AI-powered lead filtering from any data source.

1. Click Remix
2. Save your OPENAI_API_KEY as an environment variable
3. Customize PROMPT.txt. Don't remove anything from the structured output. (But you can add fields!)
4. Get the val's HTTP endpoint from main.ts:
5. Forward any leads to the val's HTTP endpoint as a POST request.
6. That's it! You now have a growing leads sqlite table with one column for your original input_data and second column for the AI evaluator's output_data.

• When a new lead comes in via POST, it is sent to the OpenAI agent in agent.ts.
• The agent uses web search and the lead data from your POST request to determine if it's a match for the criteria in PROMPT.txt. By default, the AI's output_data includes match (true or false), it's reasoning, and a normalized name for the lead.
• The input_data and agent's output_data each get a column in the leads sqlite table.
• The main.ts dashboard shows a history of all lead assessments, successful matches first. Clicking any lead shows the full inbound and output data from the assessment.

• You can add any number of fields to the AI's output instructions in PROMPT.txt and they'll be added to the output_data json in the sqlite table and be viewable in the dashboard. This can be used to add more colr to the AI's assessments or simply normalize data from different sources.
• You can also add query parameters to the end of your val endpoint. (ex: yourval.val.run?source=singup-page). They'll be bundled in the _query object of 'input_data.' This can be useful for tracking where a lead came from.