AI Chat Tools and Smart Search

Behind every answer, the GrndCtrl assistant can reach for 24 specialized tools — searching your documents and the web, querying connected systems, generating files, running code. This article explains what they do and how to read what the assistant is doing.

How many you can use depends on your organization

The tool set is determined by which apps and integrations an administrator has enabled, so a capability described here may simply not be turned on for you.

How AI Tools Work

When you ask a question, the AI decides which tools to use automatically. You’ll see a collapsible tool indicator appear inline in the response showing what’s happening in real time:

• Blue spinner — tool is running
• Green checkmark — tool completed successfully
• Red X — tool encountered an error

The indicator header shows a count and a breakdown — for example “3 tools (1 running, 2 completed)”. The breakdown lists only the states that apply, so you may also see failed. Click the chevron to expand and see each tool’s name, search query (if applicable), and execution time.

Smart Search

Smart Search is a meta-tool that orchestrates multiple search strategies in a single request. Instead of running one search at a time, it combines results from several tools simultaneously — for example, searching your company documents AND the web in parallel — then returns the most relevant combined results to the AI.

When Smart Search runs, the tool indicator shows the parent smart-search tool with sub-tools listed beneath it (e.g., web-search + query-docs), each with their own status indicator.

Available Tools

You never call these yourself — the assistant picks them based on what you asked. Knowing what exists is still useful, because it tells you what is worth asking for.

Not everyone sees every tool

The available set depends on which apps and integrations your organization has enabled. A tenant with nothing extra connected sees roughly half this list; a fully connected one sees nearly all of it. If a capability below does not work for you, it is most likely not enabled rather than broken.

Finding Things

Tool	What it does
search-tools	Finds the other tools by capability. The assistant calls this when it needs something it does not currently have — and it is always available, even when the rest of the toolset is narrowed for your organization.
query-docs	Semantic search across document content by topic or meaning. This is what runs when you ask about something rather than for a specific file.
query-project	Semantic search scoped to one project’s documents — plans, specs, RFIs, submittals, and contracts.
read-document	Reads, searches within, or links to one specific file: a chat attachment, a My Docs file, a generated PDF or report, or an image.
web-search	Searches the web for current information, choosing a search depth that matches the question.
deep-research	Runs an extended, multi-step research pass when a single search will not answer the question.
smart-search	Runs several tools at once and processes their results with code, so only the processed answer enters the conversation. See below.
retrieve-context-details	Pulls back the full content of an earlier tool result, attachment, or image by its ID, rather than repeating the work.

search-tools is the one to know about

It is how the assistant discovers capabilities it does not already have loaded. If you ask for something and are told it is not possible, asking again more specifically — naming the format or the system — often prompts it to go looking.

Your Organization’s Data

Tool	What it does
dynamo-query	Reads the data your organization has connected to chat — records, metrics, and cross-system tables an administrator has opened up in GrndCtrl Access.
nspec-query	Queries the NSPEC inspection database for field reports, observations, action items, and project data.
nspec-analytics	Queries NSPEC analytics — inspections, action items, feedback, and usage. Refreshed nightly.
nspec-admin-analytics	The administrator view of the same NSPEC analytics.
nspec-generate-report	Builds a PDF QA Field Report from NSPEC inspection data, including observations, action items, site maps, and progress photos.
beci-directory	Queries the BECI employee directory, synced from Microsoft Azure AD.
trainual-search	Searches the BECI Trainual knowledge base for policies, procedures, standards, and training content.

dynamo-query is governed by GrndCtrl Access

It can only read tables an administrator has explicitly listed on the GrndCtrl Access tab of Tenant Administration, and only the columns allowed there. If the assistant says it cannot see data you know exists, that is the place to check.

Producing Things

Tool	What it does
generate-pdf	Creates a new PDF, or refines one it made earlier.
render-pdf	Renders HTML you or the assistant already wrote straight to PDF.
generate-image	Generates a new image, or edits an existing one.
execute-code	Runs Python or JavaScript in a sandbox. Sessions persist for the length of the conversation, so later steps can build on earlier ones.
send-email	Sends an email on your behalf. Needs a recipient, a subject, and a body.
superclear-analyze	Analyzes a building envelope photo for defects using SuperClear AI.

Memory

Tool	What it does
create-memory	Saves something about you for future conversations — triggered when you say remember this, or when the assistant learns something worth keeping.
read-memory	Reads back a saved memory in full.
retrieve-memory	Semantic search across your conversation history. This is what answers “what did we decide about X last month?”

File Outputs

When the AI generates a file (PDF, image, spreadsheet, etc.), it appears as a file card in the chat with two buttons:

• View — opens the file in a new browser tab
• Download — saves the file to your device

The card shows the filename, file type, and size. Supported output formats include PDF, PNG, JPG, HTML, JSON, CSV, Markdown, XLSX, DOCX, and ZIP archives.

Extended Thinking

Extended Thinking lets the AI show its reasoning process before responding. When enabled, a collapsible “Thinking…” panel appears above the AI’s response showing the step-by-step reasoning.

How to Enable

1. Look for the lightbulb icon in the message input area.
2. Click it to toggle Extended Thinking on or off.
3. When ON, the tooltip reads: “Extended Thinking ON — Brix shows its reasoning process before responding. Uses more tokens but produces deeper analysis.”.

When to use Extended Thinking:

Turn it on for complex analysis, multi-step problems, or when you want to understand how the AI arrived at its answer. Turn it off for quick questions — it uses more tokens and takes longer to respond.

Conversation Search

Search across your entire conversation history to find past discussions:

1. In the conversations sidebar, type in the “Search conversations…” field.
2. Results appear after 2+ characters with a 400ms debounce.
3. Each result shows: conversation title, a relevant snippet, relative time, and relevance score.
4. Click a result to open that conversation.

Managing Conversations

The conversations sidebar lets you organize your chat history:

• New Chat — starts a fresh conversation
• Pin — pin important conversations to the top
• Edit Title — rename a conversation inline
• Archive — move to a collapsed “Archived” section
• Download — export the conversation as a Markdown file

Conversations are grouped by time: Today, Yesterday, Past 7 Days, and Long Ago… A Pinned group sits above all of them, and Archived collapses at the bottom with a count.

Customizing Your AI

Click the gear icon in the chat header to open Customize Responses. Eleven settings control which model answers you, how long and how formal its answers are, what it calls you, and your timezone. Click Apply to save.

Every setting is described in Customizing AI Responses.

Citations

When the AI uses information from your documents, it adds inline citation numbers (e.g., [1], [2]) in the response. Click a citation to see the source document, a relevant snippet, and a relevance score. You can open the full source document from the citation preview.

Sending Attachments

You can attach files to your messages:

• Images — upload or drag-and-drop PNG, JPEG, GIF, or WebP files
• Documents — upload PDF, DOCX, XLSX, CSV, and other file types

Uploaded images show as thumbnail previews with a processing status indicator. Documents show as filename cards with upload progress.

Context Limit

Very long conversations may reach the AI’s context limit. When this happens, a yellow banner appears: “Your conversation has reached its context limit. Please start a new conversation to continue.” The message input is disabled until you start a new chat.

Giving Feedback on Responses

On each AI response, you’ll see:

• Copy button — copies the response to your clipboard
• Thumbs up / Thumbs down — rate the response quality (helps improve the AI)

Troubleshooting

Tools are taking a long time

Solutions:

• Complex queries that use multiple tools (especially Smart Search + code execution) naturally take longer
• If a tool shows an error (red X), try rephrasing your question
• Lower the Effort Level in preferences for faster responses

Smart Search returns no relevant results

Solutions:

• Try being more specific in your question — include keywords from the document you’re looking for
• Check that the relevant documents have been uploaded to the Document Library
• Ask the admin to verify document indexing is complete

Extended Thinking makes responses very slow

Solutions:

• This is expected — Extended Thinking uses significantly more tokens and processing time
• Turn it off for quick questions; use it only for complex analysis

Related Articles

• Documents and AI Chat
• Prompt Management
• Managing Your Document Library — what the assistant can read, and which documents it favours
• Conversation Analytics — how the assistant is being used across your organization
• Admin Chat and Message Analysis — what the model saw and did on a single message
• Getting Started with GrndCtrl