Decision Gate Features & Capabilities
You’ve seen what Decision Gate enables. This page answers two practical questions: what guarantees are universal, and how Open Source differs from Enterprise.
OSS vs Enterprise
| Decision Driver | Open Source (Apache 2.0) | Enterprise (Self-Hosted) |
|---|---|---|
| Core evaluation engine (RET, tri-state, evidence modes, runpacks) | ✓ | ✓ |
| Storage backends | SQLite, in-memory, filesystem runpacks | PostgreSQL, S3-compatible storage, object lock |
| Auth & authorization | mTLS subject (local-only) | OIDC, API keys, RBAC |
| Tenancy & namespaces | — | Tenant + namespace governance |
| Audit & compliance | Runpacks | Hash-chained JSONL audit logs + export |
| Usage & quotas | — | Quotas + usage ledger + export |
| Admin & ops | CLI tooling | Admin API + lightweight HTML console, Prometheus metrics |
| Deployment | Self-hosted | Self-hosted; managed cloud limited early access |
| Support | Community (GitHub) | Integration guidance |
Guarantees
- Deterministic decisions: same spec + same evidence → same outcome
- Fail-closed evaluation: missing or untrusted evidence holds the gate
- Auditability: runpacks can be replayed offline to verify decisions
- Integrity: hashes verify artifacts; signatures and hash‑chained logs strengthen audit trails (Enterprise)
Included
- Deterministic evaluation engine with explicit requirements
- Evidence modes for fast iteration and audited verification
- Runpack export for replayable audit trails
- MCP integration + built‑in providers (
time,env,json,http)
Enterprise Adds
- Scalable storage and retention (PostgreSQL + S3 + object lock)
- Multi‑tenant auth and access control (OIDC, API keys, RBAC)
- Compliance logging and exports (hash‑chained audit logs)
- Operations tooling (admin API/console, Prometheus metrics)
- Quotas and usage enforcement
Boundaries
- Does not run tools: DG evaluates evidence; tools run elsewhere
- Does not orchestrate agents: use your agent framework; DG evaluates checkpoints
- Does not mutate external systems: evidence is read‑only
- Does not provide a generally available managed hosting service: self‑hosted by default
- Does not infer meaning: authors define conditions and comparators explicitly
Deep Dives
- RET Logic — Core evaluation semantics
- Evidence Flow + Execution Model — How evidence is sourced and evaluated
- Runpacks and Verification — Replayable audit bundles
- Built-in Providers — What ships out of the box
Learn More
- Examples — Workflows that use these guarantees
- Full Documentation — Specs, math, tooling, and reference
- Enterprise Overview — Deployment model and scale considerations