Function query_ollama 

pub fn query_ollama(
    endpoint: &str,
    model: &str,
    prompt: &str,
    timeout_secs: u64,
) -> Result<String, SsgError>

Typed Ollama generation call backed by ureq (issue #520).

All HTTP traffic flows through ureq::post(...).send_json(...) — no subprocess is spawned, so prompts containing shell metacharacters ($(, backticks, ;, &, |) traverse the transport as a JSON body byte-for-byte unchanged (AC3).

Errors

• SsgError::LlmTimeout when the call exceeds timeout_secs.
• SsgError::LlmEndpointUnreachable when the TCP connection is refused, the host is unresolvable, or any other transport failure occurs before a response is received.
• SsgError::LlmInvalidResponse when the server returns a non-2xx status, the body is not valid JSON, or the JSON does not carry a non-empty response field.

Examples

use ssg::llm::query_ollama;
use ssg::SsgError;

// No Ollama running on port 1 ⇒ deterministic transport failure.
let err = query_ollama("http://127.0.0.1:1", "llama2", "hi", 1).unwrap_err();
assert!(matches!(err, SsgError::LlmEndpointUnreachable { .. } | SsgError::LlmTimeout { .. }));