Product documentation

Getting started

Create an account, choose a product, test the workflow, and review usage before deploying anything customer-facing.

1. Create your ChatBucket account or sign in to an existing account.
2. Open the dashboard and confirm your plan, usage limits, available balance, and active products.
3. Choose the product guide below and complete the setup checklist for translation, chat agents, voice agents, or analytics.
4. Run a sample translation, conversation, or call. Review the output, transcript, and usage before going live.
5. Invite teammates only after roles, handoff behavior, and escalation paths are clear.

Authentication

Dashboard pages use your signed-in account session. Server integrations should use credentials created in API keys and stored in your backend environment, not in client-side JavaScript.

Never expose bearer tokens or API keys in browser bundles, public repositories, analytics snippets, or support screenshots.

Try a translation

1. Open Translate and choose Indian or global languages.
2. Select automatic source detection or choose the source language and script.
3. Choose the target language and output script, enter your text, and submit the translation.
4. Review names, terminology, and meaning before using the result in customer-facing workflows.

Request format

Send POST /translate to your configured API base URL with JSON content and an authorized credential. The example below is written for a server runtime.

// Server-side translation request. Keep credentials out of browser code.
const response = await fetch(`${process.env.CB_API_BASE_URL}/translate`, {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    Authorization: `Bearer ${process.env.CB_ACCESS_TOKEN}`,
  },
  body: JSON.stringify({
    text: "Hello, how can we help you?",
    src_lang: "auto",
    tgt_lang: "HINDI",
    isGlobal: false,
  }),
});

const result = await response.json();
if (!response.ok) {
  throw new Error(result.details || result.message || "Translation failed");
}

const translatedText = result.data?.output || result.data?.translated_text;

For production API-key integration, confirm your account endpoint, available languages, and credential requirements with support.

Response handling

Read translated output from the response data object and always handle non-2xx responses. Log request IDs, HTTP status, and error details on your server so support can investigate failures quickly.

Create and publish a widget

1. Create your widget. Set its basic details in the chat widget builder.
2. Add knowledge. Use business details, FAQs, documents, or website sources. Keep policies and product information current.
3. Design the widget. Match colors, labels, and greeting text to your website.
4. Shape the conversation. Configure tone, data collection, fallback wording, and handoff rules.
5. Set availability. Choose websites, pages, devices, allowed domains, and operating hours.
6. Review and publish. Test common questions, publish the widget, and install the generated snippet on your website.

Installation

Install the exact snippet generated for your published widget. The example below shows the general shape only; use the real widget URL and ID from ChatBucket.

<script
  async
  src="https://your-chatbucket-widget-url.example/widget.js"
  data-widget-id="YOUR_WIDGET_ID">
</script>

Validate the customer experience

Test a common question, an unanswered question, a policy-sensitive question, and a request for a human agent. Use widget conversation history to inspect conversations and identify knowledge gaps.

Configure your agent

1. Open the voice agent builder. Set the agent name, purpose, voice, languages, and greeting.
2. Add knowledge, actions, and instructions. Define the questions the agent should answer and the outcomes it should help customers reach.
3. Configure inbound routing, transfers, callbacks, and human takeover rules when those features are enabled for your account.
4. Test realistic questions, interruptions, silence, repeated answers, and requests outside the agent scope.
5. Review summaries, transcripts, recordings, and call outcomes before launching a live workflow.

Deployment and integration availability depend on your account configuration. Confirm phone provisioning, recording requirements, and external integrations with support before launch.

Review calls

Open voice agent history to review available call records. Use recurring questions, transfers, and incomplete outcomes to improve instructions and knowledge.

Review conversation performance

1. Open your dashboard overview after generating conversations or calls.
2. Choose the reporting period and available widget, team, channel, or agent filters.
3. Compare conversation volume and resolution across AI and human conversations.
4. Inspect conversation history to understand the questions and outcomes behind the numbers.

Keep the reporting period and filters consistent when comparing results. Unavailable feedback or response-time values do not represent a score of zero.

Security and privacy

Customer conversations can include personal or business-sensitive information. Configure each product with the minimum data needed for the workflow.

1. Use approved knowledge sources and remove outdated documents, URLs, and FAQs.
2. Do not ask for payment credentials, government IDs, passwords, or health information unless your workflow and account are explicitly prepared for that data.
3. Restrict widget domains and team access where available.
4. Review transcripts and recordings according to your internal retention and consent policies.
5. Revoke API keys, widget access, and teammate permissions that are no longer needed.

Account & usage

Manage credentials in API keys. Give each integration a recognizable name, store the key securely when it is created, and revoke credentials that are no longer needed.

Review Usage for consumption, Limits for account constraints, and Billing for payment details. Check API status when investigating service availability.

Troubleshooting