Advanced Setup¶
This guide is for source builds, Bun workflows, profile-based launches, diagnostics, and explicit provider control.
Install Options¶
Option A: npm¶
npm install -g @vantagesec/socc
Option B: From Source with Bun¶
Use Bun 1.3.11 or newer for source builds.
git clone https://github.com/nilsonpmjr/socc.git socc
cd socc
bun install
bun run build
npm link
Option C: Run Directly with Bun¶
git clone https://github.com/nilsonpmjr/socc.git socc
cd socc
bun install
bun run dev
Provider Examples¶
OpenAI¶
export SOCC_USE_OPENAI=1
export OPENAI_API_KEY=sk-...
export OPENAI_MODEL=gpt-4o
Codex via ChatGPT Auth¶
codexplan maps to the higher-reasoning Codex route and codexspark maps to the faster Codex Spark route.
If you already use Codex CLI, SOCC reads ~/.codex/auth.json automatically. You can also set CODEX_AUTH_JSON_PATH or CODEX_API_KEY.
export SOCC_USE_OPENAI=1
export OPENAI_MODEL=codexplan
socc
DeepSeek¶
export SOCC_USE_OPENAI=1
export OPENAI_API_KEY=sk-...
export OPENAI_BASE_URL=https://api.deepseek.com/v1
export OPENAI_MODEL=deepseek-chat
OpenRouter¶
export SOCC_USE_OPENAI=1
export OPENAI_API_KEY=sk-or-...
export OPENAI_BASE_URL=https://openrouter.ai/api/v1
export OPENAI_MODEL=google/gemini-2.0-flash-001
Ollama¶
ollama pull llama3.3:70b
export SOCC_USE_OPENAI=1
export OPENAI_BASE_URL=http://localhost:11434/v1
export OPENAI_MODEL=llama3.3:70b
Atomic Chat¶
export SOCC_USE_OPENAI=1
export OPENAI_BASE_URL=http://127.0.0.1:1337/v1
export OPENAI_MODEL=your-model-name
LM Studio¶
export SOCC_USE_OPENAI=1
export OPENAI_BASE_URL=http://localhost:1234/v1
export OPENAI_MODEL=your-model-name
Environment Variables¶
Variable |
Required |
Description |
|---|---|---|
|
Yes |
Enables the OpenAI-compatible provider path |
|
Usually |
API key for hosted providers; not needed for some local backends |
|
Yes |
Model name such as |
|
No |
API endpoint; defaults to OpenAI if omitted |
|
Codex only |
Codex or ChatGPT access-token override |
|
Codex only |
Path to Codex CLI |
|
Codex only |
Alternate Codex home directory |
ANTHROPIC_MODEL can still override model naming in some flows, but OPENAI_MODEL takes priority on the OpenAI-compatible path.
Runtime Hardening¶
bun run smoke
bun run doctor:runtime
bun run doctor:runtime:json
bun run doctor:report
bun run hardening:check
bun run hardening:strict
Notes:
doctor:runtimefails fast on placeholder keys and missing required credentialslocal providers can run without
OPENAI_API_KEYCodex profiles validate either
CODEX_API_KEYor the Codex CLI auth file
Provider Launch Profiles¶
bun run profile:init
bun run profile:recommend -- --goal coding --benchmark
bun run profile:auto -- --goal latency
bun run profile:codex
bun run dev:profile
bun run dev:codex
bun run dev:openai
bun run dev:ollama
bun run dev:atomic-chat
Use:
profile:initfor a first-time guided profile bootstrapprofile:recommendto compare installed models by goalprofile:autoto persist the best available recommendationdev:*commands to launch from a saved or explicit provider path