Address gaps found in adversarial verification of the in-flight balance
reservation:
- Early release: the reservation is now a ref-counted handle carried on the
request context. Billing tasks submitted via submit*UsageRecordTask take a
ref and return it only after RecordUsage has run; with reservation enabled
the balance cache is deducted synchronously inside the billing task, so the
reservation is released only after admission can see the deduction. Dropped
tasks return their ref immediately. The reservation is renewed every ttl/3
while the handler runs; after the handler returns it is no longer renewed,
so the post-handler hold is bounded by one TTL.
- Estimate: use the same billing model (channel mapping + BillingModelSource,
max over requested/mapped when ambiguous) and ModelPricingResolver (group
ModelPricing incl. wildcards, channel per_request/image/video/token
overrides), default/group/user rate + peak multiplier, plus image, video,
audio and search surcharge kinds. Unpriced requests fail open with a
throttled log; new billing.inflight_reservation.fail_closed_on_unpriced
(default false) rejects them instead.
- Wire images, embeddings, web/x search, alpha search, grok media generation,
grok realtime/voice, OpenAI-platform /v1/messages and ResponsesWebSocket
(session-level reservation with renewal).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Balance-mode admission only checked cached balance > threshold, so N
concurrent expensive requests could all pass and overdraw the user.
Add a Redis in-flight reservation: at admission estimate a conservative
per-request cost (input estimate + output price x max_tokens, times rate
multiplier) and atomically require cached_balance - inflight >= estimate
(first in-flight request always admitted). Reservations are released via
defer on every path and expire by TTL. Fails open on estimation / Redis
errors. Configurable via billing.inflight_reservation (enabled by default).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Keep the internal platform ID opencode_go. Product surface is OpenCode
with account_mode zen|go, model-aware routing across Chat Completions,
Responses, and Anthropic Messages, and X-OpenCode-Session derived from
prompt_cache_key / metadata.user_id so prompt cache can hit.
Unmatched models fall back to /v1/chat/completions. Related to #6742.
Resolve three conflicts from main's #6680 (supplementUnmappedOpenAIModels):
1. gateway_handler.go — keep PR's allowlist refactoring (ModelAllowlistEnabled,
writeAllowlistedModelsList, modelListingSource); main's OpenAI branch in
writeModelsList was auto-merged.
2. group_models_list.go (modify/delete) — delete file per PR intent; relocate
supplementUnmappedOpenAIModels to group_model_allowlist.go where gateway_service
and openai_codex_models_service can still call it.
3. openai_codex_models_service.go — combine main's supplementUnmappedOpenAIModels
wrapping with PR's ModelAllowlistEnabled() guard.
Also fix two auto-merged test files that still referenced the old
GroupModelsListConfig / ModelsListConfig types.
Copy cached API-key manifests before group-specific mutation, use DeepSeek
model IDs for Codex fallbacks, omit unsupported config.toml effort, and
drop wildcard mapping keys from generated catalogs.
Co-authored-by: Cursor <cursoragent@cursor.com>
Anthropic fallbacks previously reused Antigravity defaults, leaking Gemini models into Claude-only groups. Composite aliases were also emitted with generic metadata and could be scheduled to accounts that did not own the exact mapping, while generated configs could default to a model absent from the downloaded catalog.
Keep provider defaults separated, derive alias capabilities from unique upstream targets with conservative fallback, require exact mapping ownership during scheduling, filter media targets, and choose generated config defaults from the fetched catalog.
Codex treats the downloaded model catalog as a capability contract. Slug-only entries hide supported reasoning levels, can advertise invalid image inputs, and expose automatic or media-only models in the model picker. Generate complete route-aware metadata, preserve official manifests, filter non-agent models, and compute validators from the final catalog.
Codex model discovery expects a top-level models manifest, but Composite and other non-OpenAI groups either reached the OpenAI live-manifest handler or had no Codex-specific response at all. API key users also had no supported way to fetch that manifest and reference it from config.toml.
Generate minimal manifests from each group's effective model list while preserving the official OpenAI live path and the ordinary /models response. Add the Use Key flow for authenticated catalog download and model_catalog_json configuration without writing the API key into the downloaded file.
- Responses WebSocket composite whitelist stays openai+grok: CN accounts
cannot pass the WSv2 ingress transport filter and the WS HTTP bridge
has no Responses conversion for them, so admitting CN targets only
turns a clear policy rejection into a misleading no-available-account
- widen the responses/input_tokens and messages/count_tokens composite
gates to the shared openai-compatible text whitelist so CN targets get
the same local-estimate token counting as generation
- restore the Claude default-model fallback for standalone CN groups
(admin candidates and custom models list) while keeping composite
listings limited to CN account mapping keys
- refresh the scheduler bulk-rebuild comment and bucket capacity hints
for the 8-platform canonical set
Review fixes for the profit-control feature commit:
- Image intent no longer disables the profit gate. The shared /v1/responses
handler previously skipped the pricing context (and therefore the gate)
whenever the platform-wide image intent predicate matched, which includes
Codex's passive image_gen namespace declaration: any client could disable
admission control for anthropic/gemini/antigravity groups by declaring a
namespace tool in the request body. Both /v1/responses paths now always
install the token pricing context; image intent only drives capability
routing and image billing. Mixed token+image requests stay token-gated;
only dedicated media endpoints remain out of scope.
- Out-of-scope paths are now explicitly suppressed instead of implicitly
ungated: Grok media (billed by media multipliers; also prevents in-flight
video lookups from turning into spurious 404s), OpenAI-group count_tokens
(unbilled), and Live calls (duration billed) carry a suppress marker that
every install point honors, including the defensive scheduler-entry
install.
- The gate resolved during selection now travels back to handlers on the
AccountSelectionResult. The shared gateway installed the gate only on a
scheduler-local context, so handler-side post-slot terminal rechecks and
post-admission sticky binding were no-ops for anthropic/gemini/
antigravity/shared-grok requests (and for composite-routed member groups
on the OpenAI path). Handlers re-apply the carried gate via
ContextWithSelectionProfitGate before the terminal recheck and binding;
the WS acquired-selection branch gained the previously missing recheck.
- Sticky binding semantics restored for ungated traffic:
BindStickySessionAfterProfitAdmission falls back to the official eager
bind when no gate is installed (wait paths lost their only binding point
otherwise), reads the pre-existing binding at bind time only when gated
(removes the unconditional per-request Redis read the feature added to
the shared handlers), and the legacy engine's three selection-time
binding writes are skipped under a gate so a terminally vetoed account
can no longer become the new sticky target.
- Responses WS connections re-freeze pricingAt and re-resolve the gate at
every turn (BeforeTurn) and bill each turn with its own instant, closing
the connect-at-valley/bill-at-valley window; a turn that fails the
recheck closes the connection so the client reselects on reconnect.
- Terminal recheck no longer swaps a DB-fresh selected account for a
staler snapshot object (UpdatedAt guard), and observer counters are
documented as per-evaluation.
- Migrations renumbered 191/192 -> 192/193 after the passkey migration
landed upstream as 191.
New regressions: selection-carried gate propagation (control group proves
the pre-fix no-op), image intent not disabling the shared gate, eager
binding fallback without a gate (both services), gated
read-failure/sentinel-miss binding semantics, legacy-engine deferred
binding under a gate with eager behavior preserved ungated, turn-level
pricing refresh (config re-resolution, scheduled-group precedence,
suppress, mid-connection disable), and suppress-marker coverage for the
request pricing context.
Group pricing (rate multiplier, peak windows, per-user overrides) and
account cost (accounts.rate_multiplier) already live side by side, but
nothing stops the scheduler from handing a request to an account whose
cost multiplier exceeds what the group's pricing can profitably serve.
Add an opt-in per-group profit gate that filters scheduling candidates
by a margin rule, while ordering, scoring, stickiness and breakers keep
working unchanged among qualified accounts.
Admission rule: an account qualifies iff U <= D * (1 - min_margin -
safety_buffer) within a small relative epsilon, where U is
accounts.rate_multiplier (0 is legal; missing/negative/NaN/Inf are
conservatively rejected as invalid) and D is the requester's effective
downstream multiplier (user-group override ?? group default, times the
group peak factor) frozen at the request's pricing instant.
- groups gain profit_control_enabled / profit_min_margin /
profit_safety_buffer (migration 191); the durable auth-cache
invalidation trigger additionally watches the profit and pricing
columns (migration 192) so out-of-band group edits cannot leave
stale auth snapshots; GetByKeyForAuth explicitly projects the new
columns and the API-key auth snapshot version is bumped to force a
refresh of pre-existing snapshots
- request-level pricing instant: token entry points install pricingAt
into ctx; the profit threshold D and the RecordUsage peak factor
read the same instant, so one request never changes price mid-flight
across waits/retries/failover (media and unwired paths keep the
existing record-time semantics)
- the gate covers token requests on openai, anthropic, gemini, grok
and antigravity groups: OpenAI-family handlers via
WithOpenAIRequestPricingContext (responses incl. WS bridge, chat
completions, messages, embeddings, alpha search), the shared gateway
via WithGatewayTokenRequestPricing (messages, chat completions,
responses, gemini model actions); composite groups cannot enable it
directly; image/video/models/usage/count_tokens stay ungated and an
explicit image-generation intent suppresses the gate end to end
- post-slot recheck: after a slot is acquired the account is re-read
via SchedulerSnapshotService.GetAccount (scheduler cache, then DB;
only when both fail the check fails open with WARN + metric); a
vetoed account releases its slot and joins the request's exclusion
set for reselection; sticky bindings are written only after the
final check passes, and an over-threshold sticky account is skipped,
not deleted, so it comes back once its rate recovers
- sticky-session cache contract: GatewayCache.GetSessionAccountID now
returns ErrStickySessionNotFound on a miss (mapped from redis.Nil in
the repository implementation, mirroring ErrRefreshTokenNotFound) so
the profit sticky path can distinguish "no binding yet" from a real
read failure without importing the cache driver in service code
- cross-group re-entry (composite parent -> member group) resolves the
gate against the member group and clears a stale parent gate instead
of letting a foreign threshold veto accounts
- per-platform/group activity counters (installs, threshold vetoes,
invalid-rate vetoes, refresh failures) for observability
- admin UI: profit-control section on the five platforms' group forms
with percent input, validation and platform-switch reset; group
create/update/duplicate normalize and validate the config at a
single choke point
- cmd/profit-preview: offline what-if tool that replays the production
admission semantics over an exported config/account/override/model
dump, reports per-model admitted-account counts under the default
and the worst-case (lowest user override) D, and surfaces probe-sync
staleness as warnings without affecting admission
Tests: service unit coverage for gate resolution/veto/threshold
epsilon/pricing instant/suppress marker/scheduler filtering and
post-slot recheck (incl. -race on the profit surface), unit-tagged
handler slot-recheck and capability-mapping regressions, sqlmock and
real-PostgreSQL integration regressions for the GetByKeyForAuth
projection and the migration-192 trigger watch list, API contract
update, and frontend specs for the five-platform form helpers.
Fixes#5148: with aggregator upstreams (e.g. newapi) that end SSE
streams without a proper terminal event, every such request was
silently missing from usage logs and billing.
Root cause (tracked via the nested audit issue): the low-level
Anthropic SSE readers already return the partially collected usage
together with the stream error (missing terminal event, read error,
interval timeout), but Forward converted every such result to
(nil, err) and the handler returned before submitting RecordUsage.
Changes:
- Add partialStreamUsageResult: on stream errors, wrap observed usage
into a ForwardResult and return it alongside the error, for both the
regular Anthropic path and the API-key passthrough path. Invariants:
UpstreamFailoverError always keeps result=nil (failover retries are
billed as the successful attempt, never twice), and zero observed
usage returns no partial result (no phantom zero-usage records).
- Messages handler: hoist the usage submission block into a closure
shared by the success path and the new partial-result error path.
- Usage record worker pool: distinguish pool-stopped drops
(dropped_stopped) from operator-configured drop/sample overflow
drops; billing tasks now fall back to inline synchronous execution
only during the shutdown window, while explicit drop/sample overflow
semantics are preserved. Image usage keeps its mandatory fallback
for both drop kinds via the new mode.Dropped() helper.
Tests: Forward-level regressions for missing-terminal / read-error /
no-usage / failover-invariant on both paths, plus handler-level
stopped-pool sync fallback and drop-policy preservation tests.
feeeei