Marc-André Bombeck dfe34386be fix(ai): read a Codex sign-in's memory only under its own account ...

Every write to the per-sign-in Codex memory checked the account digest,
but the reads did not. A credential reconnected to another ChatGPT
account therefore tried the old account's working model first, sorted
by its refusals and trusted its model list, until the first write
rebound the entry, and the vision gate kept answering from the old
account in the meantime.

The client reads (working model, refusals, model list) now take the
credential and treat an entry learned under another account as empty.
The vision gate holds no account id, so the connect routes rebind the
entry and the disconnect routes drop it, and the gate reads the entry
as its process last bound it.

2026-10-29 21:36:33 +01:00

..

fixtures

feat(ai): pick Codex models from the account's live list, per sign-in

2026-10-29 20:17:42 +01:00

ai-budgets.test.ts

chore(release): v1.28.28 — OpenAI-compatible gateways work; a re-keyed night can no longer vanish

2026-07-11 10:36:02 +02:00

anthropic-client.test.ts

feat(ai): ask Anthropic models to think in the dialect their generation takes

2026-10-06 19:26:29 +02:00

anthropic-reasoning.test.ts

fix(ai): keep tools on the forced final round and narrow reasoning-dialect learning

2026-10-06 21:50:19 +02:00

briefing-grounding.test.ts

chore: strip planning tags from comments and enable the rule

2026-10-20 09:08:40 +02:00

cached-input-usage.test.ts

feat(ai): pick Codex models from the account's live list, per sign-in

2026-10-29 20:17:42 +01:00

citation-coverage-logging.test.ts

feat(i18n): register Brazilian Portuguese as a locale

2026-10-22 14:50:00 +02:00

citation-enforcement.test.ts

chore: strip planning tags from comments and enable the rule

2026-10-20 09:08:40 +02:00

coach-prompt-v423.test.ts

chore: strip planning tags from comments and enable the rule

2026-10-20 09:08:40 +02:00

codex-client.test.ts

feat(ai): pick Codex models from the account's live list, per sign-in

2026-10-29 20:17:42 +01:00

codex-live-models.test.ts

fix(ai): read a Codex sign-in's memory only under its own account

2026-10-29 21:36:33 +01:00

codex-model-catalog.test.ts

fix(ai): read a Codex sign-in's memory only under its own account

2026-10-29 21:36:33 +01:00

codex-oauth.test.ts

fix(ai): serialize Codex refreshes so a lost race never parks a live sign-in

2026-10-21 00:06:21 +02:00

codex-reasoning.test.ts

feat(ai): pick Codex models from the account's live list, per sign-in

2026-10-29 20:17:42 +01:00

codex-sign-in-routes.test.ts

fix(ai): read a Codex sign-in's memory only under its own account

2026-10-29 21:36:33 +01:00

codex-slug-cache.test.ts

fix(ai): read a Codex sign-in's memory only under its own account

2026-10-29 21:36:33 +01:00

codex-slug-chain.test.ts

fix(ai): read a Codex sign-in's memory only under its own account

2026-10-29 21:36:33 +01:00

codex-slug-fallback.test.ts

fix(ai): read a Codex sign-in's memory only under its own account

2026-10-29 21:36:33 +01:00

codex-timeout.test.ts

feat(ai): pick Codex models from the account's live list, per sign-in

2026-10-29 20:17:42 +01:00

confidence-schema.test.ts

chore: strip planning tags from comments and enable the rule

2026-10-20 09:08:40 +02:00

consent-guard.test.ts

refactor(ai): re-check every chokepoint through the one egress helper

2026-09-24 09:29:00 +02:00

daily-briefing-schema.test.ts

chore: strip planning tags from comments and enable the rule

2026-10-20 09:08:40 +02:00

effective-timeout.test.ts

fix(insights): wait as long as the response timeout allows in the browser

2026-10-22 13:01:20 +02:00

entry-effort-refused.test.ts

fix(ai): learn a refused reasoning field set by the entry itself

2026-10-25 01:01:02 +02:00

error-body-classification.test.ts

Surface what an OpenAI-compatible gateway actually refused

2026-09-05 17:56:47 +02:00

failure-kind.test.ts

fix(ai): ask a thinking model again when it answered with reasoning alone

2026-10-22 13:01:10 +02:00

feedback-attribution.test.ts

feat(api): /api/insights/feedback persists thumbs with provider attribution

2026-05-10 01:30:06 +02:00

gateway-embedded-error.test.ts

fix(ai): every surface keeps a ceiling on the day's mixed total

2026-09-11 11:11:01 +02:00

gateway-unswitchable-reasoning.test.ts

fix(ai): give background work the High room on a model that always reasons

2026-10-29 09:07:37 +01:00

insight-generator-prompt.test.ts

chore: strip planning tags from comments and enable the rule

2026-10-20 09:08:40 +02:00

insight-interpretation.test.ts

chore: strip planning tags from comments and enable the rule

2026-10-20 09:08:40 +02:00

insight-response-schema.test.ts

Delete the citation-enforcement chain that never ran

2026-07-19 05:50:57 +02:00

json-dialect.test.ts

fix(ai): keep JSON mode when an endpoint rejects only reasoning_effort

2026-10-04 10:48:35 +02:00

json-extract.test.ts

fix(ai): stop prefilling the assistant turn on Anthropic JSON calls

2026-09-02 10:28:05 +02:00

legacy-payload-detection.test.ts

chore: strip planning tags from comments and enable the rule

2026-10-20 09:08:40 +02:00

local-client-egress.test.ts

fix(security): keep ALLOW_LOCAL_AI_PRIVATE_HOSTS=true for the operator's own AI, deprecated

2026-09-26 20:59:10 +02:00

local-client.test.ts

fix(security): grant private AI endpoints by exact origin, dialled pinned

2026-09-26 19:46:11 +02:00

local-gemma-wire.test.ts

fix(ai): ask a thinking model again when it answered with reasoning alone

2026-10-22 13:01:10 +02:00

local-host-allowlist.test.ts

fix(security): keep ALLOW_LOCAL_AI_PRIVATE_HOSTS=true for the operator's own AI, deprecated

2026-09-26 20:59:10 +02:00

local-reasoning.test.ts

fix(ai): hold a refused reasoning field for a day

2026-10-29 19:39:38 +01:00

medical-reference-prompt.test.ts

chore: strip planning tags from comments and enable the rule

2026-10-20 09:08:40 +02:00

medical-references.test.ts

chore: strip planning tags from comments and enable the rule

2026-10-20 09:08:40 +02:00

model-availability.test.ts

fix(ai): name the real cause of failed briefings and stop retrying what cannot heal

2026-10-08 09:06:09 +02:00

no-default-positivity-opener.test.ts

chore: strip planning tags from comments and enable the rule

2026-10-20 09:08:40 +02:00

openai-client.test.ts

fix(ai): recognise the gpt-6 model family in every model gate

2026-10-24 10:11:16 +02:00

openai-compatible-client.test.ts

Let the operator provider reach an allowlisted private host

2026-08-27 08:48:00 +02:00

openai-compatible-provider.test.ts

fix: preserve import job wipe ownership

2026-08-06 02:11:01 +02:00

openai-model-family.test.ts

fix(ai): recognise the gpt-6 model family in every model gate

2026-10-24 10:11:16 +02:00

openai-reasoning.test.ts

fix(ai): hold a refused reasoning field for a day

2026-10-29 19:39:38 +01:00

provider-admin-host.test.ts

fix(ai): pick the admin client from the configured base URL

2026-09-02 16:17:10 +02:00

provider-chain.test.ts

feat(ai): add operator-shared central Codex connection with per-user opt-in

2026-07-10 00:10:56 +02:00

provider-contract.test.ts

feat(ai): pick Codex models from the account's live list, per sign-in

2026-10-29 20:17:42 +01:00

provider-health-ledger.test.ts

chore: strip planning tags from comments and enable the rule

2026-10-20 09:08:40 +02:00

provider-legacy-private-host.test.ts

fix(security): keep ALLOW_LOCAL_AI_PRIVATE_HOSTS=true for the operator's own AI, deprecated

2026-09-26 20:59:10 +02:00

provider-presence.test.ts

fix(consent): mint AI consent receipts only on an explicit agreement

2026-10-16 17:11:04 +02:00

provider-reasoning-effort.test.ts

feat(ai): send a per-entry reasoning effort to Local and gateway providers

2026-10-04 10:42:59 +02:00

provider-response-timeout.test.ts

fix(ai): run every model call under the response-timeout setting

2026-10-02 23:01:36 +02:00

provider-runner-background-bench.test.ts

fix(ai): give background work the High room on a model that always reasons

2026-10-29 09:07:37 +01:00

provider-runner.test.ts

chore: strip planning tags from comments and enable the rule

2026-10-20 09:08:40 +02:00

provider-state-never-persisted.test.ts

test(ai): keep provider state out of logs and storage

2026-10-06 19:26:30 +02:00

provider.test.ts

feat(ai): one provider presence probe

2026-09-24 07:39:56 +02:00

rationale-prompt.test.ts

chore: strip planning tags from comments and enable the rule

2026-10-20 09:08:40 +02:00

rationale-schema.test.ts

chore: strip planning tags from comments and enable the rule

2026-10-20 09:08:40 +02:00

reasoning-effort.test.ts

feat(ai): send a per-entry reasoning effort to Local and gateway providers

2026-10-04 10:42:59 +02:00

reasoning-passthrough.test.ts

feat(ai): reason deterministically in the mock and count reasoning in the chain runner

2026-10-06 19:26:30 +02:00

reference-id-schema.test.ts

chore: strip planning tags from comments and enable the rule

2026-10-20 09:08:40 +02:00

request-hops.test.ts

fix(ai): bound one completion call to two requests

2026-10-25 00:11:42 +02:00

resolve-chain.test.ts

style(prettier): sweep formatting across the tree

2026-05-10 11:05:40 +02:00

response-timeout-binding.test.ts

feat(ai): pick Codex models from the account's live list, per sign-in

2026-10-29 20:17:42 +01:00

retry-correction.test.ts

fix(ai): align the corrective retry prompt with the validator that grades the reply

2026-07-19 11:01:57 +02:00

server-provider-health.test.ts

docs: comments name what the code does, not how it was built

2026-09-11 11:03:04 +02:00

trend-annotations-schema.test.ts

chore: strip planning tags from comments and enable the rule

2026-10-20 09:08:40 +02:00

truncated-plain-answer.test.ts

fix(ai): keep a cut answer from a model never seen reasoning

2026-10-25 00:20:35 +02:00

usage-level.test.ts

fix(ai-usage): never regenerate more often on frugal than on standard

2026-10-16 05:38:36 +02:00

vision-capability.test.ts

fix(ai): recognise the gpt-6 model family in every model gate

2026-10-24 10:11:16 +02:00