Skip to content

Tips for DeepCode and Bug fixes etc. #1

Description

@D1egs

DeepCode Improvements & Bug Fixes

Improvements

1. Auto Model Selection

I'd add an automatic model comparison system that picks the best model for the task when Auto Model is enabled.

Example:

> optimize this React component

Selected Model:
Claude Opus 4.7

Reason:
Best model for coding tasks

Or:

Selected Model:
DeepSeek R1

Reason:
Best model for reasoning and problem-solving

This would help users who don't know which model to choose and make the experience smoother.


2. Model Comparison Command

I'd add a /modelcompare command that lets users compare how different models perform on the same task.

Example:

/modelcompare

Task:
Refactor auth middleware

Claude Opus 4.7   ⭐⭐⭐⭐⭐
GPT-5             ⭐⭐⭐⭐
Gemini 3 Pro      ⭐⭐⭐⭐
DeepSeek R1       ⭐⭐⭐⭐⭐

This would make it easier to find the best model for a specific job.


3. Improve the UI/UX

For the CLI version, I'd improve the overall design and make it feel closer to Claude Code. The current CLI already looks good, but it could use more visual polish and better organization.

For the GUI version, I'd move away from the typical AI chatbot layout. Right now it feels more like a chat app than a development tool.

I'd take inspiration from tools like Antigravity or Codex and add more developer-focused UI elements, such as:

  • Easier access to project settings
  • In-app editing for deepcode.md
  • Better navigation
  • More visible actions and shortcuts
  • Dedicated buttons for common workflows

Overall, I'd make it feel more like a coding platform and less like a chatbot.


Bug Fixes

1. Tool Calls Sometimes Appear in Chat

Sometimes when asking something like:

Can you create an EXE that shows "Hi" when opened?

The response contains raw tool calls like:

<tool>{"name":"write_file",...}</tool>

Users shouldn't see internal tool calls. The agent should either generate the files directly or provide a clean explanation of what it's doing.


2. Option Selection Doesn't Work in CLI

In the CLI, if the agent generates multiple choices such as:

1. Modern UI
2. Minimal UI
3. Dashboard UI
4. Gaming UI

there's currently no way to select one of the options.

Typing 1, 2, 3, or 4 doesn't work, and pressing Enter just returns to the prompt.

The CLI should support selecting generated options directly, either by:

  • Typing the number
  • Using arrow keys and Enter
  • Clicking options (if supported)

Right now the choices are shown, but they aren't actually interactive.

Metadata

Metadata

Assignees

Labels

No labels
No labels

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions