Replicate Skill

Alpha Software Notice

This skill is in alpha stage and under active development. APIs, features, and functionality may change without notice. Use with caution in production environments and expect potential breaking changes in future releases.

Advanced machine learning model execution via Replicate's API. Run any public model, from text generation to image creation, with secure credential management and real-time monitoring.

Features

Run ML models - Execute text, image, audio, and video models
Model discovery - Browse and get detailed information about available models
Real-time monitoring - Track prediction progress and get results
Secure API token storage via auth and KV skills
Cancel predictions to save compute costs
Per-user isolation with authentication

Quick Setup

from webagents.agents import BaseAgent
from webagents.agents.skills.ecosystem.replicate import ReplicateSkill

agent = BaseAgent(
    name="replicate-agent",
    model="openai/gpt-4o",
    skills={
        "replicate": ReplicateSkill()  # Auto-resolves: auth, kv
    }
)

Core Tools

`replicate_setup(api_token)`

Set up Replicate API credentials with automatic validation.

`replicate_list_models(owner)`

List available models, optionally filtered by owner (e.g., "stability-ai").

`replicate_get_model_info(model)`

Get detailed information about a specific model including parameters.

`replicate_run_prediction(model, input_data)`

Run a prediction with a model and input parameters.

`replicate_get_prediction(prediction_id)`

Get prediction status and results.

`replicate_cancel_prediction(prediction_id)`

Cancel a running prediction to save costs.

Usage Examples

Text Generation

messages = [
    {"role": "user", "content": "Generate a haiku about AI using LLaMA-2"}
]
response = await agent.run(messages=messages)

Image Generation

messages = [
    {"role": "user", "content": "Create an image of a dragon flying over mountains using Stable Diffusion"}
]
response = await agent.run(messages=messages)

Model Discovery

messages = [
    {"role": "user", "content": "Show me what video generation models are available from Stability AI"}
]
response = await agent.run(messages=messages)

Audio Processing

messages = [
    {"role": "user", "content": "Transcribe this audio file using Whisper: https://example.com/audio.mp3"}
]
response = await agent.run(messages=messages)

Getting Your Replicate API Token

Sign up at replicate.com
Go to Account Settings → API tokens
Create token and copy it (starts with r8_)

Model Categories

Text Models: LLaMA, GPT alternatives, chat models, code generation Image Models: Stable Diffusion, DALL-E alternatives, image editing, upscaling Audio Models: Whisper transcription, text-to-speech, music generation Video Models: Text-to-video, video editing, style transfer Multimodal: Vision-language models, image captioning

Pricing

Free tier: Limited compute credits per month
Pay-per-use: Charged based on compute time and model complexity
Model-specific pricing: Check individual model pages for costs

Troubleshooting

Authentication Issues - Verify API token starts with r8_ and is correctly copied Model Not Found - Use model discovery tools to find correct model names Invalid Input - Get model info to see required parameters and types Long Predictions - Some models take minutes; use status monitoring Failed Predictions - Check error messages and verify input data format

Architecture

The skill integrates with Replicate's REST API v1, using secure credential storage via KV skill and per-user authentication via Auth skill. Supports both synchronous and asynchronous model execution patterns.

Integration Tips

Monitor costs - Check model pricing before expensive predictions
Use appropriate models - Choose right model size for your task
Cancel unused predictions - Save costs by canceling long runs
Cache results - Store outputs to avoid re-running identical predictions
Test first - Try models with simple inputs before complex tasks