Files
ai-engineering/docs/roadmap.md
T
Nilton Constantino e65f216926 feat: support simple projects and root Git repositories
Allow registering a single Git repository at the engineering project root with --path . while retaining path safety checks.

Document Git-optional projects first and cover plain folders, root repositories, and multiple repositories in process tests and native smoke scenarios.

Validated with 23 passing tests and macOS ARM64 native smoke checks for all three layouts.
2026-11-11 16:38:18 +00:00

69 lines
3.8 KiB
Markdown

# Roadmap
## Milestone 0 — Bootstrap
Implemented: Java 25, Spring Boot 4, non-web Picocli entry point, one Gradle Kotlin DSL
project with Wrapper, Modulith boundaries, help/version smoke tests, Spring AOT,
Native Image configuration and native archive tasks. Acceptance: clean tests and JAR
build pass, CLI terminates and errors are meaningful. Native results are tracked in
the experiment log, separately from configuration claims.
## Milestone 1 — Portable engineering protocol
Implemented: project discovery/init, schema 1 configuration, linked Git/submodule
registry and default read-only policy, stable Work IDs, local artifact workspaces,
discussion history, draft DECs, evidence integrity, structural validation and stable
Markdown/JSON context. Temporary-fixture tests cover continuity, independent Works,
multiple repositories and malformed/missing/unsafe state.
Simple projects are supported directly: initialize in an existing source folder
without Git, or register a single existing Git repository using `repo add app --path .`.
No enclosing workspace or additional repositories are required. Regression coverage
includes fresh processes without Git on PATH and single-repository root discovery.
Acceptance demonstration: initialize a project with two independent Git repositories,
create a Work referencing both, append discussion, persist a draft decision and
evidence, export context, then reconstruct the same context in independent processes
using only files. No framework state or model history may be required.
Platform release targets: Linux x86_64, macOS ARM64, macOS x86_64 and Windows x86_64.
The workflow builds the three available GraalVM 25 platforms. **macOS x86_64 is blocked
by upstream GraalVM 25 support removal**, documented in the
[official release notes](https://www.graalvm.org/release-notes/JDK_25/).
Do not substitute an older Java target or label an ARM binary as Intel. A future
distribution experiment must identify and validate a maintained compatible toolchain
before enabling that native release target. CI configuration is not evidence that CI
has run; see actual local results in experiments.
## Milestone 2 — Intelligent Discussion (design only)
Introduce a replaceable AgentAdapter using the portable context contract. Begin with
deterministic fakes and persist proposed questions/alternatives/assumptions as explicit
entries. Acceptance: two different adapters continue the same Work without private
history; humans retain authority over formal commitments. No model integration yet.
Recommended next experiment: two deterministic adapters read the same schema 1 context,
one proposes a clarification and another continues from its persisted response. First
confirm the context contract is sufficient; then choose provider integrations.
## Milestone 3 — Adaptive OODA (design only)
Add observations, orientation, operational decisions and Attempts using the conceptual
contracts in protocol.md. Acceptance: experiment evidence changes the next action;
history prevents repeating an invalidated attempt without a stated new hypothesis.
No rigid Observe→Orient→Decide→Act pipeline or autonomous engine is implemented now.
## Milestone 4 — Coordinated repositories (design only)
Add per-Work permissions, independent worktrees, exact source revisions, coordinated
validation and separate commit outcomes. Acceptance: partial success in one repository
remains visible; destructive actions, push and merge require explicit authorization.
Never assume atomic commit across repositories.
## Milestone 5 — Knowledge integration and conformance (design only)
Check acceptance criteria against evidence; identify impacted canonical specs, tests,
code and documentation; expose conformance checks to agents and CI. Acceptance: a Work
can demonstrate its outcome and preserve actionable knowledge without a mandatory
Lesson model or redundant documentation trail.