An AI assistant platform running Opus 4.6 hit stream_stall (90s timeout) twice consecutively when generating docx. Root cause: skill instructions lacked one sentence — 'Write a script' — causing the model to output JS code inline instead of writing a file and executing with node. Claude.ai's official SKILL.md has that sentence, and the model consistently takes the safe path.
An AI assistant platform's image search needed three rounds of fixes: pushing node_type filtering into the ES query to stop text chunks from hogging top-k slots, making filename matching deterministic instead of relying on the LLM to pass an optional parameter, and switching from Postgres icontains to ES match for CJK-aware partial matching.
About 10% of conversations on our AI assistant platform randomly lost knowledge base tools — the agent hallucinated answers from training data instead. Root cause: a bare except Exception swallowed Elasticsearch connection failures during retriever initialization, silently skipping tool registration. Fix: retry + surface failures to system prompt + structured metadata tracking.
A user set retrieval top_k to 15, but the monitoring dashboard showed 5 and the streaming UI flashed 5 before jumping to 15. The same top_k value existed at four layers — LLM tool arguments, runtime, trace DB, and streaming payload — each requiring its own override. Three sequential fixes, each revealing the next layer was also wrong.
Parsing a 150-page PDF via Vision API took 29 minutes (one page per request). Batching cut it to 1.5×, adding 5-way concurrency brought it down to 24 seconds. One week after launch, a customer uploaded 23 PDFs at once — 70 parallel Bedrock requests triggered full throttling: 90 pages skipped, 5 files failed, 8 stuck. Fixed with Redis-based cluster-wide slots + backoff retries.
Routines created by Claude itself (created_via: meta_mcp) prompt for connector approval on every call. User-created ones (created_via: http_api) don't. Fix: recreate the routine via the RemoteTrigger API.
Training neural nets is closer to engineering than magic: look at the data, overfit a mini-batch to prove capacity exists, then regularize back the generalization; learning rate is always the highest-leverage knob.
chrome-devtools-mcp, maintained by the Chrome team, packages DevTools capability as an MCP server: performance traces and insights, Lighthouse audits, heap snapshots, extension management — none of which @playwright/mcp exposes. It runs on Puppeteer, so interactions auto-wait; the costs are Chrome-only support and usage statistics reported to Google by default.
When mobile Chrome keeps redirecting back to the login page after sign-in, the culprit isn't always OAuth or broken frontend state. In this case, the root cause was that the HTTP entry point for app-dev.daodao.so wasn't issuing a 301 redirect to HTTPS, so /auth/me requests sent with an http origin didn't include the auth_token cookie.
A hands-on log of installing Superpowers (packaged by DwainTR) for Copilot CLI on a local machine — including the diagnostic process when skills didn't appear after installation, the fix, and practical tips.
A brief error during nginx restart caused Cloudflare to mark the origin as unhealthy and stop forwarding requests, returning 502 on its own. The key clues: localhost hits to the origin return 200, and nginx access logs are completely empty. Just wait for Cloudflare to automatically re-check the origin — it recovers on its own.
When a RAG system breaks, 90% of the time it's one of these 10 failure modes. Identify which one first, then apply the matching fix — far more effective than optimizing blindly.
The hardest part of a RAG system isn't building it — it's figuring out why a particular answer went wrong. Pipeline Tracing records every step's decisions and data so debugging has a clear trail to follow.