AI inside your spreadsheet sounds futuristic — until you actually try it. With DeepSeek R1 integrated as an AI field shortcut in Lark Base, you can ask the model to summarise customer feedback, classify support tickets, generate product descriptions, or even reason through complex data — all without leaving your base.

DeepSeek R1 is one of the most talked-about AI models in 2026, known for its strong reasoning and chain-of-thought capabilities. Combined with Lark Base’s powerful data management and automation, the result is a no-code AI workflow that small teams can stand up in under an hour.

This guide walks you through the entire process — from signing up for the BytePlus account that hosts DeepSeek, to making your first AI-powered field inside Lark Base.

What You’ll Need Before You Start

Before diving in, make sure you have:

• A Lark account with access to Lark Base (desktop or web version)
• A working email address (for BytePlus signup)
• A phone number (for verification)
• About 30–45 minutes for the full setup

You do not need a credit card to get started — BytePlus allows you to skip the billing step initially and use the free trial tokens.

What Is DeepSeek R1 and Why Use It in Lark Base?

DeepSeek R1 is an advanced AI model developed by DeepSeek, recognised for its strong reasoning ability and its capacity to “think through” multi-step problems before answering. It is particularly good at:

• Analytical and logical reasoning
• Summarising complex content
• Categorising and classifying data
• Generating structured responses
• Multilingual tasks (including English and Mandarin)

When you connect DeepSeek R1 to Lark Base via BytePlus (the platform that hosts DeepSeek and other models for enterprise use), you gain a powerful AI field type that can process every row of your base automatically. Use cases include:

• Customer feedback analysis — auto-tag sentiment or themes
• Support ticket triage — classify by urgency or category
• Lead qualification — score leads based on their description
• Content generation — draft product descriptions, email replies, or summaries
• Data enrichment — extract structured information from unstructured text

Now let’s set it up.

Step 1: Sign Up for a BytePlus Account

DeepSeek R1 is accessed through BytePlus, the enterprise AI platform that hosts the model. Start here:

1.1 Go to the BytePlus signup page: www.byteplus.com

1.2 Fill in your email and create an account.

1.3 Complete Step 1 of 2 — Add account profile. Provide your:

• Organisation name (e.g. your business name)
• Full name
• Country/Region (select Malaysia)
• Address details (postal code, city, state)
• Phone number (you will receive a verification code via SMS)

1.4 Verify your phone number using the code sent to your mobile, then click Continue.

1.5 For Step 2 of 2 — Billing information, you have two options:

• Add a credit card now to unlock full credit limits
• Click Complete later to skip billing and start with free trial tokens

Tip: For first-time users, click Complete later. The free trial tokens are more than enough to test DeepSeek R1 with Lark Base before deciding whether to commit to a paid plan.

Step 2: Access ModelArk and Set Up DeepSeek R1

Once your BytePlus account is ready, you’ll need to enable the ModelArk service — this is where DeepSeek R1 is hosted.

2.1 From the BytePlus console homepage, locate ModelArk in your products list and click to access it.

2.2 Inside ModelArk, navigate to Online inference to see the endpoint configuration page.

2.3 Fill in the required endpoint details:

• Endpoint name — give it a clear name like “Lark Base DeepSeek”
• Endpoint description (optional) — describe its use case
• Model selection — this is where you’ll add DeepSeek R1

2.4 Click + Add model to choose your model.

2.5 In the model selection screen, switch to the DeepSeek tab and select DeepSeek R1, then click Confirm.

Tip: BytePlus offers multiple model providers (ByteDance, DeepSeek, Moonshot AI, and more). DeepSeek R1 is the strongest pick for reasoning-heavy tasks, but feel free to experiment with others depending on your use case.

2.6 After confirming the model, you’ll see DeepSeek R1 listed under your endpoint, ready for inference.

Step 3: Generate Your API Key

To connect BytePlus to Lark Base, you’ll need an API key that authorises the connection.

3.1 In the ModelArk console, navigate to API Keys in the left sidebar.

3.2 Click Create API Key, give it a name (e.g. “Lark Base Integration”), and confirm.

3.3 Once generated, copy the API key immediately and store it somewhere safe (a password manager or Lark Doc with restricted access works well).

Important: API keys are usually displayed only once. If you lose it, you will need to generate a new one. Treat it like a password — never share it publicly or commit it to a public file.

Step 4: Add DeepSeek R1 as an AI Field in Lark Base

Now for the part that brings it all together. With your BytePlus account set up and your API key in hand, you can finally connect DeepSeek R1 to your Lark Base.

4.1 Open the Lark Base where you want to use DeepSeek R1.

4.2 Click the + icon to add a new field.

4.3 Enter a Field title (e.g. “AI Summary” or “Sentiment Analysis”).

4.4 For Field type, scroll through the options and select DeepSeek R1 under Recommended or Explore Field Shortcuts.

4.5 Lark Base will prompt you to connect your BytePlus ModelArk account. Paste in:

• The API key you generated in Step 3
• Any other required details (such as the endpoint name)

4.6 Confirm the connection. Once connected, DeepSeek R1 is now available as an AI field in your base.

4.7 Configure your prompt. Tell DeepSeek what to do with each row’s data — for example:

4.8 Save the field. DeepSeek R1 will now run automatically for every row, populating the AI field with reasoning-powered output.

Tip: Start with a small base (10–20 rows) to test your prompt and confirm the output quality before running it across hundreds of records. Iterating prompts is much faster on small datasets.

Practical Use Cases for SMEs and Startups

Here are real-world ways SMEs are already using DeepSeek R1 inside Lark Base:

Customer feedback analysis

Auto-classify customer reviews into themes (delivery, quality, pricing, support) so you can spot patterns without reading every single review.

Support ticket triage

Have DeepSeek read incoming tickets and assign priority levels, categories, or suggested response templates — saving your team hours per week.

Lead scoring

Pass lead descriptions or notes through DeepSeek and let it score each lead’s quality based on criteria you define in the prompt.

Product description generation

Feed in product specs and let DeepSeek draft polished marketing copy for your e-commerce site.

Meeting notes summarisation

Paste raw notes or transcripts into a Lark Base field and use DeepSeek to extract action items, decisions, and follow-ups.

Multilingual content support

Translate or localise content automatically — particularly useful for businesses serving Malaysia’s multilingual market (English, Bahasa Malaysia, Mandarin).

Pro Tips for Working with DeepSeek R1 in Lark Base

• Refine prompts iteratively. Your first prompt is rarely your best one. Test, observe outputs, and improve.
• Use clear input fields. DeepSeek performs best when the input data is structured and well-labelled. Messy data leads to messy output.
• Watch your token usage. Every AI call consumes tokens from your BytePlus quota. Disable the AI field on rows where it is not needed.
• Combine with Lark automations. Use DeepSeek’s output to trigger Lark workflows — for example, auto-notify the support manager when a ticket is classified as “urgent”.
• Document your setup. Keep a Lark Doc that records your endpoint name, model version, and prompt patterns. This is invaluable when onboarding new team members.

Frequently Asked Questions (FAQ)

1. How do I use DeepSeek in Lark Base?

Sign up for a BytePlus account, set up DeepSeek R1 inside ModelArk, generate an API key, then add a new field in your Lark Base and select DeepSeek R1 as the field type. Connect your BytePlus account using the API key, configure your prompt, and DeepSeek will process every row automatically.

2. Is DeepSeek R1 free to use in Lark Base?

BytePlus offers free trial tokens for new accounts, which is enough to test the integration. After the trial, usage is metered based on tokens consumed. Pricing depends on the volume of AI calls you make.

3. Do I need a credit card to set up BytePlus?

No. You can click Complete later during signup and start using free trial tokens immediately without entering payment details.

4. What can DeepSeek R1 do inside Lark Base?

DeepSeek R1 can summarise text, classify content, extract structured information from unstructured text, generate written content, and reason through multi-step problems — all applied row by row to your Base data.

5. Is my data safe when using DeepSeek through BytePlus?

BytePlus is the enterprise-grade platform that hosts DeepSeek, with its own data security and privacy policies. Always review BytePlus’s data handling terms before sending sensitive customer or business data through the API.

6. Can I use other AI models besides DeepSeek R1 in Lark Base?

Yes. BytePlus ModelArk supports multiple models including ByteDance, DeepSeek, Moonshot AI, and others. The setup process is the same — swap DeepSeek R1 for your preferred model in Step 2.

7. What if I lose my API key?

You will need to generate a new one from the API Keys page in BytePlus, then update the connection in Lark Base. The old key will stop working once revoked.

Bring AI Into Your Daily Workflows

DeepSeek R1 inside Lark Base turns your everyday spreadsheets into intelligent, self-processing data tables.

For SMEs and startups, this is the kind of automation that used to require an engineering team — and now takes less than an hour to set up.

Start with one use case (customer feedback or ticket classification are great first projects), validate the value, then expand DeepSeek across more bases as your team gets comfortable.

Combined with Lark’s automations, dashboards, and approvals, you have a no-code AI platform that genuinely scales with your business.

Ready to Power Your Business with Lark?

Lark Base is just one part of an all-in-one platform that brings messaging, meetings, documents, approvals, and AI-powered automations together for your entire team.

As the Platinum Partner for Lark in Malaysia, Exabytes offers tailored Lark plans, hands-on onboarding, and dedicated local support to help you build AI-driven workflows like this one — and many more — for your business.

Explore Lark plans with Exabytes today

When your Lark Base supports a real workflow — a sales pipeline, an inventory tracker, or a project dashboard — making changes to it can feel risky. One wrong tweak to a formula or automation could disrupt how your entire team works.

That is exactly why Lark Base offers the Sandbox feature. It gives you an isolated testing environment that mirrors your live production base, so you can safely test and validate any configuration changes before they go live. Once you have verified everything works, you simply apply the changes to production with a single click.

This guide walks you through exactly how to use the sandbox in Lark Base — from enabling it to applying validated changes — so SMEs and startups can iterate confidently without breaking what already works.

What Is the Sandbox in Lark Base?

The sandbox is a safe, isolated environment that mirrors your production base.

You can think of it as a working copy of the base that contains the same configurations, where you can experiment freely without affecting the live data your team relies on.
Inside the sandbox, you can test:

• Changes to fields and views
• Edits to automations and workflows
• Configuration changes across tables

Once your changes are verified, applying them to production takes one click.

• Platform requirement: The sandbox feature is only available on the Lark desktop app or web version. It is not available on the Lark mobile app.
• Permission requirement: Only users with manage permissions for the base can use the sandbox feature.

Step 1: Enable and Access the Sandbox

Before you start, make sure that Edit history and Time Machine for your base are turned on. These features allow you to roll back to the pre-update version if any errors occur when applying changes to production.

Once that’s confirmed, follow these steps:

1.1 Open the base, click the ⋯ (more) icon in the upper-right corner, and select Sandbox.

1.2 Choose what to copy to the sandbox. You have two options:

• Configurations only — Copies your structure, fields, formulas, and automations without any data
• Configurations and specified records — Copies the structure plus up to the first 1,000 rows of data per table

1.3 Click Create. After the sandbox is created, you will see a “Sandbox created” message. Click Go to Sandbox in the message to enter the sandbox environment.

Tip: Always glance at the green Sandbox banner before making any change. It is easy to forget which environment you are in once you start working — and the banner is your safety net.

Step 2: Test Inside the Sandbox

Once you are inside the sandbox, click the ⋯ icon in the upper-right corner of the base, then select Sandbox > Go to Sandbox to open the sandbox. From here, you can freely test:

• Tables — Add, edit, or delete fields or views in the table. Every configuration change will be recorded in the sandbox.
• Automations and workflows — Add, edit, or delete automations or workflows. Every change is recorded for later review.

Important note about test messages: When you test sending Lark messages in automations or workflows from the sandbox, the message will be sent to the actual automation or workflow configurator — not the recipient or group. When you test sending emails, the email will be sent to the recipient(s). Automation and workflow runs triggered in the sandbox will count toward the total run count.

This means you should be careful with email-based automations during testing, as test emails reach real recipients.

Tip: For email-based automations, set up a test email address (or use your own) in your sandbox before testing, so test runs do not accidentally email your actual customers.

Step 3: View Changes

To track everything you have done in the sandbox, click View Changes at the top of the sandbox to see a complete change set.

Inside the change set, you can:

• Filter changes by module, user, or time
• Hover over the ⋯ icon to the right of any change and click View for more details
• Spot dependencies between changes — when there are dependencies, hover over the View in the Linked items column to see the details

What Are Dependencies?

Some changes rely on others to function correctly. Examples:

• If you add a new field to a table and then group records by that field in a view, the changes in the view depend on the new field.
• If you add a new field to a table and then reference a value from that field in a workflow, the changes to the workflow depend on the new field.

Changes with dependencies must be applied together. This is critical to avoid breaking your base during the apply step.

Click Refresh Changes to refresh the list. For example, if another user updates the sandbox while you are viewing changes, you will need to refresh to see the latest changes.

Tip: Use the filter feature to focus on specific modules or your own changes — especially if multiple admins are testing in the same sandbox.

Step 4: Apply Changes to Production

After completing all configurations and testing in the sandbox, you can apply the changes to your production base.

4.1 Click View Changes at the top of the sandbox, then click Apply in the lower-right corner. The system will validate all changes before applying them, checking for conflicts with production.

• If validation passes, you’ll see a message: “Validation completed. Ready to apply changes.”
• If validation fails, the failed changes will be listed. You will need to fix the errors and validate again. (Refer to Lark’s documentation on common validation errors and solutions if you encounter any.)

4.2 Once validation passes, click Apply in the lower-right corner again. All changes in the sandbox will be applied to production.

Important: Applying changes will directly affect the production base. If the changes fail to be applied, the base will automatically roll back to the pre-update version. All updates made to the production base while changes are being applied will also be rolled back. Apply your changes during off-peak hours when few or no users are using the production base, and verify the results after completion.

4.3 After applying the changes, click View Changes at the top of the sandbox and select Applied Changes to see the changes that have been applied. You can filter by module, user, or time.

Tip: Notify your team in the company chat before applying major changes — especially if it changes how they interact with the base. A short message saves a flood of “is something broken?” questions afterwards.

Step 5: Refresh the Sandbox

If significant differences develop between your sandbox and the live production base — for example, if production has been updated by other users since you started — you can sync the configuration and data from production to the sandbox.

In the base, click the ⋯ icon in the upper-right corner > Sandbox > Refresh.
Select what to copy to the sandbox and click Refresh.

Important: After refreshing, existing changes and data in the sandbox will be overwritten. Make sure you have applied or saved any sandbox changes you need before refreshing.

Step 6: Delete the Sandbox

Once you no longer need the sandbox, you can delete it.

1. Click View Changes at the top of the sandbox
2. Click Delete Sandbox in the lower-left corner
3. Confirm the deletion

Important: All changes and data in the sandbox will be permanently deleted and cannot be recovered. If you need to use the sandbox again later, you can create a new one anytime.

Real-World Examples for SMEs and Startups

Here are practical scenarios where the sandbox is invaluable:

Restructuring a CRM

You want to split your “Customers” table into “Leads” and “Customers.” A change like this affects views, automations, and reports. Test the entire restructure inside the sandbox first to make sure nothing breaks.

Adding a new automation

You want to auto-send a welcome email when a new sign-up is added. Build and test the automation in the sandbox to confirm it triggers correctly. (Just remember: test emails will reach real recipients, so use a controlled test email first.)

Updating a complex formula

A formula that calculates commissions needs adjusting. Test the new logic inside the sandbox using copied records to confirm the numbers come out correctly.

Onboarding a new admin

Use a sandbox to train a new team admin on Base configuration without giving them access to production data or risking accidental edits.

Seasonal or quarterly changes

Preparing for a year-end campaign, financial year reset, or quarterly review? Build the new structure in the sandbox in advance, then apply on the day you are ready to launch.

Pro Tips for Using the Sandbox Effectively

• Always enable Edit history and Time Machine first so you have a fallback if something goes wrong after applying.
• Plan changes before opening the sandbox. A focused sandbox session is faster and produces cleaner change sets.
• Apply changes during off-peak hours. This minimises the impact on your team and reduces the chance of conflicting updates during the apply window.
• Always check dependencies. Changes with dependencies must be applied together — skipping one will break the others.
• Use the View Changes filter regularly. It is the easiest way to keep track of who changed what, especially in shared sandboxes.
• Communicate with your team. A short heads-up before applying major changes saves confusion and builds trust.

Frequently Asked Questions (FAQ)

1. What is the sandbox in Lark Base?

The sandbox is an isolated testing environment that mirrors your production base. It lets you safely test and validate configuration changes — including fields, views, automations, and workflows — without affecting your live data. Once verified, you can apply the changes to production with one click.

2. How do I create a sandbox in Lark Base?

Open your base, click the ⋯ icon in the upper-right corner, and select Sandbox. Choose whether to copy configurations only or configurations with up to 1,000 rows of records per table. Then click Create.

3. Which modules of Base can be tested in the sandbox?

You can test changes to tables (fields and views) and automations or workflows inside the sandbox. Every configuration change is recorded in the sandbox change set for review.

4. Who can use the sandbox feature?

Only users with manage permissions for the base can create and use the sandbox.

5. Will test emails sent from the sandbox reach real recipients?

Yes. When you test sending emails from automations in the sandbox, the email will be sent to the actual recipient(s). For Lark messages, however, test messages are sent only to the automation configurator — not to the recipient or group.

6. What happens if I apply changes and they fail?

If applying changes fails, the base will automatically roll back to the pre-update version. Any updates made to production during the apply attempt will also be rolled back. This is why applying during off-peak hours is recommended.

7. Is the sandbox available on the Lark mobile app?

No. The sandbox feature is only available on the Lark desktop app or web version.

Build Better Bases with Confidence

The sandbox feature is what separates careful, professional Lark Base builders from anxious ones.

With it, you can experiment, refine, and improve your bases without ever putting your live data at risk.

For SMEs and startups whose operations rely on Lark Base every day, this is more than a convenience — it is essential infrastructure for building robust, scalable workflows.

Make it a habit: any non-trivial change goes through the sandbox first. Your team — and your future self — will thank you.

Ready to Power Your Business with Lark?

Lark Base is just one part of an all-in-one platform that brings messaging, meetings, documents, approvals, and automations together for your entire team.

As the Platinum Partner for Lark in Malaysia, Exabytes offers tailored Lark plans, hands-on onboarding, and dedicated local support to help you build, test, and scale your bases with confidence.

Explore Lark plans with Exabytes today