Build Your First Local AI App with the Ollama API

Confirm that the server and model work with curl before adding a framework. The local and cloud API base URLs are different, so verify that sensitive prompts point to localhost.

The working rule for “Make the smallest request” is: The default local base URL is `http://localhost:11434/api`. Exercise invalid input and interruption paths as well as the happy path, because application boundaries are where a working example most often fails.

After running the command or code, inspect the exit status, logs, and the file, process, or response it was meant to create. If it fails, change one input, version, permission, or resource condition at a time and repeat the same check so that the cause remains attributable.

curl http://localhost:11434/api/generate -d '{
  "model": "gemma3:4b",
  "prompt": "Explain local AI in one sentence.",
  "stream": false
}'

The chat endpoint accepts role-based messages. Your application should send the history required for each request rather than assuming the server remembers every conversation.

Summarize old turns or begin a new session when context becomes unnecessarily long.

Preserve the before-and-after state and the time of the check so that another run can reproduce the result. Include at least one failure condition—such as empty input, constrained resources, or a restart—to reveal the boundary of the step rather than documenting only the happy path.

Streaming improves perceived latency but requires the client to assemble chunks, handle cancellation, and surface mid-stream failures.

Begin with a complete non-streaming response, validate the shape, and add streaming only when the basic flow is reliable.

Define completion with an observable result instead of a general impression. Repeat the same input, and if the output changes, isolate whether the model, runtime settings, or source data changed before moving to the next stage.

Structured Outputs can constrain a response with a JSON schema. A lower temperature can improve repeatability.

Still validate parsing, required fields, value ranges, and factual claims against the source data.

Use the checklist with a date and observed result beside every item. When one check fails, record its scope before continuing, then repeat the same input after the change; that turns a list of advice into evidence that the step actually works in this environment.

• Define required fields
• Limit retries
• Validate ranges and lengths
• Log validation errors without storing sensitive prompts
• Provide a human-review path

Handle a stopped server, missing model, timeout, oversized request, user cancellation, and invalid JSON. Do not retry indefinitely.

Keep the server on localhost unless you have designed authentication, TLS, access controls, monitoring, and rate limits.

Preserve the before-and-after state and the time of the check so that another run can reproduce the result. Include at least one failure condition—such as empty input, constrained resources, or a restart—to reveal the boundary of the step rather than documenting only the happy path.