be better

Use AI Answer when the agent needs broad conversational reasoning, live public information, or action skills.

Use Data Search when the answer depends on exact structured records.

Use AI Form when information needs to be collected in a controlled sequence.

Use Router when different inquiries should be sent to different workflows or knowledge sources.

Use Human Handoff when the conversation needs to move to a person.

Document Search can be combined with all of them.