Merge pull request #163 from anurag008w/feat/llm-fallback-models

feat: LLM fallback models, May 2026 model updates, and bug fixes

Merge pull request #162 from anurag008w/feat/llm-fallback-models

fix: preserve existing manual fallbacks in config on restart

fix: remove duplicate fallback echo from JSON build block

feat: add LLM_FALLBACK_MODELS env var for automatic model failover

fix: handle WebSocket upgrade in Worker

fix(cloudflare-proxy): handle WebSocket upgrade in Worker without body/redirect

fix: IPv4 forcing, proxy fallback, browser symlink noise, model injection, WA stall recovery

Reject non-WebSocket upgrades and make devdata/openclaw prunes resilient to partial snapshots

Harden Jupyter devdata prune behavior for deleted files and copy races

Add optional reuse of suspended keys as last-resort and treat 429/402 as strike-bearing failures

Default to last-resort key reuse when all keys are suspended

Preserve fetch semantics and caller headers when injecting keys

fix retry/body safety and rpc cleanup in request patch paths

harden fetch arg builder for Request inputs and query providers

Add rotator verbosity & transient-failure handling; enable POST retries; make config/state snapshot writes safer

Avoid misleading rotator fetch errors for non-provider requests

Merge injected model providers using '*' to preserve provider fields in start.sh

Remove session size-bypass backup override

Prevent session backups from being dropped by file-size cap

Prevent remote prune during partial state snapshots

Validate restored config only when patching fails

Preserve existing openclaw config snapshots before regeneration

Harden startup config handling for invalid openclaw.json

Update env builder for key-rotator logging and POST retry labels

Align rotator docs/comments with new retry and logging behavior

Reduce rotator log noise with configurable verbosity

Harden key rotation under transient provider failures

Improve key-rotator retry classification for provider timeouts

Fix restored config merge to inject full provider settings

Key-rotator hardening, archive tools, env-builder polish & sudo wrapper

chore: fix footer — @somratpro as main author, @anurag008w as contributor

chore: remove legacy TELEGRAM_USER_IDS from env-builder

Merge pull request #21 from anurag008w/main

Merge branch 'main' into main

Remove legacy TELEGRAM_USER_ID and BACKUP_DATASET from env-builder

Fix sync robustness, devdata activation, alias cleanup, and timing-safe auth compare

fix: use GATEWAY_TOKEN as Jupyter token fallback when JUPYTER_TOKEN unset

fix: inject Jupyter token header to skip JupyterLab's own login screen

refactor: port HuggingMes auth pattern — unified /login, HMAC session, crypto.timingSafeEqual

Default Jupyter root changed to /home/node and README updated

Revert Jupyter root default to /home/node and document /home option

Improve privacy-detection timeout, robust state snapshotting, and Jupyter terminal defaults

Set default Jupyter root to filesystem root

Implement sessions_marker to monitor session changes

Fix shutdown process to handle sync loop correctly

Update comments and instructions in .env.example

Update labels for key rotation settings

Refactor key rotator and merge kimi-coding with moonshot