[Feature Request]: Multi-Modal Asset Generation (Image/Video Providers) & Interactive Planning Mode in ZCode

Overview & Product Context

Feedback submitted by solo founder/developer mvallebr detailing key feature gaps preventing users from consolidating multiple AI subscriptions (e.g., ChatGPT Plus, Codex) entirely into the Z.ai / ZCode ecosystem.

Feature Gaps & Proposed Solutions

1. Multi-Modal Provider Integration for Asset Generation (Image & Video)

• Problem: When developing web applications or games, developers frequently require real-time image and video asset generation. Currently, ZCode provider settings focus strictly on LLM text/chat endpoints.
• Proposed Solution:
  • Expand Model Settings / Providers to support dedicated multi-modal asset generation endpoints (e.g., fal.ai, Midjourney API, Kling, Qwen-VL/Omni, CogView/CogVideo).
  • Enable in-chat agent tools to generate, preview, and place images/videos directly into the local workspace project tree.

2. Native Interactive "Planning Phase" (Zero-Quota / Structured Architecture Mode)

• Problem: In competing harnesses (such as Codex/Claude), developers can start with high-level architecture/planning mode without consuming heavy execution quotas, then seamlessly hand off the plan to implementation agents. In ZCode, users currently have to generate plan markdowns externally and manually import them.
• Proposed Solution:
  • Introduce a dedicated Plan Mode inside ZCode chats (similar to planning mode / brainstorming).
  • Allow the user and model to collaboratively structure and iterate on an implementation plan before triggering file-modifying tools.

3. General Daily Companion & Voice Capability (Broader Ecosystem Parity)

• Problem: Household and non-coding tasks (voice mode, everyday mobile chat) currently require keeping OpenAI / ChatGPT Plus subscriptions active.
• Proposed Solution:
  • Enhance chat.z.ai mobile and web apps with real-time voice mode and generalist everyday assistant capabilities to allow users and their families to fully replace third-party companion subscriptions.

Origin & Discord Reference

• Author: mvallebr
• Discord Thread: https://discord.com/channels/1346756824233148527/1550067874146816051

No repository files, tests, or implementation entry points are identified. Start by splitting the three broad proposals into separately scoped issues and use the linked Discord thread for context; progress is complete only when one capability has agreed requirements and concrete acceptance criteria.