START_HERE
Start Here
Go from zero to verified HELM in 5 minutes.
Step 1 — Run HELM (30 seconds)
git clone https://github.com/Mindburn-Labs/helm-oss.git && cd helm-oss
docker compose up -d
curl -s http://localhost:8080/healthz # → OK
Step 2 — Trigger a Tool Call
Option A: OpenAI proxy (1 line change)
import openai
client = openai.OpenAI(base_url="http://localhost:8080/v1")
# Every tool call now gets a cryptographic receipt
Option B: Build + use CLI
make build
./bin/helm doctor # Check system health
Step 3 — Run Conformance (UC-012)
# Run all 12 use cases including conformance L1/L2
make crucible
# Or run conformance directly
./bin/helm conform --level L2 --json
Expected: 12/12 use cases pass, conformance L1+L2 verified.
Step 4 — Proof Loop
# Export a deterministic EvidencePack
./bin/helm export --evidence ./data/evidence --out pack.tar
# Verify offline (air-gapped safe)
./bin/helm verify --bundle pack.tar
Step 5 — See the ProofGraph
# Health + ProofGraph timeline
curl -s http://localhost:8080/api/v1/proofgraph | jq '.nodes | length'
What Just Happened
- HELM started as a kernel with Postgres-backed ProofGraph
- Tool calls were intercepted, validated (JCS + SHA-256), and receipted
- Conformance verified L1 (structural) and L2 (temporal + checkpoint)
- EvidencePack was exported as a deterministic
.tar - Offline verify proved the pack is valid with zero network access
Every step produced signed, append-only, replayable proof.
Next Steps
- README — architecture and comparison
- Security Model — TCB, threat model, crypto chain
- Deploy your own — 3-minute DigitalOcean deploy
- SDK — Python + TypeScript client libraries
- Use Cases — UC-001 through UC-012
Having Issues?
./bin/helm doctor # Diagnoses common problems
File an issue: github.com/Mindburn-Labs/helm-oss/issues