Icosa API

Build with Icosa's hosted reasoning and finetuning APIs.

Fast path to production: discover allowed models, run inference, launch finetuning jobs, evaluate asynchronously, and download model artifacts.

API Key Auth Streaming Inference Async Evaluation Model Downloads

This documentation intentionally focuses on the currently supported external surface:

• model discovery (/cr/get_allowed_models)
• inference (/cr/gcloud_predict_firebase_base_oss)
• finetuning job creation/status (/cr/generate_questions, /cr/generate_questions_status)
• async evaluation (/cr/finetune_model/evaluate_async, /cr/finetune_model/evaluate_status)
• model artifact download (/cr/finetune_model/download)

Base URL

https://cr-api.icosacomputing.com

Quick links

• Start here: Getting Started
• Auth details: Authentication
• Models endpoint: Endpoints / Models
• Inference endpoint: Endpoints / Inference

Models

GET /cr/get_allowed_models

List the models available for the current API key.

Inference

POST /cr/gcloud_predict_firebase_base_oss

Primary prediction endpoint with optional streaming and RAG.

Finetuning Jobs

POST /cr/generate_questions

GET /cr/generate_questions_status

Submit and poll long-running finetuning preparation jobs.

Evaluation

POST /cr/finetune_model/evaluate_async

GET /cr/finetune_model/evaluate_status

Async submit-then-poll evaluation flow for production reliability.

Downloads

GET /cr/finetune_model/download

Return redirect or short-lived SAS URL metadata for model artifacts.