QDivZero for developers
From the idea you want to test to the feature your users need.
You want to bring AI into your product and find out whether it solves a real problem. QDivZero gives you access to models and prepares inference for your deployments, so you can focus your development work on the application.
For your applications, agents and development tools.
01 /Remove integration roadblocks
The model you chose at the start may not be the one you need later. Your application should evolve with what you learn about it. QDivZero lets you test those decisions through a consistent access pattern.
01 / Switch models
If you use Chat Completions, reuse your client with QDivZero’s URL and key. Change the model ID and review its tool, streaming and context options without building a different integration for each provider.
See how to connect the API02 / Publish your own endpoint
You have found a model that fits the task. QDivZero checks its requirements, finds compatible GPU capacity in your chosen region and prepares inference. You get an endpoint to call from your code.
See how deployments work03 / Understand what you will pay
Compare per-token pricing with the per-second cost of a dedicated deployment. Unlimited Basic also covers usage included in its plan. Choose according to your application’s volume and continuity, with rates in view.
Explore usage options02 /Different problems. Real work.
Sometimes it is a client project. Sometimes it is a hypothesis your startup needs to validate or a process your team wants to improve. What matters is getting to something another person can use.
Add an assistant or automation to a client project. A documented integration and a defined endpoint make handover and ongoing maintenance easier, including when someone else takes over.
Put a first version in front of your users. Watch what they ask, which answers help them and where they get stuck. Use that evidence to refine the solution before extending it across the product.
Connect backends, scripts and automations to a common foundation. Your team can define how to access inference and reuse that work across applications instead of starting again in each department.
03 /AI within your workflow
Bring QDivZero into tools that support a configurable Chat Completions provider. Choose the environment you use to code or automate tasks and connect your account’s models there.
Coding from the terminal
Use your account’s models within Copilot CLI through a custom OpenAI-type provider. Your chosen model must support streaming and tool calling.
View official setupAn agent you can adapt
Keep your Pi extensions and skills while using QDivZero. Register the connection in models.json with the openai-completions protocol and add the model you want to work with.
View official setupTasks and automations
Connect Hermes workflows to a custom OpenAI-compatible endpoint. Select a model that supports the tools your agent needs.
View official setupWork on repositories
Use QDivZero in your OpenCode development sessions. Add a provider with @ai-sdk/openai-compatible and choose which models to offer in your configuration.
View official setupA workspace for Mac
Research, write code and review documents from your desktop. Follow the agent’s activity and file changes as it works through a QDivZero connection.
Explore HarnessInference inside your product
Bring responses into your application flow with Python or JavaScript clients, or HTTP requests. The example below shows a call using the Python SDK.
Open the API referenceThe tool must support a configurable endpoint. Streaming, vision, context and tool calling depend on the model and deployment. Access conditions for external tools are managed separately.
04 /Turn the example into a request
Create an API key, select a model available to your account and set these two environment variables. Run the example to check the connection before bringing it into your application.
https://api.qdiv0.com/v1/v1/chat/completionsTest with the Python SDK
1import os
2from openai import OpenAI
3
4client = OpenAI(
5 base_url="https://api.qdiv0.com/v1",
6 api_key=os.environ["QDIVZERO_API_KEY"],
7)
8
9response = client.chat.completions.create(
10 model=os.environ["QDIVZERO_MODEL"],
11 messages=[{"role": "user", "content": "Summarize the pending decisions for this project in three points."}],
12)
13
14print(response.choices[0].message.content)Install the openai package and set QDIVZERO_API_KEY and QDIVZERO_MODEL in the environment where you run app.py.
The request uses the model you configure. Usage is billed according to your active access option.
05 /The decisions that come with users
Once people start using the feature, new questions appear. Which requests need a more capable model. What content can enter the system. Where to find your own information. Address each need with the platform’s capabilities.
A simple question and a reasoning task do not have to follow the same route. Define strategies with Model Routing and prepare alternative destinations for your requests.
Organize your routesYour users’ requests form part of your system’s input. Apply LLM Firewall rules to detect prompt injection, jailbreaks and risky content before inference.
Configure your rulesYour manuals, documents and catalogs hold information the model needs for its task. Retrieval finds relevant passages through semantic search so you can use them as context.
Work with your dataWe apply zero logs to inference content by default. Chat Completions lets you work without persistent history in the platform. If you use Responses for ongoing conversations, its context is retained and can be managed. For location and isolation requirements, explore available European regions and dedicated infrastructure.
Explore security controls06 /Before connecting your project
Compatibility, setup and usage conditions to get started on a clear foundation.
It brings model access and deployment preparation together. For dedicated compute, the platform finds compatible capacity and configures inference through to publishing the endpoint. You use the API and dashboard to manage those resources.
It must let you configure a Chat Completions API base URL, key and model ID. Also check that the model supports the features the tool uses.
Use https://api.qdiv0.com/v1, a QDivZero key and a model available to your account. Follow each tool’s custom provider guide. Agents using streaming and tool calling need a model that supports both.
Beyond the URL, key and model ID, review supported parameters, tool-call formats, vision, streaming and context window. Test your application’s behavior before sending production traffic.
Yes. Dedicated compute is billed while its resources remain active. Check the price before deploying and use the scheduling or serverless options supported by its configuration.
Yes, provided it meets the technical execution requirements. Choose compatible Hugging Face or private models and review the available deployment options.
No. Copilot and other tools keep their own plans, permissions and conditions. QDivZero usage and access to those tools are managed separately.
Build a first version, share it and discover what is worth taking further.
Access the platform