Messy text is everywhere: support tickets, lead forms, emails, contracts, incident reports, call notes, Slack messages. The annoying part is that the useful data is usually in there somewhere, but not in a shape your app can trust.
I built a small Python example that uses Telnyx AI Inference to turn unstructured text into structured JSON.
Repo: https://github.com/team-telnyx/telnyx-code-examples/tree/main/extract-structured-json-with-ai-python
What it does
The app exposes a simple Flask endpoint where you send messy text plus the fields you want back.
For example, you can send something like a customer support note and ask for:
- customer name
- issue type
- urgency
- product
- next action
The model returns a structured JSON object that your app can validate, store, route, or pass into another workflow.
Why this pattern is useful
A lot of AI demos stop at “summarize this text.” That is useful, but many real apps need something stricter:
- route a ticket
- classify an incident
- extract lead details
- normalize intake forms
- prepare data for a CRM
- trigger automations based on extracted fields
Structured JSON makes the LLM output easier to use in actual software.
How it works
The example uses Telnyx AI Inference through an OpenAI-compatible client pattern.
At a high level:
- Define the schema you want back
- Send messy text to the model
- Ask the model to return JSON
- Validate the result before using it
That last part matters. Even when you ask an LLM for JSON, your app should still treat model output like external input and validate it.
Try it
Clone the repo:
git clone https://github.com/team-telnyx/telnyx-code-examples.git
cd telnyx-code-examples/extract-structured-json-with-ai-python
Install dependencies and run the app:
pip install -r requirements.txt
cp .env.example .env
python app.py
Then call the endpoint with your own text.
Why I like this example
It is small enough to understand quickly, but it maps to a very real app pattern: taking messy human language and turning it into data your system can actually act on.
Also worth noting: the Telnyx code examples repo is structured to be agent-readable, so coding agents can inspect the examples, understand the API patterns, and help you extend them into fuller apps.
Resources
Code example
Telnyx AI repo with skills/toolkits
Telnyx AI Inference
Feedback welcome, especially from folks building AI apps that need reliable structured output from messy user input.