deepseekcode

Providers

deepseekcode defaults to DeepSeek, but the runtime can also talk to OpenAI-compatible chat-completions endpoints. Provider definitions live in .deepseek/config.toml; the active provider is selected by name.

Configuration Shape

[active]
provider = "deepseek"

[providers.deepseek]
type = "deepseek"
base_url = "https://api.deepseek.com"
env_var = "DEEPSEEK_API_KEY"
first_token_timeout_ms = 45000
chunk_stall_timeout_ms = 20000

Secrets resolve in this order: api_key, then env_var, then ~/.config/deepseekcode/secrets.toml or ~/.deepseekcode/secrets.toml. Secrets files must be mode 0600 on Unix.

Examples

DeepSeek (default)

[active]
provider = "deepseek"

[providers.deepseek]
type = "deepseek"
base_url = "https://api.deepseek.com"
env_var = "DEEPSEEK_API_KEY"
first_token_timeout_ms = 45000
chunk_stall_timeout_ms = 20000

DeepSeek enables thinking, prefix-cache accounting, and JSON mode.

OpenAI

[active]
provider = "openai"

[providers.openai]
type = "openai-compat"
base_url = "https://api.openai.com"
env_var = "OPENAI_API_KEY"
default_model = "gpt-4o"

OpenAI-compatible providers drop DeepSeek’s thinking field before sending requests.

Self-hosted vLLM

[active]
provider = "vllm"

[providers.vllm]
type = "openai-compat"
base_url = "http://127.0.0.1:8000"
api_key = "local-dev-token"
default_model = "Qwen/Qwen2.5-Coder-32B-Instruct"

For local endpoints, api_key can be a placeholder if the server only checks that a Bearer token exists.

Capabilities

• thinking: provider accepts DeepSeek V4 thinking options.
• prefix_cache: provider reports DeepSeek prompt-cache hit/miss metrics.
• json_mode: provider supports response_format = json_object.
• max_ctx: approximate maximum context window used for diagnostics.

This site is open source. Improve this page.