How are RAG citations handled by this skill?

It implements Cohere's native RAG support, where documents are passed directly to the chat method and the model returns grounded answers with automatic inline citation metadata.

Does this skill support the legacy CohereClient?

No, this skill enforces the use of CohereClientV2, which is the current standard for all new Cohere integrations to ensure compatibility with the latest API features.

Does this skill help with API security?

Yes, it strictly enforces patterns that source API tokens from environment variables rather than hardcoded strings, following security best practices.

Why is the inputType parameter required for embeddings?

Using the wrong inputType (such as using search_query for document indexing) silently degrades similarity scores. This skill ensures you correctly pair search_document with search_query.

Can I use this for streaming chat interfaces?

Yes, the skill includes optimized patterns for chatStream that correctly handle content-delta events for real-time UI updates.

Cohere AI Provider SDK

Name: Cohere AI Provider SDK
Author: agents-inc

byagents-inc

•

Ciencia de Datos y ML

Integrates Cohere's advanced LLM capabilities including RAG with inline citations, semantic search, and reranking using the official TypeScript SDK.

This skill provides Claude Code with the domain-specific knowledge required to implement Cohere's V2 TypeScript SDK effectively. It focuses on modern patterns for high-performance AI applications, covering the Command-A and Command-R model families, multi-stage retrieval pipelines using embeddings and reranking, and sophisticated Retrieval-Augmented Generation (RAG) with automatic source grounding. By enforcing best practices like correct embedding input types and tool-use loops, it ensures robust, production-grade integrations with Cohere's platform across various environments including Bedrock and Azure.

Características Principales

01Advanced relevance scoring and search result ordering via Rerank v4

02Production-ready tool use patterns and function calling loops

035 GitHub stars

04Implementation of CohereClientV2 for modern chat and streaming interfaces

05RAG workflows featuring automatic inline document citations

06Semantic search optimization with specialized embedding input types

Casos de Uso

01Optimizing search relevance using multi-stage embed and rerank pipelines

02Building enterprise-grade RAG systems with verifiable document grounding

03Creating agentic workflows with Cohere Command model tool calling

Características Principales

01Advanced relevance scoring and search result ordering via Rerank v4

02Production-ready tool use patterns and function calling loops

035 GitHub stars

04Implementation of CohereClientV2 for modern chat and streaming interfaces

05RAG workflows featuring automatic inline document citations

06Semantic search optimization with specialized embedding input types

Casos de Uso

01Optimizing search relevance using multi-stage embed and rerank pipelines

02Building enterprise-grade RAG systems with verifiable document grounding

03Creating agentic workflows with Cohere Command model tool calling