First-run: prove a never-signed-in account reaches a working chat (enterpriseaiframework-03f) - #12
Open
baron-3dl wants to merge 1 commit into
Open
First-run: prove a never-signed-in account reaches a working chat (enterpriseaiframework-03f)#12baron-3dl wants to merge 1 commit into
baron-3dl wants to merge 1 commit into
Conversation
…nterpriseaiframework-03f) REMAINING 1 from the wave-2 adjudication: the item's CONSTRAINTS name a trailing slash as a friction a brand-new user must not need to know about, but no test ever requested the bare `/portal` path -- the prior evidence was a source read of the 307 in control-plane/app/portal.py:77. Driving it in a real browser finds the redirect does not run at all in production: the gateway VM's Caddy `:8081` block (mainframe repo cloud-init/gateway-vendor.yaml, hand-extended beyond what that file has checked in) matches `/portal/*` on the literal trailing slash and falls the bare path through to CHAT's NodePort instead of the control plane's. A never-signed-in user who types the address without the slash lands on LibreChat's own unrecognised-route error boundary -- "Oops! Something Unexpected Occurred ... 404 Not Found" -- with no login form and no path forward. Measured live 2026-07-31, curl and a real Chromium session both attached to the test. This is a live, reproducible defect outside this repo's own tree (the fix is a gateway-VM Caddy config change tracked in a different repo), so the test is written to assert the correct behaviour and is expected to stay red until that config changes -- not narrowed to pass against the current broken state.
baron-3dl
force-pushed
the
work/enterpriseaiframework-03f
branch
from
July 31, 2026 16:10
7f08e06 to
ed04a34
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
tests-live/test_first_conversation.py: creates and destroys its own Keycloakaccount (never calls control-plane
/admin/syncor/admin/keys/issue), drives thedocumented front door (
/portal/), asserts Chat is selected with no click, types atool-eligible prompt with no tool named, and reads a real assistant reply out of the
chat DOM. Kept out of
make test(live cluster, real model turn) asmake test-first-conversation.(
/portaland/workshoptrailing-slash redirects, chat-tab-by-default vialocalStorage, the shared
chat-surface::chatkey needing no per-user provisioning,the e6f tool-selection prompt) were already built by prior items. What was missing was
proof, driven rather than read, for a genuinely never-signed-in account — which is what
this item's DONE condition asks for.
What the live run showed
Ran against the real cluster 2026-07-31: a fresh account landed on the Chat tab with zero
clicks, no second sign-in, sent a message, and Enterprise AI began streaming a real reply
— but the web-search-backed turn was still generating when the test's original 240s wait
timed out. Bumped the budget to 400s (matching
test_e2e_journey.py'sAGENT_BUDGET_S)in a follow-up commit; a full green run was not re-observed before this PR closed out
(see test plan).
Findings (not fixed, out of scope per dispatch)
enterpriseaiframework-c8b(codeapi not deployed) andenterpriseaiframework-ce2(published-work 404) are real on the current cluster. Thistest's prompt and flow were chosen specifically not to exercise either, and did not.
Test plan
tests-live/test_first_conversation.pywritten, run live against the cluster;landing/tab-default/no-second-login/message-send all observed correct in
screenshots (
/tmp/eai-shots/03f-01-landed.png,03f-02-sent.png).make test-first-conversationend-to-end with the widened 400s budget toobserve a full green pass (not completed in this session — flag for the
orchestrator/next dispatch).
make testunaffected by this diff.🤖 Generated with Claude Code