↕ scroll inside the wheel
Problem: Open WebUI won't come up on Docker (image pull stuck, port in use, page not loading, data loss)
four break points when deploying Open WebUI on Docker — direct ghcr.io pull did not finish within a 25s limit; the ghcr.nju.edu.cn mirror pulled 1.66GB in 3m24s; an occupied host port gives "address a
open-webui-deploy-fail · domain: Local LLM · verified
Problem: First `npx ruflo init` is slow and the npx cache grows to ~2GB (disk-sensitive setups)
first npx init takes ~5 minutes and the cache grows to ~2GB-class — npmmirror makes downloads near-instant in China; clear `~/.npm/_npx` after verification (existing projects unaffected).
ruflo-4 · domain: Multi-Agent Orchestration · verified
Problem: `hive-mind spawn --claude` cannot find the claude CLI on Windows (PATH detection failure)
on Windows, spawn --claude cannot detect the claude CLI (#3071 closed — PATH semantics difference) — add the claude directory to PATH or call it by full path; the Windows path was not reproduced here
ruflo-3 · domain: Multi-Agent Orchestration · verified
Problem: ruflo install stalls on an unpublished dependency pin (@claude-flow/mcp alpha not published)
ruflo install stalls during dependency resolution (@claude-flow/mcp alpha.10 unpublished) — retry after the upstream release, or pin a published version range; not reproduced here on 2026-08-29 init (
ruflo-2 · domain: Multi-Agent Orchestration · verified
Problem: `ruflo plugins install` does not exist (the correct skills entry is npx skills add)
the `ruflo plugins install` command suggested by `ruflo migrate status` does not exist — the correct skills entry is `npx skills add ruvnet/ruflo --all`; initialize with `npx -y ruflo init` (measured
ruflo-1 · domain: Multi-Agent Orchestration · verified
Problem: pip keeps timing out on CN networks — switch to a CN mirror
pip downloads slow/time out — switch to the Tsinghua mirror for one install: `pip install <pkg> -i https://pypi.tuna.tsinghua.edu.cn/simple`.
pip-mirror-source · domain: Coding Agents & Terminal · verified
Problem: Chinese text garbled or untypeable in herdr over SSH (terminal encoding and locale)
Chinese garbled/untypeable in herdr over SSH — this is terminal encoding / remote locale territory; make remote `locale` fully UTF-8 first, then check tmux-like layers. Not reproduced here (no remote
herdr-3 · domain: Coding Agents & Terminal · verified
Problem: Installing herdr on Windows via PowerShell fails (downloads get reset; download the binary directly instead)
herdr install via Windows PowerShell fails — PowerShell Web downloads of GitHub releases get reset easily; download the matching binary directly instead (gh-proxy prefix in China), or upgrade to the f
herdr-2 · domain: Coding Agents & Terminal · verified
Problem: `herdr server start` does nothing / background server is not started (no such subcommand in v0.8.2)
`herdr server start` does nothing — v0.8.2 has no such subcommand; use `herdr` (TUI auto-starts the server) or `nohup herdr server &` for headless, then verify with `herdr status`.
herdr-1 · domain: Coding Agents & Terminal · verified
Problem: agent hooks never install or fire (the repo setup only installs git hooks)
the framework repo setup installs only git hooks; agent hooks (pre-tool-call etc.) must be registered per the agent platform conventions (#3732 material). Not reproduced here (stated honestly).
agent-framework-4 · domain: Multi-Agent Orchestration · verified
Problem: MCP connection only shows Cancelled via... (the real 401 auth error is hidden)
2026-08: MCP HTTP auth failures show only "Cancelled via..." while the real 401 is hidden (#7699/#7704/#7656, open) — check credentials/scope first, or curl -i the endpoint directly to see 401/403; up
agent-framework-3 · domain: Multi-Agent Orchestration · verified
Problem: Installing full agent-framework on Python 3.14 hangs with resolution-too-deep (dependency graph too deep)
full meta-package resolution on Python 3.14 did not converge in ~20 minutes and pip reported `resolution-too-deep` — the minimal core `agent-framework-core` (31MB / 9 deps) installed successfully; off
agent-framework-2 · domain: Multi-Agent Orchestration · verified
Problem: Same package won't install across products — Codex can't read skills / CodeBuddy channel error
no local file read permission / channel mismatch — grant local file read in Codex desktop settings first, then install per product form.
superpowers-7 · domain: Skills Management · verified
Problem: venv scripts fail with "bad interpreter: No such file or directory" after moving (python -V still works)
venv breaks after moving/copying — venvs are not portable; don't fix shebangs, freeze dependencies and recreate.
python-venv-move · domain: Coding Agents & Terminal · verified
Problem: pip install fails with "externally-managed-environment" on new Ubuntu/WSL
pip reports externally-managed-environment (PEP 668) — create a venv and install inside it: python3 -m venv .venv && source .venv/bin/activate.
pep668-externally-managed · domain: Coding Agents & Terminal · verified
Problem: Forgot the Open WebUI admin password / can't log in
forgot the Open WebUI admin password — docker exec into the container and run a one-liner password reset (bcrypt + sqlite3 on webui.db).
open-webui-forgot-password · domain: Local LLM · verified
Problem: Models keep filling up the system drive — how to move the Ollama model directory
models fill the C drive — stop the service (systemctl stop ollama), move the model dir to another drive (copy first, delete old after verifying), restart.
ollama-models-migration · domain: Local LLM · verified
Problem: npm install -g fails with EACCES (two forms, two fixes)
npm install -g EACCES — move the global prefix to your user dir (`npm config set prefix ~/.npm-global`) for a permanent fix.
npm-global-eacces · domain: Coding Agents & Terminal · verified
Problem: Build crashes with "JavaScript heap out of memory" (FATAL ERROR: Reached heap limit)
Node build fails with heap out of memory — raise the cap temporarily: NODE_OPTIONS=--max-old-space-size=8192 (size to your machine).
node-heap-oom · domain: Coding Agents & Terminal · verified
Problem: "detected dubious ownership in repository" — every git command fails
dubious ownership — `git config --global --add safe.directory <abs path>`; the whitelist is per-directory, not recursive, so other repos on the same drive need their own entry.
git-dubious-ownership · domain: Coding Agents & Terminal · verified
Problem: Container can see the mount but reads/writes fail with Permission denied (uid mismatch)
bind-mount permission denied — chown the host dir to the container uid (chown -R 1000:1000), or run with -u 1000:1000.
docker-volume-permission · domain: Coding Agents & Terminal · verified
Problem: Docker Desktop fails to start: The file cannot be accessed by the system / orphan AF_UNIX socket
Docker Desktop fails with The file cannot be accessed — stop all Docker processes, then rename (don't delete) the %LOCALAPPDATA%\Docker\run parent dir.
docker-desktop-orphan-socket · domain: Coding Agents & Terminal · verified
Problem: crontab jobs report command not found / wrong environment
crontab jobs fail with command not found because cron does not inherit your shell PATH — declare the full PATH at the top of the crontab and reload.
crontab-env-path · domain: Coding Agents & Terminal · verified
Problem: Chinese filenames garbled when unzipping Windows-created zips on Linux/WSL (GBK encoding)
garbled Chinese filenames after unzip — Windows/360 archives are GBK; extract with `unzip -O GBK`.
zip-gbk-filename · domain: Coding Agents & Terminal · verified
Problem: WSL reports Temporary failure in name resolution / no network, and VPN kills WSL connectivity
WSL name resolution fails / network dies when VPN is on — point the systemd stub at reachable upstreams (DNS=223.5.5.5 119.29.29.29) and restart systemd-resolved.
wsl-dns-resolution-failure · domain: Coding Agents & Terminal · verified
Problem: JDK old-project TLS upgrade fails with Received fatal alert: protocol_version / No appropriate protocol
TLS handshake fails after JDK upgrade — disable SSLv3/TLSv1/TLSv1.1 explicitly in jdk.tls.disabledAlgorithms (java.security) to force TLS1.2+.
tls-jdk-upgrade · domain: Security & Governance · verified
Problem: SessionStart hook error after installing superpowers
Windows SessionStart hook fails — upgrade both sides first (claude update + /plugin update), then run the hook manually to see the real error.
superpowers-4 · domain: Skills Management · verified
Problem: /plugin install superpowers says "plugin not found" (marketplace source not added)
Plugin not found — add the source, install, verify: /plugin marketplace add obra/superpowers-marketplace then /plugin install.
superpowers-3 · domain: Skills Management · verified
Problem: Only one specific website won't open (everything else works) — from curl timeout to layer-by-layer diagnosis
When only one site won't open, locate the layer with curl: getent to check whether DNS resolves -> direct connect to check the TCP layer -> --resolve to bypass DNS and hit the resolved IP -> distingui
site-open-fail-others-ok · domain: Coding Agents & Terminal · verified
Problem: UnicodeDecodeError reading files — file encoding vs default encoding mismatch
Python UnicodeDecodeError reading a file — check the real encoding with `file`, then pass it explicitly (e.g. encoding='gbk').
python-unicode-decode · domain: Coding Agents & Terminal · verified
Problem: Open WebUI knowledge base retrieval not working / embedding misconfigured (file upload fails with Ollama embed 404)
RAG retrieval not working / embed 404 — wrong embedding config: update the retrieval/embedding url and model name via the admin API (bare name, no ollama/ prefix).
open-webui-rag-embedding · domain: Local LLM · verified
Problem: External extensions fail to load — "is not host-provided" (Windows) / silent failure (darwin-arm64)
Windows/ARM release binaries fail to load external extensions — upgrade to >=17.1.5 (18.0.9 includes the fix).
oh-my-pi-3 · domain: Coding Agents & Terminal · verified
Problem: OMC hooks don't trigger (UserPromptSubmit/Stop hook error)
hooks don't fire / Windows hook errors — usually missing execute permission: find ~/.claude/hooks -name '*.sh' -exec chmod +x {} +.
oh-my-claudecode-2 · domain: Multi-Agent Orchestration · verified
Problem: npm install slow/timing out in CN — measure before switching mirrors
npm install slow/hangs — disable audit and raise the timeout: `npm install --no-audit --no-fund --fetch-timeout 600000`.
npm-mirror-registry · domain: Coding Agents & Terminal · verified
Problem: Loss spikes (loss spike) / numeric divergence (NaN, inf) during training
loss spikes/divergence during training — clip gradients (clip_grad_norm_ max_norm=1.0), resume from the last healthy checkpoint, drop the LR to 1/3~1/10.
llm-training-loss-spike · domain: Local LLM · verified
Problem: GitHub direct connection is slow/fails (clone reset, port 443 unreachable, SSH auth, release downloads interrupted)
GitHub direct connection slow/fails — diagnose in layers: curl for network, anonymous HTTPS ls-remote for config, then accelerators/proxies last.
github-direct-connect-fail · domain: Coding Agents & Terminal · verified
Problem: Codex sandbox error helper_unknown_error: setup refresh had errors (write permission/ACL)
Codex sandbox setup refresh had errors — usually directory write permission/ACL: run takeown /f <dir> /r /d y in an elevated PowerShell to take ownership.
codex-sandbox-setup-error · domain: Security & Governance · verified
Problem: "No previous sessions found" after switching third-party model auth
"No previous sessions found" after switching third-party model auth — write BASE_URL/AUTH_TOKEN into `~/.claude-mem/.env` and restart the worker.
claude-mem-6 · domain: Sessions & Memory · verified
Problem: Same-name pollution: pip installs ivanzwb / multiple foreign repos with the same name
pip installs an impostor package — use the official path npx -y skills add addyosmani/agent-skills; on GitHub trust the addyosmani org.
agent-skills-3 · domain: Skills Management · verified
Problem: yfinance history times out in CN — cookie/crumb fetch failures, curl (28)
yfinance history always times out — upgrade to >=1.7 and use a curl_cffi firefox-fingerprint session with a proxy (yf.download(..., proxy='socks5h://...')).
yfinance-history-timeout · domain: Gateway & Routing · verified
Problem: systemctl reports "System has not been booted with systemd" / services won't start in WSL
systemctl says System has not been booted with systemd — add [boot] systemd=true to /etc/wsl.conf, wsl --terminate, and ps -p 1 must show systemd.
wsl-systemd-not-running · domain: Coding Agents & Terminal · verified
Problem: WSL fails to start a distro with WslRegisterDistribution failed (0x800701bc / 0x80370102)
WslRegisterDistribution failed — 0x800701bc is a missing kernel (wsl --update on Store builds; kernel update package otherwise); 0x80370102 means virtualization is off (bcdedit /set hypervisorlaunchty
wsl-register-failed · domain: Coding Agents & Terminal · verified
Problem: Garbled output when calling wsl.exe from PowerShell (UTF-16LE / every character spaced)
garbled wsl output in PowerShell — set WSL_UTF8=1 at the top of the script (or `cmd /c set WSL_UTF8=1 && wsl ...`).
wsl-pipe-encoding · domain: Coding Agents & Terminal · verified
Problem: "No space left on device" in WSL while df shows hundreds of GB free (vhdx on a full host drive)
WSL reports No space left but df has space — the ext4.vhdx filled the Windows drive; locate it, free host space, and WSL resumes writing.
wsl-no-space-left · domain: Coding Agents & Terminal · verified
Problem: notepad.exe / cmd.exe reports command not found in WSL (interop disabled)
notepad.exe command not found in WSL — enable [interop] enabled=true + appendWindowsPath=true in /etc/wsl.conf, restart the distro.
wsl-interop-disabled · domain: Coding Agents & Terminal · verified
Problem: default user becomes root after `wsl --import` migration/reinstall — whoami changed, permissions broken
default user becomes root after wsl --import — add [user] default=<name> to /etc/wsl.conf, then wsl --terminate and re-enter the distro.
wsl-default-user-root-after-import · domain: Coding Agents & Terminal · verified
Problem: Refreshing a deep link white-screens after deploy (404) — SPA history fallback missing
history-route refresh 404 — nginx try_files falls back to the SPA entry (/app/index.html); static assets get their own location with =404.
vite-history-fallback · domain: Build & Deploy · verified
Problem: White screen for some users after release — ChunkLoadError: Loading chunk N failed (404)
occasional white screen ChunkLoadError after deploy — version mismatch: upload the new dist assets first, overwrite index.html last, keep old assets for at least one release cycle.
vite-chunkloaderror · domain: Build & Deploy · verified
Problem: Leftovers after uninstall — shared config not restored, command still resolves
leftovers after uninstall — back up shared config first (only undo option); full uninstall = remove package + tool config dir + restore shared config sections manually.
uninstall-leftover · domain: Coding Agents & Terminal · verified
Problem: tar.gz extracts with an unexpected pkg/ prefix; macOS says --strip-components not supported
tar extracts an extra top-level dir / scatters files — inspect with tar -tzf, then strip one level with --strip-components=1 (not supported on macOS).
tar-strip-components · domain: Coding Agents & Terminal · verified
Problem: Offline install fails — /plugin list shows "!" unavailable after manual zip extract
plugin shows unavailable after local/offline deploy — use it as skills: cp -r skills ~/.claude/skills/superpowers (Codex: ~/.codex/skills).
superpowers-6 · domain: Skills Management · verified
Problem: /find-skills doesn't show brainstorming / systematic-debugging / TDD after install
skills not visible after install — check /plugin list, fully exit and reopen a session, then /find-skills.
superpowers-5 · domain: Skills Management · verified
Problem: Downloading GitHub zip/tarball keeps timing out and breaking
ZIP/tarball download times out — switch to a shallow git clone: --depth 1 --branch v6.3.0 (smaller, higher success).
superpowers-2 · domain: Skills Management · verified
Problem: /plugin install superpowers fails — can't connect to github.com (app proxy doesn't cover git)
/plugin install can't reach github:443 — probe codeload with curl -sI first (1.7~1.9s tested); if reachable use tarball install, or set an explicit git proxy.
superpowers-1 · domain: Skills Management · verified
Problem: "UNPROTECTED PRIVATE KEY FILE" / "Permissions 0644 are too open" — ssh login refused
UNPROTECTED PRIVATE KEY FILE — chmod 600 the key and 700 ~/.ssh (Linux/macOS/WSL); use icacls on native Windows OpenSSH.
ssh-key-permissions · domain: Coding Agents & Terminal · verified
Problem: SMS API hammered by scripts/abuse — costs spike, endpoint collapses (no rate limiting)
SMS API being hammered — rate-limit by IP at nginx: limit_req_zone (rate=5r/s) + limit_req burst=10 nodelay inside the location, 503 above the limit.
sms-api-rate-limit · domain: Security & Governance · verified
Problem: "Skipping analyzer semantic_developer_intent: required API key is missing" — is config broken?
semantic_* analyzer warnings — missing API key: use --no-llm for static scans, or set OPENAI_API_KEY / SKILLSPECTOR_PROVIDER for semantic analysis.
skillspector-6 · domain: Security & Governance · verified
Problem: Scan report shows DO NOT INSTALL / high risk score — is the tool broken?
risk score >50 prints DO NOT INSTALL — read the Issues details (severity/confidence/rule/location) before concluding; don't trust the score alone.
skillspector-5 · domain: Security & Governance · verified
Problem: skillspector conflicts with existing packages (langchain/openai/pydantic) — afraid to install
dependency version conflicts — install in an isolated venv to avoid polluting system Python (pip install skillspector inside the venv; tested).
skillspector-4 · domain: Security & Governance · verified
Problem: skillspector mcp hangs with no output (missing optional deps)
MCP stdio hangs — confirm the CLI layer with skillspector mcp --help first, then install the full optional deps per project docs (pip install -U skillspector[mcp]).
skillspector-3 · domain: Security & Governance · verified
Problem: yara-python fails to build during pip install — missing build toolchain
yara-python build fails — install build-essential python3-dev on Debian/Ubuntu and retry.
skillspector-2 · domain: Security & Governance · verified
Problem: pip install skillspector fails — "No matching distribution" (it's not on PyPI)
pip can't find skillspector — not on PyPI: download the release wheel (gh-proxy), pip install, deps via the Tsinghua mirror.
skillspector-1 · domain: Security & Governance · verified
Problem: ./setup.sh fails with Permission denied but bash setup.sh works (missing +x)
bash: ./script.sh: Permission denied — the script lacks the execute bit; chmod +x and run.
script-permission-denied · domain: Coding Agents & Terminal · verified
Problem: ModuleNotFoundError: No module named 'distutils' on Python 3.12+
Python 3.12+ lacks distutils — `pip install setuptools` (inside the venv for venvs; sudo for system Python).
python-distutils-removed · domain: Coding Agents & Terminal · verified
Problem: "python: command not found" on Ubuntu/WSL (python3 works fine)
python: command not found — Ubuntu ships python3 only; install the official python-is-python3 package with no side effects.
python-command-not-found · domain: Coding Agents & Terminal · verified
Problem: proxy env vars set but not taking effect — some commands use the proxy, some don't (HTTP_PROXY case-sensitivity trap)
proxy env vars set but ignored — curl only honors lowercase http_proxy/https_proxy (uppercase HTTP_PROXY is deliberately ignored); pip/npm honor uppercase; NO_PROXY works for both.
proxy-env-var · domain: Coding Agents & Terminal · verified
Problem: pip install fails with SSL certificate error (CERTIFICATE_VERIFY_FAILED) — the trust chain broke, don't rush to --trusted-host
pip reports CERTIFICATE_VERIFY_FAILED — root cause is the CA trust chain does not include that certificate (self-signed / private internal CA / corporate proxy MITM); --trusted-host only skips verific
pip-ssl-cert · domain: Coding Agents & Terminal · verified
Problem: Open WebUI in Docker cannot connect to Ollama (Ollama not connected / empty model list)
Open WebUI in a container can't reach Ollama — recreate the container pointing at host Ollama, with --add-host host.docker.internal:host-gateway on Linux.
open-webui-ollama-connect · domain: Local LLM · verified
Problem: ollama pull hangs/times out (official registry unreachable from CN)
ollama pull hangs/times out — the official registry is unstable in China: download the GGUF from ModelScope and import with ollama create (e.g. qwen2.5:0.5b).
ollama-pull-timeout · domain: Local LLM · verified
Problem: ollama won't install — official installer/download unreachable from CN (slow, timeouts)
official installer stalls in China — pull the GitHub Release tar.zst via gh-proxy, extract to /usr/local, verify the version (tested).
ollama-install-slow · domain: Local LLM · verified
Problem: ollama reports "could not connect to a running Ollama instance" (server not started)
could not connect to a running Ollama instance — the service isn't started: run nohup ollama serve, then curl http://127.0.0.1:11434/api/version for readiness.
ollama-connect-failed · domain: Local LLM · verified
Problem: Mixing up oh-my-pi and pi-mono — command names don't match (omp vs pi)
oh-my-pi (omp) vs pi-mono (pi) names are easy to confuse — identify by command: if `omp --version` works after install, it's oh-my-pi.
oh-my-pi-4 · domain: Coding Agents & Terminal · verified
Problem: omp install <plugin> fails or behaves unexpectedly — the command in tutorials is outdated
omp install and omp plugin install are different — use the explicit subcommand `omp plugin install <name>` to avoid ambiguity.
oh-my-pi-2 · domain: Coding Agents & Terminal · verified
Problem: Install succeeds but omp --version reports SyntaxError (old bun)
omp --version throws SyntaxError — old bun is incompatible; upgrade bun to >=1.3.14 and re-verify.
oh-my-pi-1 · domain: Coding Agents & Terminal · verified
Problem: Multiple same/near-name repos for "oh-my-claudecode" — installing the wrong fork
same-name fork confusion — npm view to confirm versions, use the official oh-my-claude-sisyphus@5.0.0, verify with omc --version.
oh-my-claudecode-7 · domain: Multi-Agent Orchestration · verified
Problem: Claude Code itself breaks during OMC install (OOM Killed / Homebrew conflict / scripts disabled)
Claude Code environment errors — install the CLI via npm on WSL/Linux to avoid brew duplication; on Windows PowerShell set ExecutionPolicy RemoteSigned first.
oh-my-claudecode-6 · domain: Multi-Agent Orchestration · verified
Problem: Can't find "oh-my-claudecode" on npm; don't know the uninstall package name
package/plugin/command names differ — install/uninstall via the npm name oh-my-claude-sisyphus; use the omc command daily.
oh-my-claudecode-5 · domain: Multi-Agent Orchestration · verified
Problem: Abnormal behavior/config conflicts after OMC upgrade
update/config anomalies — run the built-in diagnostic omc-doctor to locate the issue; keep the output (version + node version) when reporting.
oh-my-claudecode-4 · domain: Multi-Agent Orchestration · verified
Problem: Multi-agent tasks not auto-delegated/claimed in session
agents not auto-delegated — rerun /omc-setup in the correct project root (CLAUDE.md is generated per project), then check whether rules loaded.
oh-my-claudecode-3 · domain: Multi-Agent Orchestration · verified
Problem: omc not found after install (or /omc-setup didn't produce a working CLI)
omc: command not found — confirm install (npm ls -g oh-my-claude-sisyphus), reinstall 5.0.0 via npmmirror, check PATH.
oh-my-claudecode-1 · domain: Multi-Agent Orchestration · verified
Problem: npm install fails with ERESOLVE unable to resolve dependency tree (strict peer deps)
ERESOLVE peer dependency conflict — `npm install --legacy-peer-deps` skips strict peer checks (best transition for legacy projects).
npm-eresolve-peer-conflict · domain: Coding Agents & Terminal · verified
Problem: "digital envelope routines::unsupported" building an old webpack 4 project on new Node
digital envelope routines::unsupported with newer Node + old webpack — NODE_OPTIONS=--openssl-legacy-provider as a transition; upgrade the build toolchain long-term.
npm-digital-envelope-unsupported · domain: Coding Agents & Terminal · verified
Problem: Netty direct-memory OOM: OutOfDirectMemoryError / Direct buffer memory (ByteBuf leak + unset MaxDirectMemorySize)
Netty OutOfDirectMemoryError — isolate heap and direct memory with explicit -XX:MaxDirectMemorySize and generous container limits (stopgap); fix for good with a body-cache size threshold and release f
netty-direct-memory-oom · domain: Coding Agents & Terminal · verified
Problem: MySQL data lost after accidental DELETE - recover with binlog flashback
accidentally deleted MySQL data — binlog flashback: SHOW BINLOG EVENTS to locate the DELETE transaction start, then export with --stop-position=DELETE start and replay the writes.
mysql-binlog-flashback · domain: Build & Deploy · verified
Problem: MinIO community edition is archived / unmaintained — what to do with a Docker-deployed file service (migrate to RustFS, zero data loss)
MinIO community edition archived — migrating to RustFS can be zero-loss: record the data dir with docker inspect, stop the old container (data stays), mount the same dir on the new service.
minio-to-rustfs-migration · domain: Build & Deploy · verified
Problem: Flood of "Invalid type NoneType for attribute http.port" warnings after logfire upgrade
Invalid type NoneType for http.port after upgrade — set by third-party clients (kiota/httpx); the warning is harmless and can be ignored.
logfire-2 · domain: Observability · verified
Problem: NoInspectionAvailable with logfire.instrument_pydantic + SQLModel (old logfire bug, fixed)
FastAPI+SQLModel NoInspectionAvailable — upgrade logfire to >=2.11.1 (4.41.0 includes the fix).
logfire-1 · domain: Observability · verified
Problem: Java OOM signature cheat-sheet: heap space / Metaspace / StackOverflowError / native thread / GC Overhead
Java OOM has six distinct signatures — enable HeapDumpOnOutOfMemoryError first, read GC logs to see which resource is exhausted (heap/Metaspace/stack/threads/GC overhead), then tune per signature.
java-oom-classify · domain: Coding Agents & Terminal · verified
Problem: Hugging Face model download hangs/times out in China — switch HF_ENDPOINT to hf-mirror, zero code changes
HF model downloads hang in China — huggingface.co is unreachable direct (measured timeout 000); set HF_ENDPOINT=https://hf-mirror.com and hf_hub_download succeeds in ~1.2s (662B, zero code changes).
hf-model-download · domain: Coding Agents & Terminal · verified
Problem: go get/build times out: proxy.golang.org dial i/o timeout (Google IPs)
go get times out — switch to a China proxy temporarily: `GOPROXY=https://goproxy.cn,direct go mod download`.
go-proxy-cn · domain: Coding Agents & Terminal · verified
Problem: Cloning a public repo fails: "Your account is suspended" (config/credential layer, not network)
cloning a public repo reports auth errors — verify with anonymous HTTPS first (`GIT_CONFIG_GLOBAL=/dev/null git ls-remote ...`); if refs return, it's config — unset the git@ rewrite rule.
github-clone-fail · domain: Coding Agents & Terminal · verified
Problem: One-line change shows the whole file as modified (line-ending chaos: LF/CRLF)
git line-ending chaos — add .gitattributes (* text=auto eol=lf) at repo root and `git add --renormalize .` once and for all.
git-crlf · domain: Coding Agents & Terminal · verified
Problem: LLM function calling is not trustworthy - dangerous AI-suggested calls must pass an allowlist (tested locally with qwen)
LLM function calling is uncontrollable — dangerous calls must pass a whitelist (read-only functions allowed; delete/refund/update need double confirmation), tested with local qwen.
function-calling-guard · domain: Local LLM · verified
Problem: Intermittent SIGSEGV from DBL_MAX sentinel + NULL best_key (float ABI mismatch / NaN)
sentinel pointer stays NULL when a float comparison can never hold — filter NaN/Inf and null-check the sentinel in code; fix the ABI (soft-float rebuild) at the environment level.
float-crash · domain: Coding Agents & Terminal · verified
Problem: PowerShell refuses to run scripts: "running scripts is disabled on this system"
PowerShell blocks scripts — process-scoped bypass is safest: `powershell -NoProfile -ExecutionPolicy Bypass -File script.ps1`, no registry changes.
execpolicy-restricted · domain: Coding Agents & Terminal · verified
Problem: what does docker system prune actually delete? container/image/volume/cache cleanup boundaries
prune commands differ in scope — system prune removes stopped containers + dangling images + unused networks, never volumes or build cache; volume prune only removes anonymous volumes by default (name
docker-prune-scope · domain: Coding Agents & Terminal · verified
Problem: docker run/compose fails: host port already allocated / address already in use
port already allocated — check docker ps for container usage, ss -tlnp for host processes, then switch ports.
docker-port-already-allocated · domain: Coding Agents & Terminal · verified
Problem: docker pull says "manifest for xxx:latest not found"
manifest not found — verify remote existence with `docker manifest inspect` first, then pull an explicit tag instead of betting on latest.
docker-manifest-not-found · domain: Coding Agents & Terminal · verified
Problem: docker says "invalid reference format" / "repository name must be lowercase"
invalid reference format — print the image name docker actually receives, then compare against a guaranteed-valid command like `docker run ubuntu` to find case/spaces/quote residue.
docker-invalid-reference-format · domain: Coding Agents & Terminal · verified
Problem: docker pull fails on campus/home network (connection reset / dead mirror lists)
docker pull connection reset — back up daemon.json, write a working registry-mirrors accelerator, restart docker.
docker-hub-pull-fail · domain: Coding Agents & Terminal · verified
Problem: Docker container cannot reach a service on the host (localhost inside the container is the container itself, not the host)
can't reach host services from a container — localhost is the container itself; run with --add-host host.docker.internal:host-gateway and use host.docker.internal inside.
docker-host-access · domain: Coding Agents & Terminal · verified
Problem: docker exec reports the input device is not a TTY / container is not running / executable file not found
docker exec reports not a TTY — drop -t for non-interactive runs: `docker exec <container> sh -c "cmd"`.
docker-exec-tty · domain: Coding Agents & Terminal · verified
Problem: docker commands report Cannot connect to the Docker daemon at unix:///var/run/docker.sock / failed to connect to the docker API (can't reach the daemon)
Cannot connect to the Docker daemon — unset DOCKER_HOST first to rule out a stale env var, then confirm Docker Desktop is actually running.
docker-daemon-connect · domain: Coding Agents & Terminal · verified
Problem: container time is 8 hours off (date shows UTC) / Chinese text garbled (UnicodeEncodeError)
container time is 8h off / Chinese garbled — use -e TZ=Asia/Shanghai (TZ=UTC-8 for images without tzdata, sign inverted); fix locale for mojibake.
docker-container-timezone-locale · domain: Coding Agents & Terminal · verified
Problem: container process killed suddenly (OOMKilled / exit code 137) — memory limits and troubleshooting
container killed with exit 137 — check `docker inspect` OOMKilled: true = raise memory limits; false = killed by someone else.
docker-container-oom-killed · domain: Coding Agents & Terminal · verified
Problem: docker says "Conflict. The container name /xxx is already in use"
container name conflict — `docker rm -f` the old container holding the name (exited ones too), then rerun the original command.
docker-container-name-conflict · domain: Coding Agents & Terminal · verified
Problem: DNS resolution fails inside Docker containers (wget: bad address / Temporary failure in name resolution / Could not resolve host)
DNS resolution fails inside containers — rerun with public DNS: --dns 223.5.5.5 --dns 119.29.29.29.
docker-container-dns · domain: Coding Agents & Terminal · verified
Problem: docker compose stack fails to start — dependency not ready (Connection refused / unhealthy / depends_on ignored)
depends_on only waits for container start, not readiness — add healthchecks to dependencies and use the long-form depends_on with condition.
docker-compose-depends-not-ready · domain: Coding Agents & Terminal · verified
Problem: docker-compose / docker compose says command not found
docker-compose command not found — on Ubuntu/Debian install the docker-compose-v2 plugin (tested 2.40.3), then verify with `docker compose version`.
docker-compose-command-not-found · domain: Coding Agents & Terminal · verified
Problem: Docker build cache not working (install layers re-run on every code change) - COPY order and .dockerignore
Docker build cache keeps missing because of COPY order — COPY dependency manifests first, RUN install, and COPY the full tree last.
docker-build-cache · domain: Coding Agents & Terminal · verified
Problem: agent unresponsive after install (needs init + provider config)
agent unresponsive after install — missing init + provider: run deepcode init && deepcode provider and configure model/key.
deepcode-6 · domain: Multi-Agent Orchestration · verified
Problem: MCP client won't start (Brave MCP WinError 2 / npx didn't run)
MCP client won't start — confirm npx works, predownload the MCP package (npx -y <pkg>), then configure it into deepcode.
deepcode-5 · domain: Multi-Agent Orchestration · verified
Problem: Old Python can't install (requires-python)
old Python can't install — requires-python: 3.10+ installs directly with pip; below 3.10 upgrade via pyenv/conda first (don't touch the system default).
deepcode-4 · domain: Multi-Agent Orchestration · verified
Problem: curl certificate error (--ssl-no-revoke)
Windows curl cert error — add --ssl-no-revoke to skip revocation checks.
deepcode-3 · domain: Multi-Agent Orchestration · verified
Problem: git clone connection reset (source install path)
source install clone reset — clear stale git proxies and pip install deepcode-hku directly; no need to clone source.
deepcode-2 · domain: Multi-Agent Orchestration · verified
Problem: curl resume fails: (33) HTTP server does not seem to support byte ranges
interrupted curl downloads can't resume — probe with a Range header first (206 = supported); if supported use `curl -C -`, otherwise switch source or redownload.
curl-resume-break · domain: Coding Agents & Terminal · verified
Problem: CSS fails to load (page unstyled/layout broken, console may show nothing)
CSS fails to load — on nginx it's usually a missing `include mime.types;` (.css not served as text/css); add it, nginx -t, then reload.
css-load-failure · domain: Build & Deploy · verified
Problem: bash script reports $'\r': command not found / bad interpreter: /bin/bash^M - Windows CRLF mixed in
bash `$'\r': command not found` means Windows CRLF leaked into the script — fix with one line: `sed -i 's/\r//' script.sh`.
crlf-command-not-found · domain: Coding Agents & Terminal · verified
Problem: Codex desktop app fails to update (in-app updater error / manual MSIX install)
Codex desktop update fails — force-quit all ChatGPT/Codex processes first (required), then Add-AppxPackage the official MSIX manually.
codex-update-msix · domain: Coding Agents & Terminal · verified
Problem: ChatGPT Windows client won't open: Unable to locate the Codex CLI binary / Codex CLI not found
ChatGPT client cannot locate the Codex CLI — run `codex --version` first; if missing, install via the official PowerShell script, then set CODEX_CLI_PATH to the real path (don't copy paths from the we
codex-cli-chatgpt-not-found · domain: Coding Agents & Terminal · verified
Problem: Direct GitHub slow/fails (scenarios depending on GitHub)
GitHub direct connection slow — the core flow is offline: code-review-graph build --repo . works without network; use the Tsinghua mirror for pip.
code-review-graph-4 · domain: Code Review · verified
Problem: "My config got changed" after install / suspected conflict with omc, agt
config changed after install — those writes are features, not faults; use code-review-graph uninstall for official cleanup if unwanted.
code-review-graph-3 · domain: Code Review · verified
Problem: Name confusion: juspay/code-review-graph-rescript same-name variant
same-name variant confusion — pip install code-review-graph==2.3.8; output 2.3.8 is correct, rescript content means the wrong variant.
code-review-graph-2 · domain: Code Review · verified
Problem: opencode install API mismatch
opencode install API mismatch — run `claude-mem doctor` first; on opencode prefer waiting for a plugin update or a compatible implementation.
claude-mem-7 · domain: Sessions & Memory · verified
Problem: Chroma process leak / zombie chains
Chroma process leaks/zombie chains — `pkill -f chroma` to clear residue and restart the worker; upgrading the plugin to 13.16.1 fixes it for good.
claude-mem-5 · domain: Sessions & Memory · verified
Problem: npm dependency install fails
npm dependency install fails — reinstall with the npmmirror registry and confirm the version via `claude-mem --version`.
claude-mem-4 · domain: Sessions & Memory · verified
Problem: Repeated SessionStart hook errors in the first session (worker race)
Repeated SessionStart hook errors in the first session are a worker warm-up race — run `claude-mem start && claude-mem status` before opening a session.
claude-mem-3 · domain: Sessions & Memory · verified
Problem: worker/UI port in use: "Is port 37777 in use?"
Port-in-use errors — use the actual port from install output (default 37700; old docs say 37777), free it with `kill -9`, then restart via `claude-mem start`.
claude-mem-2 · domain: Sessions & Memory · verified
Problem: marketplace clone fails: SSH authentication failed / clone stuck
Marketplace clone failure is not an SSH key issue — public repos clone anonymously; force HTTPS + gh-proxy turns a 90s stuck clone into seconds.
claude-mem-1 · domain: Sessions & Memory · verified
Problem: Multiple tools write ~/.claude.json - do they conflict when installed together? (measured compatible)
multiple tools writing ~/.claude.json — verified compatible: install via each official command, then python3 -m json.tool to confirm the JSON parses.
claude-json-multi-tool · domain: Multi-Agent Orchestration · verified
Problem: JSON / API config errors make the service fail to start
service fails to start on bad JSON config — back up the config first, validate with jq empty, then ccr restart.
claude-code-router-6 · domain: Gateway & Routing · verified
Problem: Docker deployment: ports / volumes / permissions
Docker deploy broken/lost config/permission — follow docker/README.md: -p 8080:8080 + named volume + -u 1000:1000.
claude-code-router-5 · domain: Gateway & Routing · verified
Problem: Local model api_base_url misconfigured (Ollama etc.)
local models unreachable — confirm names with ollama list, set api_base_url to the full URL (http://localhost:11434/v1/chat/completions), then ccr restart.
claude-code-router-4 · domain: Gateway & Routing · verified
Problem: First run shows No available models / gateway won't start
No available models on first run — start the gateway with ccr start, then add a provider in the web UI.
claude-code-router-3 · domain: Gateway & Routing · verified
Problem: node: not found inside WSL (npm came from the Windows side)
node: not found in WSL — npm came from Windows; install native Node 22 in WSL via nvm and reinstall the global package.
claude-code-router-2 · domain: Gateway & Routing · verified
Problem: EBADENGINE: Node version doesn't meet the engines requirement
EBADENGINE means Node doesn't meet engines — nvm install 22, switch, and reinstall the global package.
claude-code-router-1 · domain: Gateway & Routing · verified
Problem: Install source confusion: npm same-name package / other cc-switch projects
The npm package with the same name is an impostor — use the official repo farion1231/cc-switch and install the deb from Release.
cc-switch-5 · domain: Sessions & Memory · verified
Problem: Launch failure / white screen troubleshooting
Launch failure/white screen — run `ldd /usr/bin/cc-switch | grep not found` to find missing dynamic libraries and install them.
cc-switch-4 · domain: Sessions & Memory · verified
Problem: Architecture mismatch (ARM devices like Orange Pi)
ARM devices (Orange Pi etc.) reject x86 packages — check `dpkg --print-architecture` first, then install the matching deb.
cc-switch-3 · domain: Sessions & Memory · verified
Problem: GitHub Release download TLS drop / timeout
GitHub Release TLS drop/timeout — prefix the URL with gh-proxy; 13.5MB downloaded in seconds in local tests; drop the prefix if the mirror is down.
cc-switch-2 · domain: Sessions & Memory · verified
Problem: dpkg -i missing dependencies: libwebkit2gtk etc. (Linux install/launch fails)
dpkg install of cc-switch fails on libwebkit2gtk etc. — run `sudo apt install -f` to auto-fix dependencies, then reinstall (locally tested install/uninstall cycle).
cc-switch-1 · domain: Sessions & Memory · verified
Problem: apt-get update fails with 404 Not Found / broken source (Err:1 ... 404)
apt-get update 404 means a dead source — locate the bad entry from the Err: line, verify with curl, then remove/comment it and update again.
apt-404-broken · domain: Coding Agents & Terminal · verified
Problem: Can the full Agent toolchain be installed together? (17 tools, 57 measured-compatible pairs)
can the whole agent toolchain coexist — yes: 17 tools, 57 pairs verified compatible; run 站内文档 --check first, no forced order.
agent-toolchain-coexist · domain: Multi-Agent Orchestration · verified
Problem: Skills not visible after install / document-skills install does nothing
skills not visible after install — skills are directory copies: ls ~/.claude/skills/agent-skills first (expect 24 dirs), fully exit and restart the agent, then check permissions/version.
agent-skills-4 · domain: Skills Management · verified
Problem: Install troubleshooting: Node version / pnpm / npm cache
install fails — check Node version first, retry npm via the npmmirror registry (PONG 0.32s tested): npx -y skills add addyosmani/agent-skills.
agent-skills-2 · domain: Skills Management · verified
Problem: Official CLI install fails: GitHub clone reset/stuck (common in CN)
official CLI install fails on GitHub clone reset — use option B: fetch the codeload tarball via gh-proxy and copy manually (works in China, tested).
agent-skills-1 · domain: Skills Management · verified