Agents Manager

​

Acts as the link between Olivier and the sub-agents. Receives requests, identifies the most relevant sub-agent(s), and delegates tasks to them. 

Sub-agent Brainstorming

​

High-level strategic advisor, objective and bias-free. Provides clear analysis and concrete, actionable recommendations.

Sub-agent Social Media

​

Plans social media content and strategies. Proposes detailed action plans for execution and monitors analytics.

Sub-agent AI Coach

​

Suggests the content of coaching sessions based on the client, their activity, and their knowledge of AI, all in an anonymized manner.

Sub-agent Visuals Manager

​

Generates, edits, and manages image and video content, including avatar-based or avatar-free creation, transcription, image generation, and video editing.

Sub-agent Internet Research

​

Performs targeted or in-depth web research, generates summaries and detailed reports with sources.  

Sub-agent Tech Stack Manager

​

   Oversees the technical inventory (tools, systems, software) of Keiko for Business and the tools used by our clients. Automates monitoring through dedicated tools.

Sub-agent Workflow Manager

​

Manages Keiko for Business automations of repetitive tasks.

Sub-agent Knowledge Base Manager

Monitors continuously new AI models to ensure our AI models knowledge is always up to date.

Google: Gemini 2.0 Flash

​

  Specialized in ultra-fast multimodal* processing with a context window** of 1 million tokens***, this model excels particularly in the simultaneous analysis of text, images, audio, and video.

Anthropic: Claude Sonnet 4

​

  A cutting-edge model optimized for complex programming tasks. It features exceptional long-term planning capabilities.

Google: Gemini 2.5 Pro

​

An experimental model focused on deep reasoning and solving complex ethical problems, seamlessly integrating text, images, and search results.

Anthropic: Claude 3.7 Sonnet

​

  The first “hybrid reasoning” model capable of smoothly switching between immediate responses and in-depth, step-by-step analysis depending on task complexity.

OpenAI: GPT-4.1

​

A flagship model optimized for complex tasks. Its improved adherence to structured instructions (such as JSON**** schemas) significantly simplifies the creation of agents and automated workflows.

Wan-video: wan-2.1-1.3b

​

A model specialized in high-quality video generation. It supports multiple tasks: Text-to-Video, Image-to-Video, video editing, and even Text-to-Image and Video-to-Audio generation.

Anthropic: Claude Opus 4

​

Ultra-high-performance coding model. It integrates long-term planning and autonomous reasoning capabilities, making it ideal for use cases requiring sustained effort.

black forest labs : flux 

A suite of revolutionary image generation models, offering realism and exceptional detail that often surpass similar models.

OpenAI: o3

​

A highly versatile model, very strong in mathematics. Its integrated visual reasoning capabilities allow it to combine visual and textual analysis directly within its chain of reasoning.

OpenAI: GPT-4.1 Mini

​

An intermediate model balancing performance and cost. Optimized for fast and economical interactions, it excels at processing long documents and retrieving specific information.

OpenAI: GPT-4.5

​

  A conversational model focused on fluidity. Unlike chain-of-thought reasoning models, it favors linguistic intuition for more natural and conversational responses. Ideal for writing tasks.

OpenAI: GPT-4o

​

Multimodal "omni" model* capable of simultaneously processing audio, vision, and text with fast response times. It excels in advanced speech and visual recognition.