The migration e2e harness only ever used one image: it seeded the database
with the candidate build and then applied synthetic migrations on top. That
proves the migration machinery (locking, crash recovery, legacy baselining,
pooling) but never executes the real schema of release N against the real
migrations of release N+1, which is the path operators actually run.
Adds a baseline image alongside the candidate, so a test can seed with a
published release and upgrade with the build under test.
Suites:
- test_upgrade.py: the candidate applies the pending release migrations,
keys minted by the baseline release survive, and concurrent replicas
upgrade a baseline database exactly once.
- test_rolling_upgrade.py: a baseline replica keeps serving virtual-key
auth while the candidate migrates underneath it, and both releases serve
and resolve each other's keys during the overlap. This is the reported
failure: a new column on LiteLLM_VerificationToken invalidates prepared
plans on pods still running the old release, which the proxy reads
whole-row, and auth starts failing until those pods leave service.
- test_shaped_database.py: the upgrade completes and preserves rows on a
populated spend log, rather than on the empty database every other
migration test starts from.
Every upgrade assertion is gated on the candidate having actually applied
migrations the baseline had not, so a stale pin fails loudly instead of
passing on an empty delta.
CI adds two jobs to the migration_startup workflow. The baseline defaults
to a committed release pin and is overridable per pipeline, matching how
migration_candidate_image already works; only the upgrade jobs pull it.
Verified against a real v1.101.0 -> v1.102.0 upgrade: 6 passed, with the
baseline seeding 165 migrations and the candidate applying the 6 that
landed between the two releases.
The deleted wrapper put cache: {"no-cache": true} on every request body, so the
gateway's response cache never answered a re-sent prompt. The SDKs send nothing
of the sort, and the mid-conversation prompt-cache priming loop re-sends an
identical body until the provider reports a warm cache, which a cached reply
never does. NO_PROXY_CACHE in sdk_clients.py restores the field as extra_body
on every messages, responses, completions, and embeddings call.
The wrapper also returned a 4xx as a value where the SDKs raise. The Bedrock
safety_identifier test judges the captured Converse body, and Claude on Bedrock
rejects the forwarded field with a 400, so the /v1/responses leg now suppresses
openai.BadRequestError the way the chat leg carries the same 400 as a Result.
The shared settings-store work this branch was stacked on landed on main,
along with source reporting for /config/field/info and /config/list.
Resolved by taking main for the store, the config routes and their tests,
and re-applying only the read surfaces main does not cover:
/alerting/settings, /get/ui_settings and /router/settings.
Claude Code's server-side auto-mode classifier sends a `safeguards` body field
together with the `dangerous-tool-use-2026-09-03` beta. PR #42152 made the
first-party anthropic route pass them through, but the beta header mapping
left the other two Claude platforms at null, so Bedrock Invoke dropped both
(classifier silently disabled) and Vertex forwarded the body field without
the beta, which the platform rejects with "safeguards: Extra inputs are not
permitted" (a 400 Claude Code hides by retrying without them).
Map the beta for bedrock and vertex_ai in the beta headers config and add
`safeguards` to the Bedrock Invoke request allowlist so the pair reaches
both platforms unchanged. Nothing is injected: a client that sends
`safeguards` without the beta still gets the platform's 400, exactly as
api.anthropic.com answers it.
A fallback hop that dies before its first chunk surfaces inside the streaming
iterator, where the chain lookup is keyed by the hop's own group. That group has
no chain of its own, so the remaining entries of the original list were never
tried. Resume the original group's chain as the last lookup key; attempted_targets
already skips the entries that were tried.
Resolves LIT-7400
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
