convez.ai
AgenteIntegração & APIsIntermediário
KRoperUK

Command Code Go Pi Provider

KRoperUK/command-code-go-pi-provider

Acesso unificado a 21 modelos de IA via Command Code API

Ver no GitHub
InicialPluginScore · de 0 a 100

Provedor nativo de integração com a API Command Code para Oh My Pi, permitindo acesso a 21 modelos de IA (Claude, GPT, DeepSeek, Gemini, Qwen e outros) através de um único endpoint e chave de API. Implementa streaming SSE customizado com suporte a raciocínio, pensamento e chamadas de ferramentas.

Instalação

$npx convez-ai KRoperUK/command-code-go-pi-provider

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
1
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 2 dias
Versão
command-code-go-pi-provider-v1.1.0
Issues abertas
1
Contribuidores
2

Casos de uso

  • Usar múltiplos modelos de IA sem trocar chaves de API
  • Executar agentes Oh My Pi com seleção dinâmica de modelos
  • Monitorar uso e billing de forma centralizada
  • Explorar capacidades de raciocínio em modelos de diferentes provedores

Veredito

Solução técnica bem-construída para agregar múltiplos provedores de LLM sob uma interface única. Reduz fricção operacional em ambientes que precisam de flexibilidade de modelos, embora exija uma conta Command Code separada e adicione uma camada de abstração extra na pilha.

README

A tradução para o português deste README ainda está na fila.

command-code-go-pi-provider

CI npm license

Command Code API provider for Oh My Pi. Use Claude, GPT, Gemini, DeepSeek, Qwen, Kimi, GLM, MiniMax, Step, and other models through a single Command Code API key — all 21 models, one endpoint.

Quick Start

1. Install

omp plugin install command-code-go-pi-provider

Or from git:

omp plugin install https://github.com/KRoperUK/command-code-go-pi-provider

For local development:

git clone https://github.com/KRoperUK/command-code-go-pi-provider
cd command-code-go-pi-provider
omp plugin link .

2. Set API Key

export COMMANDCODE_API_KEY="your-key-here"

Or create ~/.commandcode/auth.json:

{ "apiKey": "your-key-here" }

3. Use

omp --model commandcode/deepseek/deepseek-v4-pro "Hello!"

Or inside a session:

/model deepseek/deepseek-v4-pro
/commandcode status
/commandcode models
/commandcode usage

Available Models

21 models across all major providers:

ProviderModels
AnthropicClaude Haiku 4.5, Claude Sonnet 4.6, Claude Opus 4.7
OpenAIGPT-5.3 Codex, GPT-5.4, GPT-5.4 Mini, GPT-5.5
DeepSeekDeepSeek V4 Flash, DeepSeek V4 Pro
GoogleGemini 3.1 Flash Lite, Gemini 3.5 Flash
QwenQwen 3.6 Max Preview, Qwen 3.6 Plus, Qwen 3.7 Max
KimiKimi K2.5, Kimi K2.6
GLMGLM-5, GLM-5.1
MiniMaxMiniMax M2.5, MiniMax M2.7
StepStep 3.5 Flash

Use omp models find commandcode or /commandcode models for the full list with context windows and reasoning support.

Commands

CommandDescription
/commandcode or /commandcode statusShow provider status and API key config
/commandcode modelsList all available models with context windows
/commandcode usageShow billing usage (requires /login commandcode)
/commandcode refresh-modelsReload the model catalog
/login commandcodeSet up API key and session token

Configuration

VariableDescription
COMMANDCODE_API_KEYCommand Code API key (required)
COMMANDCODE_PROJECT_SLUGProject identifier sent in API headers (default: oh-my-pi)

Auth files checked (in order):

  1. ~/.commandcode/auth.json
  2. ~/.pi/agent/auth.json

Auth file format:

{ "apiKey": "sk-...", "sessionToken": "..." }

Architecture

This package is a native omp extension that registers a commandcode provider via pi.registerProvider(). It implements custom SSE streaming for the Command Code /alpha/generate endpoint, available on all Command Code plans. Message conversion handles text, reasoning/thinking, and tool calls bidirectionally.

Development

bun install
bun run typecheck    # requires omp types (available when linked into omp)
bun run lint         # biome check
bun run lint:fix     # biome check --write
bun test             # run unit tests
bun run check        # lint + test

CI runs lint and test on every push and PR via GitHub Actions.

Model catalog sync

models.json drifts as Command Code adds or renames models. Refresh it from the public provider catalog:

bun run sync           # update names/context windows, append new models
bun run sync --check   # report drift without writing (runs weekly in CI)

New catalog models are appended with placeholder pricing that must be filled in manually; models missing from the catalog are reported but never deleted. Override the endpoint with CC_CATALOG_URL.

License

MIT

Ver o original no GitHub

Bom para

Relacionados