AskTable AI Observation
Make the "thinking process" of AI visible, accurate to find, and quick to modify.
What is AskTable AI Observation Service?
During the use of AskTable, have you encountered these issues:
- The query results are inaccurate, but you don't know where the problem is?
- Some users' questions take a long time, or even result in errors, but you don't know how to troubleshoot?
- You want to uniformly optimize certain types of questions, but you can't understand how the AI "thinks"?
The AI observation service is precisely designed for this. It can fully record and visualize the "reasoning chain" and "thought process" of AskTable when processing each query — including model calls, key steps, generated content, time consumption and resource usage, error points, etc., making AI go from a "black box" to a "transparent glass box".
Why do you need it?
| Functional Module | What is it useful for you? |
|---|---|
| 🌐 Full-chain Visualization | Step by step, unfold each layer of the reasoning process from user questions to final results, helping you understand how the system "thinks". |
| 🔍 Accurate Problem Troubleshooting | Supports step-by-step identification of bottlenecks, high time consumption, and errors, suitable for operations personnel and data engineers to self-troubleshoot. |
| 📈 Query Effect Optimization | Understand the logic of AI in selecting tables, generating SQL, filtering data, etc., to identify optimization points. |
| 🧠 Transparent and Trustworthy AI | Demonstrate the "interpretability" of AI results to business stakeholders and management, building trust. |
Example Screenshots: Transparent Intelligent Agent "Reasoning Chain"
The following image shows a screenshot of AI observation logs over a period of time:
The next two images show the real reasoning process of a user's query from two perspectives:
- The time consumption, input, and output of each model call are clearly visible.
- Which steps depend on external data? Have they hit the cache? Have any errors occurred?
- The final generated SQL and summary results can be directly compared with the database execution results.
Differences Between Versions
| Deployment Version | Includes Observation Service | Notes |
|---|---|---|
| Standalone Experience Edition | ❌ Not Supported | This version retains only basic functions to reduce resource usage. |
| Private Deployment Enterprise Edition A | ✅ Supported | ✅ Recommended version, complete features, high cost-effectiveness |
| Private Deployment Enterprise Edition B | ✅ Supported | Supports offline operation and deep customization, suitable for large customers |
Why Choose Private Deployment Enterprise Edition A as Your First Choice?
- ✅ Possesses complete observation capabilities, ensuring AI is controllable, adjustable, and interpretable.
- ✅ Offers private deployment + data desensitization encryption, ensuring data privacy and compliance.
- ✅ 3-day delivery and launch + dedicated technical support team for assistance.
- ✅ High cost-effectiveness: Starting price of 20,000 yuan per year, including a gifted computing power package, suitable for small and medium-sized enterprises' budgets.
Scenario Recommendations
- ✅ Data-intensive departments (such as data platforms, IT departments)
- ✅ Technical teams that need to review or optimize AI outputs
- ✅ Organizations with internal SQL or BI capabilities, aiming to further improve efficiency and transparency