Kingy AI
Subscribe, 2.1M subscribersWork with Kingy
AI News

Creating a Real-Time AI Brain with ChatLLM Teams

November 24, 2024 AI News
Purple and White AI Tips YouTube Thumbnail editorial image

Updated October 5, 2026 UTC. This guide began in November 2024. For an AI project, start by deciding how a chatbot should read documents. ChatLLM Teams now documents two different connector approaches. Imported datasets do not guarantee immediate updates. The embedded video is a related historical GitHub integration tutorial, not a Google Drive walkthrough.

Pricing and tutorial context

Abacus advertises $7 for the first month, then $10 per user/month. Its billing FAQ lists Basic at $10 with 20,000 monthly credits and Pro at $20 total with 30,000; a $7 promotion may provide only 14,000 credits. Agent access and credit limits differ by tier. These are published terms, not a purchase or account test.

Choose the document access method

  • For questions about your own Drive files, use the Google Drive user connector with RouteLLM.
  • Abacus describes user connectors as reading the source on each request under the signed-in user’s permissions.
  • For a curated shared document collection, use an organization connector to import a dataset. Abacus distinguishes scheduled refresh from supported realtime-sync configurations; simply selecting a connector does not establish either.

Historical AI Engineer screenshot

The November 2024 screen above offers “Create a custom chatbot” for customized behaviour/contextual search and “Create an AI Agent [Beta]”; its agent option warns that generated code may need editing. It shows those choices, not a freshness test.

Define the assistant’s job

For example, call it “Company FAQ” and instruct it to answer from approved documents, identify its sources and say when the answer is missing. Keep a small set of questions with known answers for evaluation.

Decide whose documents it should see

The current Custom Chatbots guide separates searching with user-connector tools from ingesting a shared document set. The choice affects permissions and how documents enter the chatbot’s knowledge base.

The original Dataset Wizard

The screenshot below records the earlier dataset-selection interface:

  • Select Existing Dataset(s).
  • Upload a file/zip.
  • Use a connector.
  • Those options describe how data is supplied; they do not show when a later document edit becomes available.

Historical screenshot: the AI Engineer message offers available datasets or new company documents. The open Dataset Wizard lists Select Existing Dataset(s), Upload a file/zip and Use a connector. It contains no displayed refresh interval or synchronization result.

Current route for personal Drive questions

  • Open ChatLLM’s Connectors link, or Profile → User Connectors.
  • Select Google Drive and review the permissions on the Google consent screen before authorizing it.
  • Start a RouteLLM chat and ask a question referencing your Drive documents. This route does not require training a custom chatbot.

For a shared document dataset:

  • In the Abacus developer platform, create a Custom Chatbots project.
  • Choose a descriptive project name; “Company FAQ” is an example, not a required value.
  • Under Datasets, choose Create Dataset and the List of Documents type.
  • For Drive ingestion, use Read from External Service and select the Google Drive application connector.

Set up the organization connector

  • The Drive import guide uses Manage Connectors → Add New Connector → Google Drive.
  • For its folder-level option, Abacus provides a service-account address; the guide instructs you to share that folder with Viewer permission.
  • Review which files and people should have access before connecting a company folder.

Select the import scope

  • Specify the required file-location expression, or use the application browser to select the file or folder.
  • Begin with a small approved folder so you can check exactly what was imported.
  • Complete the import options and schema mapping described in the Drive guide.
  • The current guide does not specify an immediate refresh guarantee for this Google Drive dataset path.

Inspect the imported documents

  • The Custom Chatbots guide places imported documents in Feature Groups.
  • Materialized Data shows the document rows; Features shows the mapping.
  • For ingested RAG documents, create a Document Retriever for the feature group.
  • A user-connector-only chatbot does not require that document retriever.

Configure and deploy

  • Select the model’s document source and behavior/response instructions, then Train Model.
  • When the model is ready, follow Deploy. The Dashboard opens an external chat interface.
  • A dedicated bot can instead use Google_Drive_tool under Advanced Options → Tool Use; each end user signs in under their own identity.

Test freshness before relying on it

  • Ask a question with a known answer and check the document and version supporting it.
  • Change a harmless test fact in Drive, record the edit time, then check retrieval and the answer again. For an imported dataset, verify the actual refresh/retrieval state. Record any delay; do not assume a successful connection proves immediate updates.

Useful evaluation cases

Try these before sharing the assistant:

  • FAQs: compare answers with the approved source and include a question the documents cannot answer.
  • Policies: check both the latest approved version and what a less-privileged user can retrieve.
  • Project documents: edit one test fact and measure when the assistant can find it.

What the evidence establishes

The original title’s “real-time” language should not be read as a guarantee for the dataset workflow. Current Abacus documentation describes request-time user access and separate ingestion paths. This update reviews those documents and the original screenshots; it does not report a new connected-account or latency test.

Official documentation

Read Abacus’s connector comparison, Google Drive user-connector guide, Drive dataset-import guide, Custom Chatbots guide and billing FAQ.

ChatLLM Platform Guides

ChatLLM Agents

ChatLLM Productivity