01 / prepare
Prepare the host
The guided command detects a missing Ollama installation on macOS, installs local dependencies, starts the service, and downloads one starter model. It does not contact TEKIZ.AI.
Community provider review
Run the host check, verify your account, and create a one-time setup code in Account. Enrollment creates a pending node. It can receive work only after capacity and security review.
Operator controls. A node serves only selected models, uses an outbound authenticated connection after enrollment, and can be removed at any time. Do not use personal OAuth subscriptions or unreviewed model sizes as Community capacity.
01 / prepare
The guided command detects a missing Ollama installation on macOS, installs local dependencies, starts the service, and downloads one starter model. It does not contact TEKIZ.AI.
02 / verify
Start with the installed chat model that fits your measured resident-memory budget. You can add a second, approved model for a specific task later; one local model remains resident by default.
03 / submit
Verified TEKIZ.AI accounts can create a short-lived host setup code in Account. Enrollment creates a pending node; capacity and security review must approve it before it can receive work.
Provider credit terms
The closed beta may issue TEKIZ.AI Credits for accepted Community work. Credits apply to Pro or Max routing first, then Community inference. Rates and payouts are not active.
Credit is tied to accepted work, not uptime. Provider rates and payouts remain under manual review.
Assignment eligibility uses recent accepted results, task fit, availability, and speed. New approved nodes may receive limited evaluation work to establish a record.
Read the draft compensation policy.Request review
This form records your request; it does not send an automatic email, create a node token, or grant customer work. If Account is not yet showing Community setup, you are still waiting for approval.
Host check
This verified command saves the public setup script, checks its SHA-256 digest, installs Ollama and jq if they are missing, pulls a local 7B starter model, and checks it. On macOS, Ollama uses Apple Silicon's Metal support. No GPU driver setup is needed.
For a constrained host, keep one resident model and choose a smaller tag. Mixture-of-Experts tags are judged by total resident weights, not active parameters.
curl -fSLo enroll_community_node.sh https://tekiz.ai/downloads/enroll_community_node.txt && curl -fSLo enroll_community_node.sh.sha256 https://tekiz.ai/downloads/enroll_community_node.sha256.txt && (shasum -a 256 -c enroll_community_node.sh.sha256 || sha256sum -c enroll_community_node.sh.sha256) && chmod 700 enroll_community_node.sh && ./enroll_community_node.sh --setup
TEKIZAI_COMMUNITY_MODELS=qwen2.5:7b-instruct \ ./enroll_community_node.sh --check
Optional task mapping
After both models pass the capacity check, map tasks to enrolled local tags. Fusion can send implementation work to the smaller model and research work to the larger model.
The map selects the requested tag; it does not keep both models in memory. A constrained node still runs one job and one resident model at a time.
TEKIZAI_COMMUNITY_MODELS=qwen2.5:7b-instruct,qwen2.5:14b-instruct \ TEKIZAI_COMMUNITY_TASK_MODEL_MAP=implementation=qwen2.5:7b-instruct,research=qwen2.5:14b-instruct \ ./enroll_community_node.sh --check
The public checker is available now. A verified Account can issue a one-time host enrollment code; enrollment never activates capacity or makes a paid listing by itself.
# Run after Account issues a one-time host enrollment code. TEKIZAI_COMMUNITY_NODE_LABEL=my-node \ TEKIZAI_COMMUNITY_MODELS=qwen2.5:7b-instruct,qwen2.5:14b-instruct \ TEKIZAI_COMMUNITY_TASK_MODEL_MAP=implementation=qwen2.5:7b-instruct,research=qwen2.5:14b-instruct \ TEKIZAI_COMMUNITY_ENROLLMENT_TOKEN=paste-token-from-account \ ./enroll_community_node.sh
Account provides the one-time enrollment code and authenticated worker instructions. The public checker remains a local-only safety check; review is required before any host can receive work.
Check enrollment in AccountKeep it safe
--check again after changing hardware, model
server configuration (Ollama or vLLM), or residency settings. A
successful check is not an instruction to run a larger model.