Crate prompt_hash 

prompt-hash

Deterministic cache key for an LLM prompt.

Inputs:

• model — model id (already normalized by claude-cost/openai-cost if you care).
• messages — [(role, content)].
• temperature — quantized to 2 decimal places to avoid floating point cache misses.

Output: 64-char hex SHA-256.

Use as a request-level cache key. For semantic cache keys (matching prompts with the same meaning but different wording), pair with semantic-cache-key.

Example

use prompt_hash::key;
let k1 = key(
    "claude-sonnet-4-5",
    &[("user", "what is 2+2?")],
    1.0,
);
let k2 = key(
    "claude-sonnet-4-5",
    &[("user", "what is 2+2?   ")], // trailing ws ignored
    1.0,
);
assert_eq!(k1, k2);
assert_eq!(k1.len(), 64);

Functions

key

Compute a 64-char hex cache key.