feat: make model age admission-only

README.md

@@ -62,8 +62,7 @@ Optional tuning:
 - `MODELHUB_ARCHITECTURE_COMMUNITY_LATEST_LIMIT` default `5000`
 - `MODELHUB_ARCHITECTURE_BOOTSTRAP_WORKERS` default `8`
 - `MODELHUB_ARCHITECTURE_BOOTSTRAP_MAX_LOGS` default `0` (unlimited)
-- `MODELHUB_RECENT_MODEL_RESERVE_SLOTS` default `10` per account
-- `MODELHUB_DYNAMIC_OLD_MODEL_CLEANUP_RESERVE_SLOTS` default `5` per account
+- `MODELHUB_RECENT_MODEL_RESERVE_SLOTS` default `5` per account
 - `MODELHUB_RECENT_MODEL_DAYS` default `7`
 
 ## Adaptive GPU Strategy
@@ -171,37 +170,45 @@ same model or infer failure from historical similarity. The cleanup repeats
 every 120 poll cycles by default and writes its full evidence report to
 `.modelhub_state/queue_cleanup_latest.json`.
 
-Architecture cleanup joins each active task to the locally recorded submission
-or ledger entry to recover its exact framework and task type, then reads the
-model's `config.json`. Only an exact GPU + framework + task type + architecture
-blacklist hit can authorize cancellation. Matching waiting tasks are stopped;
+Architecture cleanup reads the live queue task level and joins local submission
+or ledger metadata when available, then reads the model's `config.json`. An
+exact GPU + framework + task type + architecture blacklist hit can authorize
+cancellation. For legacy waiting tasks whose framework is absent from the API
+and local state, every currently listed framework must have an explicit matching
+block; a partial match is never enough. Matching waiting tasks are stopped;
 running tasks remain protected and their state is rechecked again immediately
 before the stop call. Failed outcomes are synchronized every three poll cycles.
 When fixed `MODEL_NOT_SUPPORTED` text adds a new blacklist entry, a lightweight
 architecture-only cleanup runs immediately without repeating repository-size or
 model-age scans.
 
-Each account dynamically reserves its last 10 known-capacity positions for
+Each account dynamically reserves its last 5 known-capacity positions for
 models updated within seven days. If an account's discovered limit is 100, 200,
-or 500, older models stop at positions 90, 190, or 490 respectively. Old-model
+or 500, older models stop at positions 95, 195, or 495 respectively. Old-model
 reservations are made under the same account lock as capacity reservations, so
 concurrent submissions cannot enter the reserved suffix. Once every account's
 old-model allowance is exhausted, discovery is capped at the seven-day window.
 Unknown model timestamps are treated as old for new submissions. When capacity
 probing raises an account's known limit, its old-model boundary moves with it.
+If an account's active-count read fails, old-model admission is fail-closed for
+that account while other accounts remain eligible. Date-deferred candidates are
+recorded as `age_policy_skipped`, release their temporary claim, and do not enter
+failure statistics or permanent exclusions.
 
-On startup, the worker first stops deterministic OOM tasks, recalculates each
-account's surviving task order by numeric task ID, and removes
-older-than-seven-days tasks beyond that account's current limit minus 10. Later
-scheduled cleanup uses limit minus 5, retaining a small hysteresis buffer that
-avoids repeatedly over-pruning valid work. For a 100-task account the two
-boundaries are 90 and 95; for a 500-task account they are 490 and 495. Recent
-overflow tasks are always kept. Age-based cleanup applies only to tasks that are
-still waiting; running validation tasks are protected even beyond the boundary.
-ModelScope metadata failures fail closed and never trigger cancellation.
-Immediately before the age-only stop batch, task ownership, active status, and
-post-OOM queue position are checked again. Deterministic OOM cleanup may still
-stop a running task because it cannot fit the selected GPU.
+Model age is admission-only. Startup and scheduled queue cleanup never cancel a
+task because it is older than seven days, beyond a capacity-minus-five boundary,
+or above a temporarily reduced platform limit. Existing waiting and running
+tasks remain untouched by date policy. Cleanup still removes deterministic OOM
+tasks and exact learned architecture incompatibilities under their existing
+state and evidence checks, preventing the old submit/cleanup/resubmit loop.
+
+A full 12-account failure audit found that 31.25% of failures had no usable log
+and another 23.52% had only ambiguous runtime evidence. These unresolved results
+now remain visible as `unresolvedFailureCount` but do not reduce GPU/framework
+decision success rates, create model/GPU cooldowns, or open attributable-failure
+circuits. Explicit OOM, repository, context, tokenizer, model-load, operator, and
+framework-architecture evidence remains attributable; identified platform
+failures continue to use short infrastructure circuits instead.
 
 The verified capacities, safe repository-size boundaries, evidence hierarchy,
 and source links are recorded in
@@ -311,12 +318,18 @@ Version `2026.08.12.5` enriches every cleanup from the live active queue's task
 level metadata. If a legacy waiting task has no recoverable framework, cleanup
 queries the current ModelHub framework catalog and cancels it only when its
 architecture is explicitly blocked on every listed framework.
+Version `2026.08.12.6` makes model age admission-only: each account reserves its
+last five dynamic-capacity positions for seven-day models, unknown account
+counts fail closed for older candidates, and no startup or periodic cleanup can
+cancel a task by date. It also keeps unclassified/ambiguous historical failures
+neutral in GPU/framework success feedback while preserving deterministic OOM
+and architecture cleanup.
 
 ## Deploy
 
 Create a tag and submit the repository URL plus tag in "我的适配智能体".
 
 ```bash
-git tag agent-v23
-git push origin agent-v23
+git tag agent-v24
+git push origin agent-v24
 ```