cef-inference

Run inline ML model inference in CEF agent handlers via context.models.

1|Updated Mar 30, 2026

Search Tags:#runtime #streaming #handlers #inference #ml-models #cef

Ready to run

npx skills add https://github.com/CEF-AI/skills --skill cef-inference

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: cef-inference
Source: https://github.com/CEF-AI/skills/tree/main/skills/inference
Command: npx skills add https://github.com/CEF-AI/skills --skill cef-inference

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Enable inline ML model inference inside CEF agent handlers, allowing models to run directly in the runtime without external API calls.

Core Features & Use Cases

Inline inference via context.models.<alias>.infer(...) for non-streaming results and .stream(...) for streaming outputs.
Supports the full 16-model registry (yolo, yoloXL, whisper, whisperTiny, whisperLarge, llm, llamaVision, qwenVision, qwenCoder, mistral7b, mistralSmall, embedding, emotionClassifier, sentimentAnalysis, plateDetector, plateOcr) with input/output schemas and practical handler examples.
Streaming consumption enables per-chunk results and real-time feedback within agent workflows.
Useful in building end-to-end agent pipelines: vision, ASR, NLP, and code-generation inside a single handler.

Quick Start

Create a CEF agent handler and call context.models.<alias>.infer or context.models.<alias>.stream to perform in-model inference.

Frequently Asked Questions about cef-inference

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I run ML model inference inside an agent handler without external API calls?▼

Inline ML model inference allows models to run directly in the runtime by calling context.models.<alias>.infer() within your CEF agent handler, eliminating the need for external API calls.

Can I stream real-time ML outputs within agent pipelines?▼

Yes, streaming ML outputs within agent pipelines uses context.models.<alias>.stream() to enable per-chunk results and real-time feedback during inference.

What ML models are supported for inline inference in CEF handlers?▼

The 16-model registry supports inline inference for vision, ASR, NLP, and code generation, including yolo, whisper, llm, mistral7b, embedding, emotionClassifier, and plateOcr.

How do I build an end-to-end agent pipeline for vision and ASR tasks?▼

You build end-to-end agent pipelines for vision and ASR by creating a CEF agent handler and invoking context.models.<alias>.infer or .stream to execute models like yolo and whisper.

Does inline inference support both non-streaming and streaming use cases?▼

Inline inference supports both non-streaming and streaming use cases by using .infer() for complete results and .stream() for per-chunk real-time feedback within agent workflows.