Agent platforms
Connect tools that support an OpenAI-compatible base URL. AdaptiqAI remains the model provider, while your chosen agent owns tools and execution.
The engine is not the agent harness
AdaptiqAI preserves provider semantics across frameworks.
Generic OpenAI-compatible setup
text
Base URL: https://api.adaptiqai.com/v1
API key: ak-...
Model: adaptiq-deepseek-v4.1-flashEnvironment variables
bash
export OPENAI_API_KEY="$ADAPTIQ_API_KEY"
export OPENAI_BASE_URL="https://api.adaptiqai.com/v1"
export OPENAI_MODEL="adaptiq-deepseek-v4.1-flash"Conversation identity and memory
Use a stable provider session or conversation identity when the framework supports it. The provider deduplicates replayed chat history and captures new user, tool and assistant events into the scoped Native Memory timeline.
- Do not share one session ID between unrelated users.
- Keep API keys and tenant identity outside prompt text.
- Select shared Knowledge explicitly in the Console.
- Use the model ID returned by /v1/models.
Frameworks
Tested with Pi, DeepSeek Harness (dsh) and Hermes. Any other client that lets you set an OpenAI-compatible base URL, API key and model is configured the same way.
Turn off the harness's own compaction
Native Memory keeps the history outside the prompt. A harness that summarises or trims the conversation on its own rewrites that history, so turn its compaction or auto-summary setting off.