Error category
Troubleshoot Ollama local server, model load, cloud API, timeout, runner crash, memory, GPU, and connection errors for local and hosted inference.
This Ollama troubleshooting hub collects real error signatures, quick fixes, common causes, and step-by-step debugging paths for developers who need a practical answer instead of a broad overview.
106 articles in this category.
Start by matching the exact error message, then check the technology, environment, credentials, network path, and deployment context. The pages below are grouped so you can move from broad Ollama symptoms to specific root-cause families without relying on client-side search.
Fix LangChain + Ollama integration where streaming completions incorrectly use stop_sequences as finish_reason, causing early termination or garbled output in production LLM pipelines. Includes evidence for Ollama troubleshooting demand.
OllamaLLM streaming in LangChain returns stop sequences as finish_reason instead of proper completion signal, breaking streaming pipelines Developers running local Ollama models via Codex App notice extreme slowdowns; root cause is Codex App not reading model parameter num_ctx and defaulting to max context sizes Includes evidence for Ollama troubleshooting demand.
Codex App sends requests with context_window=128000-262144 tokens instead of model's configured num_ctx (e.g., 32768) — causing severe generation slowdowns on local models Fix model download failure showing EOF error when Ollama run out of disk space during pull Includes evidence for Ollama troubleshooting demand.
pulling manifest Error: EOF Fix Ollama model returning empty string response from generate() API call despite model being installed Includes evidence for Ollama troubleshooting demand.
Ollama model returning empty string from ollama.generate() — model installed but generate() returns empty output Fix MCP server connection failure between VS Code Continue extension and Ollama running in Docker on Windows WSL2 Includes evidence for Ollama troubleshooting demand.
MCP server connection failed — Continue extension does not show or trigger MCP tools when Ollama runs in Docker on WSL2 Fix Ollama ignoring http_proxy environment variable when downloading models that use plain HTTP manifests, causing connection refused behind proxies Includes evidence for Ollama troubleshooting demand.
request failed: Get "http://registry.ollama.ai/v2/...": dial tcp: connect: connection refused Fix Ollama model not supporting tools error — resolve invalid_request_error for tool-calling with local models Includes evidence for Ollama troubleshooting demand.
BadRequestError: Error code: 400 — registry.ollama.ai/library/llama3:latest does not support tools Fix Ollama server crashing with OOM during GGUF metadata parsing after upgrading to v0.24.0 Includes evidence for Ollama troubleshooting demand.
runtime: out of memory during GGUF metadata parsing (readString/n_kv allocation) in Ollama v0.24.0 Fix Ollama v0.24.0 crash on startup due to OOM during GGUF metadata parsing with corrupted model headers Includes evidence for Ollama troubleshooting demand.
runtime: out of memory — readString/n_kv allocation in fs/gguf.readString tries to allocate 32GB+ RAM on corrupted model header Fix Ollama codex-app browser trust error: launching codex-app via ollama breaks built-in browser with pipe bridge trust failure Includes evidence for Ollama troubleshooting demand.
Error: privileged native pipe bridge is not available; browser-client is not trusted Fix Ollama MLX runner infinite repetition loop hang during large-context inference with no detection or circuit breaker Includes evidence for Ollama troubleshooting demand.
Ollama MLX runner enters repetition loop at 60K+ tokens context, no abort mechanism, streams same phrase hundreds of times indefinitely Fix Ollama gemma4 repetition loop when generating structured JSON with format parameter Includes evidence for Ollama troubleshooting demand.
Word repetition loop fills num_predict budget, JSON unterminated — repeat_penalty has no effect Fix Ollama structured output being ignored when think=false is set Includes evidence for Ollama troubleshooting demand.
think=false with format parameter causes format constraint to be silently ignored — outputs plain text instead of JSON Fix Ollama embedding endpoint returning 500 Internal Server Error with EOF when generating embeddings Includes evidence for Ollama troubleshooting demand.
Server error: POST http://localhost:11434/api/embeddings resulted in a 500 Internal Server Error — EOF fix Ollama 500 Internal Server Error model failed to load gemma4 Apple M5 exit status 2 Includes evidence for Ollama troubleshooting demand.
500 Internal Server Error: model failed to load, this may be due to resource limitations or an internal error fix Ollama gemma4 tool call parsing failed invalid character error Includes evidence for Ollama troubleshooting demand.
gemma4 tool call parsing failed error="invalid character" Fix Ollama 500 timed out waiting for llama-server to start on ROCm after upgrade Includes evidence for Ollama troubleshooting demand.
timed out waiting for llama-server to start Fix Ollama GGML_ASSERT(buffer) failed crash when loading multimodal models with CUDA Includes evidence for Ollama troubleshooting demand.
GGML_ASSERT(buffer) failed during loading of multimodal model Fix Ollama 400 unexpected EOF error when using cloud proxy models via /v1/chat/completions Includes evidence for Ollama troubleshooting demand.
Ollama /v1/chat/completions returns 400 'unexpected EOF' intermittently for cloud proxy models fix ollama 500 error unable to load model after successful pull Includes evidence for Ollama troubleshooting demand.
500 Internal Server Error: unable to load model: <path> Continue through the full static archive for this hub. Every listed page is crawlable and keeps its existing canonical URL.