Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 14 additions & 0 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -24,6 +24,11 @@ its own PR.

## What This Owns

The [public repository readiness audit](docs/public-repo-readiness-2026-10-12.md)
tracks all 83 public non-fork projects, verified functional PRs, exact tested
source trees, and outstanding recording gates. It distinguishes a working
scoped workflow from an unverified complete product.

- `ecosystem.toml`: repo registry, target toolchains, storage policy, expected
binaries, and verification commands.
- `AGENTS.md`: migration rules for coding agents.
Expand Down Expand Up @@ -60,6 +65,10 @@ After the migration queues are complete, every repo should agree on:

## Commands

Check out the product repositories next to `ecosystem`. For CI or isolated
worktrees, set `ECOSYSTEM_REPOS_ROOT` to the directory containing those product
checkouts. The control-plane manifest is always read from this repository.

```sh
scripts/ecosystem-pipeline.sh report
scripts/ecosystem-pipeline.sh gate
Expand All @@ -74,3 +83,8 @@ scripts/ecosystem-smoke.sh
`check-ecosystem-uniformity.py` is expected to fail until the repo-local
migration tasks are completed. That failure is useful: it is the remaining work
list made executable.

The smoke gate executes scenario checks; listing their names is not verification.
It currently fails because live cross-product scenarios are not wired. The same
pipeline correction is tracked separately in [PR #4](https://github.com/jadenfix/ecosystem/pull/4).
A green metadata check must not be used to authorize a live product demo.
33 changes: 33 additions & 0 deletions docs/demo-readiness.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,33 @@
# Functional work and demo recording handoff

Each project is recorded from its tested code commit, after its functional work
is complete. Recheck the commit when recording begins; changes after a test run
invalidate that recording handoff.

The functional owner supplies:

- Repository, branch, full commit SHA, and PR link.
- A reproducible install/start/reset command and prerequisites.
- Verification commands and their actual exit statuses. Separate unit tests,
contract tests, frontend builds, and live end-to-end checks.
- A complete core workflow with its expected result and persistence behavior.
- Known unfinished features, simulated data, hardware requirements, external
credentials, and checks that could not run.

The media owner confirms the working tree matches that commit before filming.
Use local or seeded data for demos. Label mock/paper/replay output where it is
shown, and demonstrate the actual interface for projects with a frontend.
Verify the finished GIF/video represents the tested workflow and link it from
the project's README after that verification.

Empty repositories and historical coursework are tracked separately from
runnable applications. A visual concept is not proof that a product exists.
Failed tests or unimplemented advertised functionality keep that workflow on
hold until the code is fixed and retested. Partial test coverage only authorizes
the specific verified lane, not an untested cloud, hardware, or live API lane.

`scripts/ecosystem-smoke.sh` now fails while live cross-product scenario
execution is unwired. The pipeline correction is coordinated in
[PR #4](https://github.com/jadenfix/ecosystem/pull/4).
`run-ecosystem-e2e.py --list` remains an inventory command.
Neither a scenario list nor metadata checks establish end-to-end readiness.
13 changes: 13 additions & 0 deletions docs/e2e-execution.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,13 @@
# Executable capability gates

`--list` reports the inventory and never verifies an integration. The pipeline E2E stage now executes declared `e2e_command` arrays from each owning sibling checkout. Missing commands, missing checkouts, nonzero exits, timeouts, dirty worktrees, and a changed HEAD fail the gate.

```sh
ECOSYSTEM_REPOS_ROOT=/path/to/products python3 scripts/run-ecosystem-e2e.py
ECOSYSTEM_REPOS_ROOT=/path/to/products python3 scripts/run-ecosystem-e2e.py --capability memory --timeout 600
python3 -m unittest discover -s tests -v
```

Each selected scenario writes its command, owner HEAD, dirty status, exit status, and bounded log tail to `artifacts/e2e-receipt.json`. A selected subset receipt is scoped to that subset and does not claim ecosystem completion. Keep source checkouts clean and use `CARGO_TARGET_DIR` outside them when needed.

The memory scenario invokes the real `remi/scripts/e2e-memory.sh` CLI/HTTP/SQLite workflow. Five additional owner workflows are registered: Palette product-loop trace/MCP/SDK/compose gate (Docker required, skips disabled); Tempo exact live-CDP CI stages (real browser required, missing/skipped test assertions rejected); Cradle native Wasm CLI/API/MCP/SQLite; temp.js authenticated remote MCP bridge and journal; tempOS token-gated admission/execution/receipt verification. Each requires its real native runtime/build dependencies. Settlement remains blocked until a transaction → receipt → indexer workflow exists; validator or mocked SDK tests do not substitute for that intent. Controller tests exercise real subprocesses in a tiny committed fixture repo; they verify gating behavior and are not product E2E evidence.
Loading