Rebuild agent on original pooled runner

modelhub_submmit_api/README.md

@@ -6,7 +6,7 @@ It currently supports:
 - one-shot submission planning via `main.py`
 - daily batch execution via `run_daily.sh`
 - continuous queue refill via `run_poll.sh`
-- a single ModelHub token read from environment variables or `KEY.md`
+- multiple ModelHub tokens read from `KEY.md` and `KEYS.md`
 - automatic task/framework/template selection across the supported GPU catalog
 
 ## Layout
@@ -15,8 +15,8 @@ It currently supports:
 - `daily_runner.py`: daily wave orchestration
 - `poll_runner.py`: long-running queue refiller
 - `runner_common.py`: shared token / key file loading
-- `modelscope_discovery.py`: ModelScope model discovery and inspection
-- `modelhub_client.py`: ModelHub API client
+- `hf_discovery.py`: ModelScope model discovery and inspection (keeps the legacy module name)
+- `modelhub_client.py`: ModelHub API client and token-pool routing
 - `history_stats.py`: online history aggregation, ranking, and warnings
 - `template_selector.py`: template lookup and GPU normalization
 - `task_registry.py`: task-type and framework selection rules
@@ -24,11 +24,12 @@ It currently supports:
 
 ## Key Files
 
-- `KEY.md`: optional local ModelScope and ModelHub tokens
+- `KEY.md`: primary ModelScope and ModelHub tokens
+- `KEYS.md`: optional supplemental ModelHub tokens
 - `templates/public_submit/adapt_task_templates.jsonl`: public submit templates
 
-The agent platform deployment should use environment variables instead of key
-files. Set `MODELHUB_XC_TOKEN` or `XC_TOKEN` for the single submitting account.
+The runner reads both files automatically. Add more accounts by appending
+`XC_TOKEN3`, `XC_TOKEN4`, and so on to `KEYS.md`.
 
 Template lookup is also relative. The selector searches from the current working
 directory and the module directory. The primary project layout is:
@@ -72,7 +73,7 @@ bash run_poll.sh --dry-run
 
 - The runner auto-discovers all safe GPU/template combinations from the public submit catalog.
 - Each model can be submitted at most once per GPU.
-- One ModelHub account is used per deployed agent instance.
+- Multiple ModelHub tokens are pooled and used to route submissions to the account with available async capacity.
 - History stats are online-only. The local ledger is used for local accounting, but platform history is only used after the local ledger reaches the configured threshold.
 - The default history threshold is `500` records.
 
@@ -83,7 +84,7 @@ Common flags:
 - `--daily-target`: total target submissions for the day; `0` means unlimited
 - `--min-downloads`: ModelScope download floor
 - `--history-stats-threshold`: local ledger threshold before using online history stats
-- `--max-scan-models`: hard cap on scanned ModelScope models for a run (0 = auto)
+- `--max-scan-models`: hard cap on scanned HF models for a run (0 = auto)
 - `--scan-multiplier`: multiplier used for auto scan cap derivation from quota/queue capacity
 - `--read-concurrency`: concurrent HTTP reads while scanning model candidates (default 4)
 - `--max-submits-per-run`: max tasks to submit per run cycle (0 = unlimited)
@@ -95,9 +96,9 @@ Common flags:
 
 `run_poll.sh` adds:
 
-- `--poll-interval-seconds`: sleep when the account has no available slots
+- `--poll-interval-seconds`: sleep when all accounts are saturated
 - `--idle-interval-seconds`: sleep when a cycle submits nothing
-- `--max-scan-models`: hard cap on scanned ModelScope models for this cycle (0 = auto)
+- `--max-scan-models`: hard cap on scanned HF models for this cycle (0 = auto)
 - `--scan-multiplier`: multiplier used for auto scan cap derivation from quota/queue capacity
 - `--max-submits-per-run`: max tasks to submit per poll cycle (0 = unlimited)
 - `--skip-outcome-sync`: skip outcome sync before scanning