Connect an OpenAI-compatible agent.
Create a scoped API key, add prepaid credit and set FLONNO_API_KEY in your server environment. Set the base URL to https://www.flonno.com/v1 and the model to qwen3.8-27b. Compatibility covers documented Chat Completions features; check your framework’s endpoint and parameter requirements.
For Python, install the OpenAI SDK with pip install openai, then run:
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["FLONNO_API_KEY"],
base_url="https://www.flonno.com/v1",
)
result = client.chat.completions.create(
model="qwen3.8-27b",
messages=[{"role": "user", "content": "Classify this ticket: My invoice is wrong. Reply with JSON containing category and needs_review."}],
response_format={"type": "json_object"},
max_tokens=200,
extra_body={"reasoning_effort": "none"},
)
print(result.choices[0].message.content)Parse the returned JSON and validate both the category and the review flag before using them. JSON mode guarantees neither your application schema nor factual correctness.
Read the request and error reference →Keep tool execution under application control.
Supply a small tool allowlist with clear parameter descriptions. The model proposes a tool call; your application checks its arguments and permissions, executes the approved action and returns a tool result. Put an iteration limit, timeout and token budget around every loop.
Use read-only tools first. Require approval for payments, deletion or outbound messages. Handle streamed tool-call arguments only after they are complete. Never pass provider credentials through a model prompt.
Tool, n8n and coding-client examples →Choose an affordable workflow you can measure.
Start with ticket classification, product-description drafts, structured document extraction, retrieval-grounded summaries or coding suggestions. Keep a small set of accepted outputs, measure errors and record the complete cost of retries and human review.
At 1,000 input tokens and 300 billed output tokens per request, 10,000 requests cost an estimated $1.05 in tokens before taxes and retries. Reasoning is included in billed output. Use reasoning_effort: "none" deliberately for straightforward requests; evaluate thinking settings for harder tasks.
Understand the service before sending data.
Qwen3.8 Flash 27B runs on our own GPU system in Switzerland. API requests go from the Flonno gateway to Flonno's inference runtime; they are not sent to another AI API provider for model inference. The current deployment uses an RTX 3090 and has no automatic external-provider fallback. Hosting infrastructure remains part of the processing chain; authentication and payments use separate service providers described in the privacy notice.
The public API is a hosted service. Scoped keys and HTTPS support access control and transport security; they do not establish correctness, attestation or legal suitability for every use case. Check the processing and retention notice for your workload.
Privacy and processing details →Read the docs as plain text or JSON.
The same reviewed product facts and prices are available without JavaScript or a login. The text files are convenient documentation formats; support varies by tool and they do not guarantee inclusion in any search assistant.