A little clarity goes a long way
Common Questions
Something not working? Find the cause, follow the checks, and get back to building.
AI questions
Troubleshoot local models, document answers, MCP integrations and AI workflows.
- 3 min read
Why does my AI agent repeat tool calls and keep spending tokens?
Diagnose repeated AI tool calls, define observable progress and set application-enforced limits without mistaking a higher step cap for a fix.
Read the answer - 3 min read
Why does AI transcription invent words during silence?
Trace invented transcript text to silence, segmentation or earlier context, then test speech detection without removing quiet real speech.
Read the answer - 3 min read
Do I need to reindex my documents after changing the embedding model?
Plan an embedding-model change with a separate index, paired retrieval tests and rollback instead of mixing old document vectors with new queries.
Read the answer - 3 min read
How do I choose a smaller LLM quantization without losing the quality I need?
Compare smaller quantizations using your own failure cases, memory limits and correction effort instead of assuming one format always wins.
Read the answer - 3 min read
How can I check whether my local AI workflow really works offline?
Audit generation, embeddings, OCR and tools, then test a complete offline workflow without assuming that a local model makes every step local.
Read the answer - 3 min read
Why is my MCP server connected but no tools appear?
Diagnose an MCP connection with missing tools by checking discovery, schema acceptance, permissions and host exposure in order.
Read the answer - 3 min read
Why does MCP login succeed but tool calls return unauthorized?
Trace MCP unauthorized errors through token audience, scopes, expiry and downstream credentials after a browser login appears successful.
Read the answer - 3 min readIllustrated guide
Why does my MCP server report invalid JSON after I add logging?
Find startup banners and debug logs contaminating MCP stdio messages, then separate protocol output from diagnostics safely.
Read the answer - 3 min read
Why does an MCP tool work in one client but fail in another?
Compare MCP tool schemas and identical inputs across clients to isolate compatibility failures without disabling validation or broadening access.
Read the answer - 3 min read
Why does RAG find English documents but miss French or Arabic ones?
Test multilingual RAG at extraction, retrieval and answer stages, using paired questions instead of assuming fluent output proves good search.
Read the answer - 3 min read
Why does Ollama run out of memory in a longer chat?
Find why a local model fits initially but fails in longer chats, using a context and concurrency test that separates memory causes.
Read the answer - 3 min read
Why is the first Ollama response slow but later ones fast?
Separate model loading from prompt processing and streaming delays to diagnose why the first local AI response takes much longer.
Read the answer - 3 min read
Why is my AI output valid JSON but still wrong for my application?
Validate AI output in three stages: JSON syntax, schema and real meaning, with explicit handling for missing facts and failed validation.
Read the answer - 3 min readIllustrated guide
Why is Ollama using my CPU instead of my GPU?
Diagnose Ollama CPU fallback by checking model placement, GPU visibility, context and memory before changing drivers or buying hardware.
Read the answer - 3 min read
Why does my AI answer include citations that do not support it?
Check whether each AI citation supports its exact claim, and distinguish retrieval references, page offsets and unsupported conclusions.
Read the answer - 3 min read
Why does my AI assistant still mention a document I removed?
Trace old document facts through chat history, retrieval collections and caches to understand why removed material still influences answers.
Read the answer - 3 min read
Can a document tell my AI agent to ignore instructions or use tools?
Use a harmless document fixture to test prompt-injection handling and enforce tool permissions outside the model's interpretation of text.
Read the answer - 3 min readIllustrated guide
Why does RAG say it cannot find an answer in a document I uploaded?
Trace a missing RAG answer through extraction, indexing, selection and retrieval instead of repeatedly uploading the same document.
Read the answer - 3 min readIllustrated guide
Why can my AI assistant read one PDF but not a scanned PDF?
Diagnose scanned-PDF failures by checking OCR, reading order, indexing and retrieval before changing prompts or replacing the language model.
Read the answer - 3 min readIllustrated guide
Why does my RAG assistant get table totals wrong?
Find whether incorrect AI table totals come from parsing, missing rows or arithmetic, then choose a retrieval or calculation workflow.
Read the answer
Software questions
Diagnose Docker, networking, webhooks, authentication and development errors.
- 3 min read
Why does AI streaming arrive all at once behind a reverse proxy?
Trace incremental AI output from the application through the proxy to the browser, separating buffering from idle timeouts and rendering delays.
Read the answer - 3 min read
I committed an API key to Git. Is deleting the file enough?
Respond to a committed API key by revoking or rotating it, checking use and cleaning history without confusing file deletion with containment.
Read the answer - 3 min read
An API request timed out. Is it safe to retry?
Decide whether to retry a timed-out API operation by checking side effects, idempotency support and the remote outcome before sending it again.
Read the answer - 3 min read
Why does my API work in curl but fail in the browser?
Diagnose browser-only API failures through origins, preflight requests and response headers without using wildcard credentials or no-cors as a fix.
Read the answer - 3 min read
Why can my container read a mounted folder but not write to it?
Diagnose Docker bind-mount write failures through mount mode, user identity and host policy without making the whole folder writable to everyone.
Read the answer - 3 min read
Why did my Docker Compose environment variable not change after restart?
Trace Docker Compose variables from interpolation to the running container and understand when configuration changes require recreation.
Read the answer - 3 min readIllustrated guide
Why can my Docker container not reach an API running on my laptop?
Map container-to-host connections correctly and separate hostname, port and bind-address failures when an API works only on your laptop.
Read the answer - 3 min read
Why does my app fail at startup even with Docker depends_on?
Separate container startup from database readiness, then add meaningful health checks and bounded reconnection for later outages.
Read the answer - 3 min read
Why is Docker filling my disk even when my database is small?
Identify whether Docker disk growth comes from logs, images, build cache or persistent data before choosing a targeted retention policy.
Read the answer - 3 min read
Is my Docker development server exposed to other devices?
Check application binding, Docker port publishing and firewall rules to determine whether a development service is reachable beyond your machine.
Read the answer - 3 min read
Why does my Docker app look empty after I rename its project folder?
Investigate Docker data that appears missing after a project rename by checking volume identity, mount targets and application configuration.
Read the answer - 3 min read
Why does login work locally but the cookie disappears in production?
Trace a production login cookie through response headers, browser storage and later requests to isolate domain, Secure and SameSite problems.
Read the answer - 3 min read
Why does NGINX return 502 when my application container is running?
Diagnose a 502 by checking the upstream address, listener, protocol and logs from the proxy's environment rather than relying on container status.
Read the answer - 3 min read
Why does npm ci fail in CI when npm install works locally?
Separate lockfile mismatch from platform, registry and lifecycle-script failures when npm ci breaks in CI but local installation succeeds.
Read the answer - 3 min read
How do I know my PostgreSQL backup will actually restore?
Verify PostgreSQL recovery with an isolated restore, representative data checks and application tests instead of relying on a successful backup log.
Read the answer - 3 min read
Why did my scheduled job run twice, and how do I prevent duplicate work?
Investigate duplicate scheduled work through scheduler identity, overlap, retries and time zones, then protect the business operation from repetition.
Read the answer - 3 min readIllustrated guide
Why does my webhook process the same event twice?
Handle duplicate webhook delivery with durable event tracking and idempotent effects, including the crash window between processing and acknowledgment.
Read the answer - 3 min readIllustrated guide
How should my app handle webhooks that arrive out of order?
Handle delayed webhook events without moving application state backward, using provider state, versions and explicit transition rules.
Read the answer - 3 min read
Why are npm and Git slow when my WSL project is under /mnt/c?
Compare WSL project locations, executable paths and file operations to diagnose slow npm or Git without assuming the laptop is underpowered.
Read the answer - 3 min read
Why does the internet work in Windows but fail in WSL on a VPN?
Separate WSL DNS, routing and certificate failures when Windows works on a VPN, using a controlled comparison and supported network settings.
Read the answer