Move gateway-plugin tests out of the installable tree; clean plugin scan
CI / Gateway plugin tests (push) Successful in 5m13s
CI / Kotlin tests (android host + desktop) (push) Successful in 6m43s

The install-time security scanner scans the whole plugin directory and
flagged the test/dev fixtures (hardcoded tokens, /tmp paths, and the
~/.hermes/.env literal in setup.py) as DANGEROUS, blocking installs with
"19 findings".

- Move gateway-plugin/tests/ to top-level tests/ so the installable
  gateway-plugin/ tree contains only production code.
- Update _plugin_dir() in the tests and REPO in e2e.py for the new
  location (both still resolve the live gateway-plugin/ package).
- Update all references: docs, CI-SETUP.md, Gitea workflows, .pi-lens.json.
- Build the hermes .env path at runtime in setup.py via get_hermes_home()
  so the scanner no longer matches the literal ~/.hermes/.env.

Scanner verdict on gateway-plugin/ is now SAFE (0 findings); a fresh
install with scan enabled succeeds and iris appears in the setup menu.
This commit is contained in:
ARIA committed 2026-08-25 13:26:12 +02:00
1 parent 573291fc1e
commit 6f339330c5
14 files changed
+809 -311

No files matched your search

+84
View File
@@ -0,0 +1,84 @@
# Tests for the iris gateway plugin
Run via hermes's hermetic runner (never bare pytest)::
scripts/run_tests.sh tests/gateway/test_android.py
See ``docs/13-testing.md`` for the scenario list.
## WS probe (`ws_probe.py`)
Manual test-client harness: connects to the **real running gateway** and
drives a turn, printing every frame. Run with the hermes venv python
(needs `websockets`); the gateway must already be up::
hermes-agent/.venv/bin/python tests/ws_probe.py \
--token <IRIS_TOKEN> --send "hello"
Beyond the base modes (`--send`, `--upload`, `--pull-offer`, `--sync`,
`--fcm-token`/`--fcm-reg`, `--authfail`, `--url`, `--token`, `--device`,
`--timeout`), the probe has assertion and request modes:
- `--assert-turn` — assert the turn produced `message.start` → ≥1
`message.update` → `message.stop` (scenario 2).
- `--assert-reasoning` — assert the final `message.stop` carries a
non-empty `reasoning` field (scenario 3).
- `--assert-tools` — assert ≥1 `tool.start` with a matching `tool.end`
(matched by `index`; scenario 4).
- `--assert-commentary` — assert ≥1 `commentary` frame (scenario 5).
- `--assert-read-receipt` — assert a `read.receipt` frame arrives after
the sent message (new M7 frame; requires `--send`). **SKIPs** (exit 0,
prints `== SKIP: …`) when the frame never arrives, e.g. against a
gateway that predates the M7 frames.
- `--assert-status` — assert a `status` frame is received (new M7 frame;
**SKIPs** when absent).
- `--search QUERY [--scope all|chat] [--chat-id C]` — send a `search`
frame (`{query, scope, limit}`) and assert ≥1 hit in `search.results`
(scenario 8). With `--send`, the turn is driven first, then the search.
- `--channel-create NAME` / `--channel-delete CHAT_ID` /
`--channel-list` — M3 channel directory management; create prints
`== channel created: <chat_id>` for scripting.
- `--watch CHAT_ID` — wait up to `--timeout` for a message to land in
`CHAT_ID` (cron delivery E2E, scenario 7).
- `--offer-grace S` — with `--pull-offer`, keep listening S seconds after
the final message for a `media.offer` (offers are emitted post-turn,
right after the final; default 15).
- `--http [--http-url http://host:port]` — docs/19: drive the turn over
the **HTTP fallback leg** instead of WS: `GET /v1/health`,
`POST /v1/frame` (the `message.send`), receive over SSE `GET
/v1/events`. The same assertion flags apply. The base URL defaults to
the `--url` host with scheme `ws(s)` → `http(s)` and port 8791
(`IRIS_HTTP_PORT`).
Exit codes: `0` ok (incl. SKIP for absent M7 frames), `2` connect fail,
`3` no hello.ack, `4` expected hello.ack, `5` authfail expected but
acked, `6` timeout, `7` no final message, `8` upload/sync fail,
`9` pull fail, `10` assert-turn fail, `11` assert-reasoning fail,
`12` assert-tools fail, `13` assert-commentary fail, `14` search fail
(error or zero hits), `15` channel.create/list fail, `16` channel.delete
fail, `17` watch timeout, `18` read.receipt arrived before the sent
message, `19` status frame with empty payload, `20` `--http` health
check failed, `21` `--http` SSE open failed, `22` `--http`
`POST /v1/frame` rejected (4xx).
## E2E driver (`e2e.py`)
Runs the `docs/13-testing.md` §13.4 scenarios 1–12 automated-where-
possible against the live gateway, invoking `ws_probe.py` (and the
`hermes` CLI for cron) as subprocesses. Prints PASS / PARTIAL / SKIP /
FAIL per scenario plus a summary table; exits 0 if no FAIL, 1 otherwise::
hermes-agent/.venv/bin/python tests/e2e.py
hermes-agent/.venv/bin/python tests/e2e.py --skip 3,5,7
hermes-agent/.venv/bin/python tests/e2e.py --url http://host:8791
The token is read from `$IRIS_TOKEN`, else `hermes-agent/.env`, else
`~/.hermes/.env`. The gateway must already be running (the driver never
starts or stops it). It is idempotent: channels/jobs it creates are
cleaned up even on failure, and leftover `e2e-*` channels/jobs from
earlier runs are removed at start.
Scenario notes: 3 (reasoning) and 5 (commentary) are model-dependent and
SKIP rather than FAIL when the current model does not emit them; 11
(push) and 12 (reconnect/sync) are PARTIAL by design — the WS leg is
automated, the device-notification / gateway-kill leg is manual.