ds4codex 0.1.1

Responses API to Chat Completions proxy for Codex-compatible model providers.

ds4codex

ds4codex is a small local proxy that lets Codex use DeepSeek models through a Responses-to-Chat translation layer.

The current shape is intentionally simple:

• only one human-edited config file: ~/.codex/config.toml
• one generated file for /model: ~/.codex/ds4codex-model-catalog.json
• only two CLI commands: init and run

Install

cd /Volumes/data/github/seqyuan/ds4codex
pip install .

or:

pipx install .

Commands

Initialize Codex config and model catalog:

ds4codex init --apikey sk-your-deepseek-api-key

Start the proxy:

ds4codex run

What init Writes

ds4codex init updates ~/.codex/config.toml and generates:

~/.codex/ds4codex-model-catalog.json

It writes three managed blocks into ~/.codex/config.toml:

# BEGIN DS4CODEX ROOT
model = "deepseek-v4-flash"
model_provider = "ds4codex"
model_context_window = 1048576
model_catalog_json = "/home/you/.codex/ds4codex-model-catalog.json"
# END DS4CODEX ROOT

# BEGIN DS4CODEX SETTINGS
[ds4codex]
port = 8099
target_url = "https://api.deepseek.com/v1/chat/completions"
thinking = "disabled"
# END DS4CODEX SETTINGS

# BEGIN DS4CODEX PROVIDER
[model_providers.ds4codex]
name = "DeepSeek via ds4codex"
base_url = "http://127.0.0.1:8099/v1"
wire_api = "responses"
experimental_bearer_token = "sk-your-deepseek-api-key"
# END DS4CODEX PROVIDER

Where Settings Live

The proxy now reads its local runtime settings from:

[ds4codex]
port = 8099
target_url = "https://api.deepseek.com/v1/chat/completions"
thinking = "disabled"

That means ~/.config/ds4codex/config.toml is no longer needed.

thinking is only the default when Codex does not send an explicit reasoning level.

Accepted practical values are:

• disabled
• enabled
• high
• max

Compatibility mapping is applied for Codex-style reasoning levels:

• low, medium, minimal -> high
• xhigh -> max

Why the Proxy Is Still Required

Moving settings into ~/.codex/config.toml makes configuration cleaner, but it does not remove the need for the server itself.

Codex custom providers send:

• wire_api = "responses"

DeepSeek currently documents:

• chat.completions

So the proxy is still the protocol adapter between Codex Responses requests and DeepSeek Chat Completions requests.

/model Support

ds4codex init generates a model catalog JSON and points Codex at it through model_catalog_json.

That makes /model show:

• DeepSeek V4 Flash
• DeepSeek V4 Pro

and the catalog advertises these reasoning levels:

• low
• medium
• high
• xhigh

The proxy maps those levels to DeepSeek-compatible request fields before forwarding upstream.