Saltar al contenido
TakoAPI
Volver a skills

gradient-inference

por simondelorean

DevOps & Cloud
478Descargas según clawskills.sh
1Vistas
0 me gustaInicia sesión para dar me gusta a esta skill
—Aún sin valoraciones
Tu valoración

Resumen

Community skill (unofficial) for DigitalOcean Gradient AI Serverless Inference.

Descripción

Qué hace esta skill

Connects to DigitalOcean's Gradient AI Serverless Inference API to run chat completions, generate images, and look up available models and pricing. The endpoint is OpenAI-compatible. Prompt caching is available via the Responses API to reduce costs on repeated context.

Because the endpoint is OpenAI-compatible, existing SDK code works with only a base URL change, and prompt caching via the Responses API reduces spend on follow-up queries that reuse the same context.

Cuándo usarla

  • Running LLM queries without managing GPU servers
  • Checking model pricing before committing to a workload
  • Generating images from text prompts via serverless API
  • Using prompt caching to cut costs on repeated context
  • Browsing and filtering available models before hardcoding an ID

Flujo de trabajo de ejemplo

ENTRADA

User asks: summarize this earnings report and compare it to last quarter

AGENTE
  1. 1

    Run gradient_pricing.py to check model costs before choosing

  2. 2

    Select a model based on context length and cost requirements

  3. 3

    Send the earnings report via the Responses API with store: true to cache the context

  4. 4

    Send the follow-up comparison query, reusing the cached context

SALIDA

Summary of the earnings report, then a cost-efficient comparison to last quarter with shared context billed only once due to prompt caching

Requisitos

  • GRADIENT_API_KEY environment variable (DigitalOcean Model Access Key from cloud.digitalocean.com)
  • DigitalOcean account

Instalación

clawhub install simondelorean/gradient-inference

Ejecuta el comando de la CLI o pega el prompt en Claude Code, Cursor, Codex o el agente de programación que uses.

Enlaces