Basalt: quickly build and optimize cue words in AI applications

General Introduction

Basalt is a cue building and management platform designed to help teams efficiently build high-quality AI cues and apply them to their products. It provides a code-free editor, intelligent optimization suggestions, and versioning capabilities that enable product managers, engineers, and domain experts to quickly design, test, and refine cues. Through structured evaluation and batch testing, Basalt ensures that prompts are efficient and reliable, supporting a seamless transition from prototype to production. Whether it's optimizing customer interactions or improving the efficiency of content generation, Basalt accelerates the development process of AI features and has become a highlight of the HX24 startup ecosystem.

Function List

• No Code Prompt Word Creation: Quickly write and adjust AI cues through an intuitive interface, with no programming skills required.
• Version tracking and management: Record each modification of the cue word, support rollback and compare the effect of different versions.
• Smart Cue Optimization: The AI assistant provides real-time suggestions to improve the clarity of the cue words and the quality of the output.
• Batch test function: Test multiple cue word versions simultaneously, analyze their performance and filter for the best results.
• Data-driven assessment: Built-in evaluation tool to quantify the accuracy and usefulness of cue words.
• One-click deployment support: Push optimized cue words to production environments quickly.
• Real-time effect monitoring: Track the operational performance of deployed cues and make timely adjustments.
• Teamwork model: Support for multiple co-editors and management of cue word projects.

Using Help

Basalt is an online platform that requires no complicated installation, just sign up and start building and managing high-quality prompts. Here's a detailed guide to help you get started quickly and get a deep grasp of its core features.

Registration and Initial Setup

1. Open your browser and visit https://www.getbasalt.ai/The
2. Click "Sign Up" in the upper right corner, enter your email and password to complete the registration process.
3. Check your mailbox and click on the verification link you received to activate your account.
4. Return to the official website, log in with your registration information and enter the main screen.

Building high-quality cue words

1. Go to Project Management
   After logging in, click "Dashboard" on the left side to see the overview of all projects. Click "New Project", enter a name (e.g. "Smart Quiz Reminder") and click "Create".
2. Creating Prompts
   Select the Prompts tab within the project to enter the editor. Enter an initial prompt, such as "Provide concise and accurate answers to user questions". The editor will display a live preview.
3. Leveraging Intelligent Optimization
   Click on the "Co-pilot" button on the right and the AI will give you suggestions for optimization based on your inputs, e.g. "Provide a concise and accurate answer to the user's question, up to 20 words". Click "Apply" to take the suggestion.
4. Saved version
   When you have finished modifying, click "Save Version", the system will store the current cue word automatically. You can compare the differences between different versions in the "Version History".

Testing and refining cue words

1. batch test
   On the "Prompts" page, click "Test" to upload test data (CSV or manually enter a list of questions). The system will run all versions of the prompts and generate results.
2. analysis and assessment
   After testing, go to the "Evaluations" tab to see the scores (e.g. accuracy, response length) for each cue word. Select the best version and click "Set as Active".
3. Iterative optimization
   Return to the editor based on the evaluation results, adjust the cue word details (e.g., add specific constraints), save and test again until the desired effect is achieved.

Deployment and Applications

1. Deployment Cues
   On the project page, click "Deploy" and select the optimized version of the prompt.
2. Integration into products
   Click "Generate API" to copy the key and code provided by the system. Embed the code into your product (e.g. website backend or application logic) to complete the integration.
3. go online
   Once saved, the cue word is ready to run in the production environment and users can experience the results directly through your product.

Monitoring and Maintenance

1. Viewing Operational Status
   Click on "Monitoring" in the left navigation bar to view real-time data on deployed prompts, including the number of invocations and success rates.
2. Exception handling
   If you notice a drop in output quality, click on "Alerts" to see details of the problem. Return to "Prompts" to adjust the prompts or switch versions and redeploy.
3. Teamwork
   In the "Team" tab, enter a colleague's email address to invite them to join the project, assign editing or viewing rights, and work together to optimize the prompt.

Core Functions

• No Code Prompt Word Creation
  The editor supports natural language input and parameter adjustment. For example, if you type "Generate a 200-word product description", you can select the model and output style from the drop-down menu, and complete a high-quality prompt in a few simple steps.
• Version tracking and management
  Each time you save a version, the system will record the time of modification and a summary of the content. By clicking "Version History", you can view specific changes (e.g., "increased word limit") and switch to any version with one click.
• Smart Cue Optimization
  "Co-pilot suggests more precise expressions based on the context. For example, if you type "Answer a user question", it might suggest "Generate a 50-word, friendly answer to the user's question".
• Batch test function
  When testing, you can upload a dataset of 100 questions, and the system will automatically run all the versions and generate a comparison chart to help you visually pick the best cue words.

Practical advice

• Be as specific as possible when writing prompts, e.g., "List 3 characteristics and explain" is preferable to "Provide some information".
• Regularly check the version history to ensure that the best solution is retained.
• Sharing testing and optimization tasks in team mode improves efficiency.

With these steps, you can easily build and manage high-quality cues to apply them to real-world scenarios.Basalt's core strength is in simplifying and accelerating cue development, allowing teams to focus on creating value.