HUB
Sign inSign up

teamintelligence/inference:0.1.0-dev.1

Multi-platform
Manifest digest

sha256:e396193009576920c04d787fba1015e08ce2741665980e2c70af9753afb1663a

Last pushed

3 days by ankurrustagi75524

Type

Sandbox Kit

Manifest digest

sha256:e396193009576920c04d787fba1015e08ce2741665980e2c70af9753afb1663a

MIXIN

Use Docker-hosted models with your coding agent.


Arguments
NameRequiredDefaultDescription
auth_tokenOptionalproxy-managed

Placeholder replaced by SBX's host-side credential proxy.

claude_modelsOptional[ { "model": "docker/deepseek-v4.1-flash-1017", "label": "Docker DeepSeek V4.1 Flash" }, { "model": "docker/kimi-k3.1-pro", "label": "Docker Kimi K3.1 Pro" }, { "model": "docker/glm-5.3", "label": "Docker GLM 5.3" }, { "model": "docker/glm-5.3-flash", "label": "Docker GLM 5.3 Flash" } ]

Bundled fallback choices and labels; startup refreshes the picker from the gateway.

codex_configOptionalmodel = "docker/deepseek-v4.1-flash-1017" model_provider = "docker-managed" model_reasoning_effort = "high" model_catalog_json = "/home/agent/.codex/docker-managed-models.json" web_search = "disabled" [model_providers.docker-managed] name = "Docker Inference" base_url = "https://${DOCKER_MANAGED_GATEWAY_HOST}/v1" env_key = "DOCKER_INFERENCE_TOKEN" wire_api = "responses" requires_openai_auth = false supports_websockets = false

Codex provider configuration for the managed gateway.

codex_modelsOptional[ { "id": "docker/deepseek-v4.1-flash-1017", "name": "Docker DeepSeek V4.1 Flash", "input": ["text", "image"], "context": 1048576, "reasoning": ["low", "medium", "high", "max"] }, { "id": "docker/kimi-k3.1-pro", "name": "Docker Kimi K3.1 Pro", "input": ["text", "image"], "context": 1048576, "reasoning": [] }, { "id": "docker/glm-5.3", "name": "Docker GLM 5.3", "input": ["text"], "context": 1048576, "reasoning": ["none", "minimal", "low", "medium", "high", "xhigh", "max"] }, { "id": "docker/glm-5.3-flash", "name": "Docker GLM 5.3 Flash", "input": ["text", "image"], "context": 1048576, "reasoning": ["none", "minimal", "low", "medium", "high", "xhigh", "max"] } ]

Managed models and their reviewed Responses API reasoning levels for Codex's picker.

context_tokensOptional1048576

Maximum context window; startup narrows it to the gateway models' limits.

disable_experimental_betasOptional1

Keep unqualified Anthropic beta features out of requests.

gateway_hostOptionalinference.docker.com

Gateway hostname at docker.com or a subdomain; HTTPS on port 443.

harnessOptionalauto

Harness to configure; auto uses SBX_AGENT or the only installed supported agent.

Options: auto, claude, codex, opencode, hermes
hermes_modelOptionaldocker/deepseek-v4.1-flash-1017

Initial Hermes managed model; a saved managed selection takes precedence.

hermes_modelsOptional{ "docker/deepseek-v4.1-flash-1017": { "context_length": 1048576 }, "docker/kimi-k3.1-pro": { "context_length": 1048576 }, "docker/glm-5.3": { "context_length": 1048576 }, "docker/glm-5.3-flash": { "context_length": 1048576 } }

Bundled Hermes model IDs and context lengths.

hermes_pythonOptionalauto

Python with the Hermes YAML library; auto uses the Docker Hermes virtualenv or python3.

modelOptionaldocker/deepseek-v4.1-flash-1017

Initial managed model; a saved /model choice takes precedence.

opencode_configOptional{ "$schema": "https://opencode.ai/config.json", "provider": { "docker-managed": { "npm": "@ai-sdk/openai-compatible", "name": "Docker Inference", "options": { "baseURL": "https://{env:DOCKER_MANAGED_GATEWAY_HOST}/v1", "apiKey": "proxy-managed" }, "models": { "docker/deepseek-v4.1-flash-1017": { "name": "Docker DeepSeek V4.1 Flash", "limit": { "context": 1048576, "output": 262144 }, "modalities": { "input": ["text", "image"], "output": ["text"] }, "attachment": true }, "docker/kimi-k3.1-pro": { "name": "Docker Kimi K3.1 Pro", "limit": { "context": 1048576, "output": 262144 }, "modalities": { "input": ["text", "image"], "output": ["text"] }, "attachment": true }, "docker/glm-5.3": { "name": "Docker GLM 5.3", "limit": { "context": 1048576, "output": 262144 } }, "docker/glm-5.3-flash": { "name": "Docker GLM 5.3 Flash", "limit": { "context": 1048576, "output": 262144 }, "modalities": { "input": ["text", "image"], "output": ["text"] }, "attachment": true } } } }, "model": "docker-managed/docker/deepseek-v4.1-flash-1017", "small_model": "docker-managed/docker/glm-5.3-flash" }

OpenCode provider configuration for the managed gateway.

refresh_modelsOptional1

Refresh model choices from the gateway on each sandbox boot.

small_modelOptionaldocker/glm-5.3-flash

Preferred managed model for Haiku/background tasks.


CapabilitiesExpand a row to see its full configuration.
TypeRequiredDescription
com.docker.sandbox/credential@1Required—
com.docker.sandbox/network-policy@1Required—
com.docker.sandbox/lifecycle@1Required—

Apply this mixin to a sandbox

sbx run <agent> --kit teamintelligence/inference:0.1.0-dev.1

Make sure you have docker sbx installed

Run the following command to install sbx on your machine.

macOS
brew install docker/tap/sbx
Windows
winget install Docker.sbx
Learn more about docker sbx⁠