Configure Ollama API payloads and model discovery for local inference.
npx skills add https://github.com/berad217/human-training --skill ollama-berad217
Please help me install this Agent Skill. Skill: ollama Source: https://github.com/berad217/human-training/tree/main/skills-drafts/ollama Command: npx skills add https://github.com/berad217/human-training --skill ollama-berad217
Writing code against a local Ollama daemon frequently encounters undocumented footguns including hallucinated model names, broken vision payloads, incorrect structured output modes, and misconfigured sampling parameters that cause empty responses, wasted GPU resources, and unpredictable inference behavior.
/api/tags endpoint to avoid referencing uninstalled or incorrectly named models.num_predict values with image inputs.Use the ollama skill to write a Python script that safely queries a local Ollama vision model to describe an image, using the correct payload format to avoid empty responses.
High-intent search queries and answers about installing and using this skill.
Ollama vision models frequently return empty responses when using legacy JSON format or positive num_predict values with image inputs. You can fix this by applying correct payload formatting and verified workarounds for these pipeline bugs.
To prevent hallucinated model names in Ollama code, perform discovery-first model selection by automatically verifying installed models via the /api/tags endpoint before referencing any model names in your application logic.
To generate deterministic structured output from a local LLM using Ollama, apply safe default guidance for structured output modes and sampling parameters to ensure reproducible local LLM evaluation and benchmarking tasks.
Common Ollama sampling parameter footguns include misconfigured values that cause unpredictable inference behavior and wasted GPU resources. You can avoid these by enforcing safe defaults for sampling parameters and thinking model configuration.
Yes, you can use the OpenAI-compatible API shim with a local Ollama daemon. Proper shim configuration is supported to ensure reliable runtime behavior when integrating local LLM inference into existing OpenAI-compatible workflows.
To safely build a local image captioning pipeline with Ollama, use the correct payload formatting for vision endpoints to avoid empty responses, ensuring the vision model reliably returns captions without wasting GPU time.
Cogni-AI-OUManage local LLM lifecycle by listing, pulling, and running models via the ollama CLI.
tiankong0101-byteAnalyze, edit, generate, and convert images using Python scripts.
TehminanazIntegrate the OpenAI Agents SDK with Groq's endpoint in FastAPI applications.
JuanMS20Deploy high-throughput LLM serving endpoints with vLLM and OpenAI-compatible APIs.