Skip to content

SCRIPTS.md

Latest commit

 

History

History
140 lines (110 loc) · 5.29 KB

SCRIPTS.md

File metadata and controls

140 lines (110 loc) · 5.29 KB
Raw

Scripts Reference

Last updated: 2026-05-12

Naming Conventions

Prefix/Pattern Role
vla-* VLA pipeline (robotics/embodied AI)
ai-app-* or ai-* AI agent / app pipeline
prep-<pipeline>.py Phase 1: collect + filter candidates
run-<pipeline>-two-phase.py Phase 2: LLM agent runner (two-phase)
post-<pipeline>.py Phase 3: deterministic post-processor / TG push
_*.py Private shared module (imported by other scripts)
gh-*.py GitHub Contents API operations

VLA Pipeline

vla-rss-collect.py          → vla-rss-YYYY-MM-DD.json
rate-vla-daily.py           → vla-daily-rating-in/out-YYYY-MM-DD.json (Phase 1.5)
run-vla-daily-two-phase.py  → vla-daily-hotspots.json + TG push

prep-vla-theory.py  → agent → post-vla-theory.py   (theory dissection)
prep-vla-social.py  → run-vla-social-two-phase.py → post-vla-social.py
prep-vla-sota.py    → run-vla-sota-two-phase.py   → post-vla-sota.py
prep-vla-release.py → run-vla-release-two-phase.py → post-vla-release.py
prep-vla-weekly.py  → run-vla-weekly-two-phase.py → post-vla-weekly.py

vla-daily-rerate-push.py    (re-push with ratings after keyword-fallback day)
vla-trend-snapshot.py       (weekly trend snapshot for quality review)
backfill-vla-history.py     (one-off historical backfill utility)
_vla_expert.py              (shared: get_api_key, call_qwen, fetch_handbook_context)

AI Agent Pipeline

ai-app-rss-collect.py            → ai-app-rss-YYYY-MM-DD.json
ai-daily-pick-collect.py         → ai-daily-pick-sources-YYYY-MM-DD.json

prep-ai-app-rss-filtered.py → write-ai-app-daily.py   (tool news / 日报)
prep-ai-app-social.py       → run-ai-app-social-two-phase.py  (社交情报)
prep-ai-app-workflow.py     → run-ai-app-workflow-two-phase.py  (工作流灵感)
prep-ai-deep-dive.py        → agent → post-ai-deep-dive.py  (深度解析)
prep-ai-weekly.py           → run-ai-weekly-two-phase.py → post-ai-weekly.py

post-ai-app-daily.py        (deterministic daily stats → ai-app-daily-stats.json)
prep-ai-app-dedup.py        (shared dedup helper)

Calibration and Quality

prep-calibration-check.py   (daily 11:00 assumption scan)
monthly-calibration-agg.py  (monthly aggregation + confidence update)
quality-drift-check.py      (7-day rolling baseline + 30-day sustained-decay)
ai-field-state.py           (mechanical zero-LLM trigger gate; 6 trigger types →
                             memory/field-state-YYYY-MM-DD.json)
cross-domain-rule-engine.py (v2: 7 built-in rules R001-R007 + LLM significance →
                             memory/cross-domain-insight.json)
entity-tracker.py           (90-day rolling index of authors/labs/methods/benchmarks)
upstream-signal-monitor.py  (track 1-2 upstream domains for early signals)

GitHub Issues Adoption Sensor

A 4-script pipeline that watches an OSS-repo registry and infers adoption phases, the Daily Field Index (DFI), and cross-repo convergence.

collect-github-issues.py    (daily; tier-1 repos)
       ↓
  memory/gh-issues-YYYY-MM-DD.json
       ↓
compute-gh-adoption.py      (Fri; tier-1 + tier-2 → adoption phases + DFI)
       ↓
  memory/gh-adoption-YYYY-MM-DD.json
       ↓
update-gh-field-notes.py    (push back to PULSAR_FIELD_NOTES_REPO)
prep-community-context.py   (bundle community notes + adoption snapshot
                             into tmp file consumed by weekly reports)
_gh_issues_config.py        (shared: registry of monitored repos,
                             tier flags, method-family tags)

Semantic Memory

semantic-index-builder.py   (DashScope text-embedding-v3, batch=10, incremental)
semantic-search.py          (pure-Python cosine top-k)

Watchdog and System Health

daily-watchdog.py            (16 checks, self-healing, lockfile)
emit-system-health.py        (emit system-health.json)
evaluate-shadow-config.py    (A/B shadow config evaluator)
gateway-preflight.py         (gateway connectivity check)
server-health-check.sh       (shell-level health check)
system-moltbot-monitor.sh    (moltbot process monitor)

Memory Management

memory-janitor.py    (age-out old records from JSON stores)
memory-snapshot.py   (daily tar.gz snapshot → snapshots/)
memory-upsert.py     (atomic upsert helper)

GitHub Operations

gh-contents-upload.py                   (generic GitHub Contents API PUT)
gh-app-index-update.py                  (update AI app index on GitHub)
gh-paper-index-update.py                (update VLA paper index on GitHub)
gh-handbook-changes-collect.py          (collect VLA-Handbook git changes)
gh-agent-library-blog-readme-patch.py   (patch Agent-Playbook README)

Shared Helpers

_heartbeat_run.py                    (heartbeat keepalive for long-running agents)
_paper_index_input_from_hotspots.py  (extract paper index input from hotspots)

Known Naming Exceptions

  • write-ai-app-daily.py — writes to ai-app-daily.json; named "write" not "post" because it is a pure atomic writer with no agent dependency. Referenced in jobs.json by this name.
  • rate-vla-daily.py — Phase 1.5 between RSS collect and two-phase runner; no prep/run/post framing because it runs inline before the LLM agent step.
  • vla-daily-rerate-push.py — special re-push called by watchdog on keyword-fallback days.