Skip to content

Commit 8f50694

Browse files
committed
Simplify fake Responses test surface
1 parent 2f3bf21 commit 8f50694

13 files changed

Lines changed: 42 additions & 632 deletions

.github/workflows/ci.yml

Lines changed: 1 addition & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -69,13 +69,7 @@ jobs:
6969
- name: Validate default ChatGPT/Codex Envoy config
7070
run: ./test/validate-openai-config.sh
7171

72-
- name: Run narrow fake Responses proxy test
73-
run: ./test/e2e-openai-responses.sh
74-
75-
- name: Run fake observability validation
76-
run: ./test/observability-openai-responses.sh
77-
78-
- name: Run broad fake Responses regression
72+
- name: Run fake Responses regression
7973
run: ./test/e2e-openai-responses-full.sh
8074

8175
- name: Capture docker compose state on failure

AGENTS.md

Lines changed: 1 addition & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -138,10 +138,6 @@ overrides them.
138138
- For Rust changes, run targeted tests or `cargo test` when feasible.
139139
- Static default config validation:
140140
`./test/validate-openai-config.sh`
141-
- Narrow fake Responses proxy test:
142-
`./test/e2e-openai-responses.sh`
143-
- Fake observability validation:
144-
`./test/observability-openai-responses.sh`
145-
- Broader fake regression before real smoke:
141+
- Fake Responses regression before real smoke:
146142
`./test/e2e-openai-responses-full.sh`
147143
- If a relevant test is skipped, state why in the final response.

README.md

Lines changed: 0 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -136,8 +136,6 @@ Local fake tests:
136136

137137
```bash
138138
./test/validate-openai-config.sh
139-
./test/e2e-openai-responses.sh
140-
./test/observability-openai-responses.sh
141139
./test/e2e-openai-responses-full.sh
142140
```
143141

codex-blackbox-cli/src/main.rs

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -780,7 +780,7 @@ fn codex_stack_config_available() -> Option<bool> {
780780

781781
fn fake_openai_e2e_available() -> Option<bool> {
782782
[
783-
"test/e2e-openai-responses.sh",
783+
"test/e2e-openai-responses-full.sh",
784784
"test/fake-openai.py",
785785
"test/docker-compose.openai-responses.yml",
786786
"test/envoy.openai-responses.e2e.yaml",
@@ -944,7 +944,7 @@ async fn run_doctor() -> i32 {
944944
print_check("⚠", "fake OpenAI Responses e2e files missing");
945945
push_unique(
946946
&mut fixes,
947-
"Run from the Codex Blackbox repository if you need `./test/e2e-openai-responses.sh`.",
947+
"Run from the Codex Blackbox repository if you need `./test/e2e-openai-responses-full.sh`.",
948948
);
949949
}
950950
None => {

docs/reference/developing.md

Lines changed: 1 addition & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -10,8 +10,6 @@ cargo fmt --check
1010
cargo test --workspace
1111
cargo clippy --workspace -- -D warnings
1212
./test/validate-openai-config.sh
13-
./test/e2e-openai-responses.sh
14-
./test/observability-openai-responses.sh
1513
./test/e2e-openai-responses-full.sh
1614
```
1715

@@ -32,6 +30,7 @@ codex-blackbox postmortem last
3230
Release-facing claims require a real Codex model turn observed by
3331
`codex-blackbox-core` with `provider="codex_responses"`. Fake fixture tests
3432
validate local parser, persistence, API, watch, and dashboard contracts only.
33+
Keep the fake regression in CI, but do not describe it as live support proof.
3534

3635
Before changing behavior, read the repository `AGENTS.md` and the module map in
3736
that file. The highest-risk paths are request parsing, Responses SSE

test/README.md

Lines changed: 0 additions & 108 deletions
This file was deleted.

test/docker-compose.openai-responses.yml

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,4 @@
1-
# Phase 5A fake OpenAI Responses E2E override.
1+
# Fake OpenAI Responses E2E override.
22
# Routes only local test traffic to test/fake-openai.py; production Envoy
33
# remains unchanged and real OpenAI/Codex is not contacted.
44

test/dogfood-codex-sessions.sh

Lines changed: 9 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -4,7 +4,7 @@
44
# In --mode real this launches real Codex sessions through Codex Blackbox's
55
# ChatGPT/Codex subscription proxy path. It uses the existing local Codex
66
# ChatGPT login, can contact chatgpt.com, and must not edit ~/.codex/config.toml.
7-
# Run the fake Phase 9A gate first:
7+
# Run the fake regression gate first:
88
#
99
# ./test/e2e-openai-responses-full.sh
1010
#
@@ -40,7 +40,7 @@ usage() {
4040
Usage: ./test/dogfood-codex-sessions.sh [options]
4141
4242
Options:
43-
--mode real|fixture real launches Codex; fixture delegates to Phase 9A fake e2e
43+
--mode real|fixture real launches Codex; fixture delegates to the fake regression
4444
--sessions N number of sessions to launch in real mode, 1-4 (default: 4)
4545
--repos same|mixed run only the current repo or include another repo (default: mixed)
4646
--include-mcp include an MCP-oriented prompt when MCP config exists
@@ -716,9 +716,9 @@ for cmd in docker curl python3 "$CODEX_BIN"; do
716716
done
717717

718718
if [ "$MODE" = "fixture" ]; then
719-
record skipped "real_codex_sessions" "--mode fixture delegates to Phase 9A fake e2e"
720-
CODEX_BLACKBOX_FULL_E2E_REPORT_DIR="$REPORT_DIR/phase9a" ./test/e2e-openai-responses-full.sh
721-
record passed "phase9a_fake_e2e" "see $REPORT_DIR/phase9a"
719+
record skipped "real_codex_sessions" "--mode fixture delegates to fake regression"
720+
CODEX_BLACKBOX_FULL_E2E_REPORT_DIR="$REPORT_DIR/fake-regression" ./test/e2e-openai-responses-full.sh
721+
record passed "fake_responses_regression" "see $REPORT_DIR/fake-regression"
722722
finish_and_exit 0
723723
fi
724724

@@ -758,7 +758,7 @@ if [ "$STACK_ALREADY_READY" = "0" ]; then
758758
record passed "stack_preclean" "removed stale Codex Blackbox Compose services before real smoke"
759759
fi
760760

761-
SMOKE_PROMPT="$PROMPT_MARKER preflight: Read AGENTS.md and docs/remaining-phases.md, then summarize the current next phase in 3 bullets. Do not edit files."
761+
SMOKE_PROMPT="$PROMPT_MARKER preflight: Read AGENTS.md and docs/reference/developing.md, then summarize the current validation rules in 3 bullets. Do not edit files."
762762
preflight_cmd=(
763763
env -u COMPOSE_FILE CODEX_BLACKBOX_COMPOSE_FILE="$COMPOSE_PATH"
764764
"$CODEX_BLACKBOX_BIN" preflight codex-subscription --
@@ -806,7 +806,7 @@ add_case \
806806
"same-read-a" \
807807
"same" \
808808
"$SAME_REPO" \
809-
"$PROMPT_MARKER same-read-a: Read AGENTS.md and docs/remaining-phases.md. Answer with exactly three bullets about the current phase gate. Do not edit files."
809+
"$PROMPT_MARKER same-read-a: Read AGENTS.md and docs/reference/developing.md. Answer with exactly three bullets about the validation boundary. Do not edit files."
810810

811811
add_case \
812812
"same-tool-b" \
@@ -825,7 +825,7 @@ else
825825
"same-read-c" \
826826
"same" \
827827
"$SAME_REPO" \
828-
"$PROMPT_MARKER same-read-c: Read docs/codex-traffic-contract.md and summarize two current unknowns. Do not edit files."
828+
"$PROMPT_MARKER same-read-c: Read README.md and summarize two things this project observes directly. Do not edit files."
829829
fi
830830

831831
if [ "$INCLUDE_MCP" = "1" ] && [ "$MCP_CONFIGURED" = "1" ]; then
@@ -839,7 +839,7 @@ else
839839
"same-context-d" \
840840
"same" \
841841
"$SAME_REPO" \
842-
"$PROMPT_MARKER same-context-d: Read docs/automated-feedback-testing.md and report the smallest real dogfood gap still called out there. Do not edit files."
842+
"$PROMPT_MARKER same-context-d: Read docs/reference/developing.md and report the difference between fake and live evidence in one paragraph. Do not edit files."
843843
fi
844844

845845
: >"$WATCH_SSE"

0 commit comments

Comments
 (0)