Docs · Infrastructure
Inference setup.
Every product runs its agents on APG, and APG draws inference from one of three tiers. Pick the tier that matches your privacy and cost posture, then set it with a config edit. For the business rationale, see the infrastructure overview.
Tier 1 · SF default
Route through Solutions Factory.
The default on a fresh install. APG proxies inference through Solutions Factory; you are billed per token, itemized. Nothing to procure.
$ apg config set inference.tier sf-default $ apg config set inference.sf_token $SF_INFRA_TOKEN inference: sf-default · metered + itemized · zero-data-retention proxy
Tier 2 · BYOK
Bring your own Anthropic key.
Point APG at your own Anthropic key. Anthropic bills you directly; Solutions Factory takes no inference margin.
$ apg config set inference.tier byok $ apg config set inference.anthropic_key $ANTHROPIC_API_KEY inference: byok · billing direct to Anthropic · SF margin: none
Tier 3 · Sovereign
Self-host the model on your GPU.
Install ollama, pull a 7B-or-larger model at Q4 quantization or better, and point APG at it. No prompt content leaves your network. A single GPU with 12–16 GB of VRAM handles an 8B model at Q4.
$ ollama pull llama3.1:8b-instruct-q4_K_M $ apg config set inference.tier sovereign $ apg config set inference.ollama_model llama3.1:8b-instruct-q4_K_M inference: sovereign · local ollama · nothing leaves your network
Deeper guidance — model selection, quantization trade-offs, and multi-GPU — is being polished post-launch.