DeepNexus Code
DeepNexus Code is a minimalist AI coding assistant for the terminal. It is a single Python file with no third-party dependencies, works with any OpenAI-compatible endpoint, and is small enough to read in an afternoon and extend however you like.
Install
Requires Python 3.10 – 3.13.
curl -fsSL http://localhost:3000/install/code.sh | bash
The script downloads dncode into ~/.local/bin (override with DNCODE_DIR) and makes it executable. Make sure that directory is on your PATH.
Configure
export DEEPNEXUS_API_KEY="sk-dn-..."
Optional settings:
| Variable | Default | Description |
|---|---|---|
DNCODE_MODEL | deepseek-v4-pro | Model to use |
DNCODE_BASE_URL | http://localhost:3000/v1 | Any OpenAI-compatible base URL, e.g. http://localhost:11434/v1 for Ollama |
DNCODE_API_KEY | $DEEPNEXUS_API_KEY | API key for the base URL |
Run
cd your-project
dncode
DeepNexus Code · deepseek-v4-pro · ~/your-project
You > add error handling to user.py
▸ read_file user.py
▸ edit_file user.py
DeepNexus Code > Wrapped the database calls in try/except and added logging.
Run a single task non-interactively with dncode -p "find all TODO comments".
Tools
| Tool | Description |
|---|---|
read_file | Read a file, optionally a line range |
write_file | Create or overwrite a file |
edit_file | Replace an exact snippet in a file |
list_files | List files, respecting .gitignore-style ignores |
search | Search file contents with a regular expression |
run_shell | Run a shell command in the project (asks for confirmation) |
Every file change and shell command is shown before it runs. Shell commands require confirmation unless you start with --yes.
Commands
| Command | Description |
|---|---|
/model <id> | Switch model mid-session |
/save [name] / /load [name] | Save or resume a session from ~/.dncode/sessions |
/changes | List files modified in this session |
/compact | Summarise the conversation to free up context |
/clear | Start a fresh conversation |
/exit | Quit |
Long conversations are compacted automatically before they exceed the model's context window.