Skip to main content
The Pioneer REST API lets you run inference, manage datasets, start training jobs, and evaluate models programmatically. All resources live under a single base URL, and every request is authenticated with your API key. Pioneer also exposes OpenAI- and Anthropic-compatible endpoints so you can integrate with any tool that already supports those formats. Base URL: https://api.pioneer.ai

​
Authentication

Every request requires an X-API-Key header. Generate your key at pioneer.ai → Settings → API Keys. See the Authentication page for full details, including key management endpoints. Follow this sequence when you’re building a fine-tuned model end to end:
1

Create or upload a dataset

Use POST /generate to generate synthetic labeled data — the response includes a job_id, poll GET /generate/jobs/{job_id} until status returns ready.To upload your own labeled data, use the three-step upload flow:
  1. Call POST /felix/datasets/upload/url to get a presigned S3 URL,
  2. PUT your file directly to S3
  3. Call POST /felix/datasets/upload/process to trigger processing.
Poll GET /felix/datasets/{name}/{version} until status is ready.
See the Datasets page for full details.
2

Start a training job

Call POST /felix/training-jobs with your model_name, datasets, and base_model. The response includes a job_id — save it for the following steps.
3

Poll training status

Call GET /felix/training-jobs/:id repeatedly until the status field is complete. Completed jobs include F1, precision, and recall metrics.
4

Run an evaluation

Call POST /felix/evaluations with your training job ID and a held-out dataset name to validate performance before you serve traffic.
5

Run inference

Call POST /inference with base_model set to your training job ID and dataset_name set to your held out dataset. You can also use any base model ID (for example fastino/gliner2-base-v1) without fine-tuning.
Replace placeholder values such as YOUR_DATASET_NAME and YOUR_TRAINING_JOB_ID with real values before running any example commands.

​
Endpoint groups

Inference

Run predictions against fine-tuned or base models using the Pioneer format, or via drop-in OpenAI- and Anthropic-compatible endpoints.

Training & Data

Manage datasets, generate synthetic training data, start and monitor training jobs, and run evaluations.

Projects

Organize resources into projects, deploy trained models, and run inference against a project endpoint.

API Keys

Create, list, and revoke API keys programmatically.

​
OpenAI and Anthropic compatibility

Pioneer exposes drop-in replacements for the OpenAI and Anthropic SDKs. Point your existing client at https://api.pioneer.ai/v1 and use your Pioneer API key — no other changes are required.
SDKBase URLNotes
OpenAIhttps://api.pioneer.ai/v1Set base_url; pass Pioneer-specific fields via extra_body
Anthropichttps://api.pioneer.ai/v1Set base_url; supports streaming
Pass Pioneer-specific fields like schema using extra_body (OpenAI SDK) or the equivalent extra-parameters mechanism for your client library.