ollama-api-skill

Practical guidance for integrating with Ollama's REST API and building agents that dynamically detect and use model capabilities.

What this skill covers

• Native REST API (/api/chat, /api/generate, /api/embed, /api/show, /api/ps, etc.)
• Capability detection via /api/show with confirmed, inferred, and pending states
• Model routing patterns for vision, audio, embeddings, and main tasks
• Streaming & structured outputs (JSON schema, tool calling, thinking)
• OpenAI compatibility layer (/v1/ endpoints)
• Multimodal inputs (images, audio) via base64
• Image generation (diffusion) via redirection, SSE streams, and OS compatibility handling
• Troubleshooting common issues (GPU, remote access, proxy, CORS, MLX runners)

Structure

File	Purpose
ollama-api/SKILL.md	Entry point: when to use, quick reference, troubleshooting, resources
ollama-api/references/api-reference.md	Complete endpoint docs, request/response schemas, error codes
ollama-api/references/capabilities.md	Detecting model capabilities from metadata and projector info
ollama-api/references/image-models.md	Diffusion-based image generation, payload options, SSE response format, and compatibility
ollama-api/references/examples.md	Code snippets in cURL, Go, Python, JavaScript/TypeScript
ollama-api/references/cloud.md	Cloud models, web search API, authentication, and IDE integrations
ollama-api/references/modelfile-capability-unlock.md	Modelfile instructions on how Ollama detects and unlocks model capabilities

When to use this skill

• Building an application that calls Ollama locally or remotely.
• Detecting which capabilities a model supports (completion, tools, vision, audio, embeddings, thinking, image).
• Implementing chat with streaming, tool calling, structured outputs, multimodal inputs, or image generation.
• Designing a model router that assigns different models to different tasks.
• Handling graceful fallbacks when a capability is unavailable.