Coding assistants
Plan, edit and test loops send large prompts many times per task.
The shape of the workload
- Large input (file context), modest output.
- Many calls per task, often in a tight loop.
- Latency matters more than for batch work — the developer is waiting.
Choosing a model
The catalog includes a dedicated coding model alongside the frontier and balanced tiers. Compare input and output prices per million tokens in Models: with file context, the input price dominates.
Pointing an existing tool at Astrodyne
Most editor assistants that speak the OpenAI Chat Completions API accept a custom base URL and key. Set them to:
Config
Base URL: https://api.astrodyne.ai/v1 API key: $ASTRODYNE_API_KEY Model: YOUR_MODEL_ID
Check the field list first
Tools sometimes send fields Astrodyne does not implement. The
compatibility page lists exactly which
request fields are supported.
Stream for perceived speed
Set "stream": true so tokens arrive as they are produced. The
charge is settled after the stream closes, from the tokens actually delivered.