Roadmap

Documentation index

The long-term goal is users as an API: give an application a believable, stateful population of users and get back evidence about what they changed. The developer preview builds the verification core that everything else depends on.

Available now (0.1.0.dev0)

• doctor, inspect, verify, diff
• init (init tenant-isolation): generates a tenant-isolation bundle (own access, cross-tenant read/write rejection, victim read-back compare) from a small JSON config
• approve and verify --require-approval/--approved-digest: local approvals bound to the scenario digest and target origin
• dashboard: a read-only local page on 127.0.0.1:8780 for runs, series, activities, per-scenario p95 trends and diffs, with filters, auto-refresh and a light/dark theme
• prune: run retention by count or age
• Explicit phases and seeded spiky, random-burst, and sustained-burst profiles
• Declared budgets with static and post-run checks, including counted write attempts and k6 vus_max
• Assertion evidence, verdicts, lifecycle, HTML report
• Run-owned HTTP fixtures, fixtures.command setup/teardown hooks, and a per-journey fixtures.pool
• Final observations with matchers, bounded until polling for eventually consistent effects, ${planned_journeys} expected-value expressions, and a second origin from allowed_origins with headers_env secrets
• Actor auth recipes: per-actor-class HS256 JWTs passed to k6 as LT_TOKEN_<CLASS> (or, with per_identity, count distinct identities as LT_TOKENS_<CLASS>, picked with tokenFor()), redacted, with every declared credential, from kept artifacts and output
• min_overlap as a client-side overlap check: observed peak in-flight requests per operation (metrics.overlap) checked against journey min_overlap. It shows requests were in flight together, not that the server executed them concurrently
• Per-operation expected_statuses and metrics.unexpected_http_failure_rate
• Examples emit expected/actual values on failing evidence samples
• Repeated seeds (consecutive or --same-seed) with dispersion stats, and baseline/candidate comparison with an optional p95 gate and per-operation p95
• Local, reachable HTTP(S), and existing E2B sandbox targets
• Five conformance examples, each with a correct and a faulty server, built on a bundled k6 runtime helper
• up: a foreground background activity of repeated bounded verify slices, reported with mode=background and no verdict

Next

• macOS release qualification and a PyPI release

Later

• Repository discovery and assisted scenario authoring
• Population mode with persistent synthetic users across slices
• Managed sandbox provisioning

Items under Next and Later are plans, not commitments or shipped features. Open an issue to discuss priorities.