Fenix 0.4.0: new capabilities for my AI assistant
Fenix is a general-purpose AI assistant. Version 0.4.0 expands its capabilities with document tools and visual template selection, while security remains the priority.
Introducing Fenix 0.4.0, a new version of my general-purpose artificial intelligence assistant, designed to help me with different tasks through a single conversation.
In this update, I am demonstrating one of the capabilities I am adding: working with documents. I can now choose a design, request changes and review the result with the agent. This functionality adds to the broader scope of Fenix.
New in this version: documents and edits from chat
This version brings together tools for working with PowerPoint, Word, Excel and PDF. I can request text changes in documents and presentations, update spreadsheet cells and perform PDF operations such as rearranging pages, cropping or filling supported form fields.
Capabilities depend on the format and operation. They do not reproduce every Office feature or provide unrestricted editing of any existing PDF text. The goal is to complete specific tasks through the agent and review the results.
Choose a design before generating
Fenix includes a visual template catalog in the right-hand panel. When requesting a presentation or interface, I can select a composition before generating the content, or skip the selection and describe my own design.
The templates are created and integrated within Fenix. Choosing one does not require an external template provider.
In this demonstration, I request four slides for a fictional travel brand, choose “Culture journal” and ask for two improvements: simplifying the second slide and changing the closing message on the fourth. The agent applies the changes, which can then be reviewed in the presentation.
Real demonstration on a local instance, approximately 55 seconds long. The original interface and narration are in Spanish; Spanish and English subtitles are available. Waiting and one attempt that replied without applying the change were cut. The presentation shown belongs to Design; the video does not demonstrate automatic PPTX export.
An integrated browser and an agent-produced demo
Fenix also has its own integrated browser, allowing the agent to navigate and interact with pages within its working environment.
I also worked with an AI agent to produce this video: Codex automated real navigation, recorded the screen and added narration using the web-recording skill. The voice is synthetic, which is why it may sound a little robotic. The edit includes highlights to point out the changes.
Security is the current priority
For now, I am not publishing this version’s source code or opening a publicly accessible web application. I am focused on strengthening security before making it available to more people.
I am sharing the progress and demonstration while that work continues. I want Fenix to be useful for real tasks, with appropriate controls over the agent’s access to documents, files and browsing.
What task would you like to handle with an assistant like Fenix?