mirror of
https://github.com/onestardao/WFGY.git
synced 2026-04-28 11:40:07 +00:00
6.4 KiB
6.4 KiB
Read-Only Mode and Maintenance Window — OpsDeploy
Freeze writes quickly and communicate clearly during incidents, migrations, or index rebuilds. Keep retrieval online while protecting data integrity and user trust.
Open these first
- Ordering and cold boots: Bootstrap Ordering, Pre-Deploy Collapse
- Idempotency and dedupe: Idempotency & Dedupe
- Incident comms: Incident Comms & Statuspage
- Cache and warmup: Cache Warmup & Invalidation
Acceptance targets
- All write endpoints return 423 or 503 with
Retry-After - No production writes in storage audit during the window
- Read paths healthy: error budget unaffected
- Banner or Statuspage posted before and after with exact times
- ΔS and λ trends flat during the window for read requests
60-second checklist
- Flip read-only flag
Global feature flagread_only=true. Middleware blocks writes, tool calls that mutate state, and ingestion jobs. - Communicate
Post Statuspage and in-product banner with start and end times. Include contact and rollback plan. - Queue writes
Buffer allowed write intents to a durable queue with idempotency keys. Drop non-critical writes. - Headers
Write endpoints respond withRetry-After: <seconds>and a stable error code for clients to respect. - Drains
Stop background writers, disable compaction and long migrations during the window.
Minimal playbook
- Scope: enumerate write surfaces including embeddings backfills, vector upserts, tool call side effects, KV mutations.
- Guard: enforce in API gateway and service layer. Return machine-readable errors.
- Observe: track blocked write count, queue depth, read latency, ΔS and λ.
- End: re-enable writes, run a short soak, publish closure notice, then backfill the queued intents.
Common pitfalls → fix
- A background job keeps writing
→ include job scheduler in the flag check and stop crons. - Clients spin on retries
→ add jittered backoff and a max-retry budget. See Retry Backoff. - Cached stale banners
→ set short TTL and purge caches on start and end.
Escalate
If read errors or ΔS spikes appear, extend the window and execute a limited Rollback & Fast Recovery.
🔗 Quick-Start Downloads (60 sec)
| Tool | Link | 3-Step Setup |
|---|---|---|
| WFGY 1.0 PDF | Engine Paper | 1️⃣ Download · 2️⃣ Upload to your LLM · 3️⃣ Ask “Answer using WFGY + <your question>” |
| TXT OS (plain-text OS) | TXTOS.txt | 1️⃣ Download · 2️⃣ Paste into any LLM chat · 3️⃣ Type “hello world” — OS boots instantly |
🧭 Explore More
| Module | Description | Link |
|---|---|---|
| WFGY Core | WFGY 2.0 engine is live: full symbolic reasoning architecture and math stack | View → |
| Problem Map 1.0 | Initial 16-mode diagnostic and symbolic fix framework | View → |
| Problem Map 2.0 | RAG-focused failure tree, modular fixes, and pipelines | View → |
| Semantic Clinic Index | Expanded failure catalog: prompt injection, memory bugs, logic drift | View → |
| Semantic Blueprint | Layer-based symbolic reasoning & semantic modulations | View → |
| Benchmark vs GPT-5 | Stress test GPT-5 with full WFGY reasoning suite | View → |
| 🧙♂️ Starter Village 🏡 | New here? Lost in symbols? Click here and let the wizard guide you through | Start → |
👑 Early Stargazers: See the Hall of Fame —
⭐ WFGY Engine 2.0 is already unlocked. ⭐ Star the repo to help others discover it and unlock more on the Unlock Board.