Guides

Guides take one task from start to finish with complete, runnable code against the Syntic API. Each one explains the decisions behind the code, how to test the result, and what to add before it goes to production.

If you have not made a request yet, start with the Quickstart.

Build a feature

Guide	What you build	Model used
Build a support assistant	A customer chat that answers from your policies, looks up orders through a function, and hands over to a person.	Amara S1.2
Route support tickets	A classifier that assigns team, priority and a summary to each incoming ticket, with an accuracy check.	Amara Quick
Summarize long documents	Structured summaries of contracts and reports, including documents too long for one request, plus follow-up questions.	Amara S1.2
Build a shopping assistant	A catalog-grounded product advisor that compares real specs and only adds to the basket after confirmation.	Amara S1.2
Moderate user content	A pre-publication check that maps forum posts to your rules and decides publish, hold or reject.	Amara Quick
Classify text	A multi-label review tagger with sentiment, confidence and a per-label accuracy report.	Amara Quick
Extract data to JSON	Validated invoice records from free text, with a corrective retry and arithmetic checks.	Amara S1.2
Answer from your documents	Question answering over a handbook with keyword retrieval, numbered sources and an honest “not covered” path.	Amara S1.2, Amara Quick
Build a coding assistant	A pull-request reviewer that reads files, searches code and runs tests through safe tools.	Syntic Code
Build an agent	A complaint-handling agent with budgets, human approval for refunds, error recovery and a trace.	Amara S1.2
Chatbot with memory	A tutor that keeps recent turns, a running summary and long-term facts the user can delete.	Amara S1.2, Amara Quick
Translate and localise	UI string translation with a glossary, placeholder checks and plural handling.	Amara S1.2

Core techniques

Guide	You will learn to
Write a tool loop	Let the model call your functions repeatedly until a task is done, in Python and Node.js.
Rate limits & retries	Know which failures are safe to send again and back off with jitter.
Handle usage limits	Tell rate limits, plan windows and budget limits apart, and react to each correctly.
Control costs	Use caching, history trimming, model choice and spend limits to keep costs low and predictable.

Conventions used in the guides

• Code reads the API key from the SYNTIC_API_KEY environment variable. Create a key as described in Get an API key.
• Examples call https://app.syntic.ai/v1/chat/completions directly or through the OpenAI client libraries; see OpenAI-compatible clients.
• Most examples set "stream": false to keep the code short. The API streams by default; see Streaming when a person is waiting on the answer.
• Prices quoted in the guides apply to requests paid from usage credits. Requests within your plan’s allowance cost nothing extra.

More ways to learn

• Prompting: how to write instructions, examples and output formats that hold up.
• Cookbook: short, runnable recipes for streaming, retries, cost tracking and more.