mindfork 0.19.0

A terminal AI chat written in Rust: local models via llama.cpp, or OpenAI, Anthropic, Gemini, Grok and OpenRouter in the cloud, with persistent memory, notes, RAG and tools.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

//! The Gemini arm of the cache probe (docs/research/cached-tokens.md §7): the
//! client's own body, sent raw so the `cachedContentTokenCount` the client
//! drops can be read — and whether a miss is a zero or an absent key.
//!
//! `MINDFORK_GEMINI_KEY=… [MINDFORK_CACHE_PROBE_GEMINI_MODELS=a,b] cargo test
//! gemini::cache_probe -- --ignored --nocapture --test-threads=1`

use super::wire::build_request;
use crate::shared::api::cache_probe_tests::{models, probe_models};

/// The newest Flash (a 4096-token minimum) and the 2.5 generation (2048).
const MODELS: &str = "gemini-3.8-flash,gemini-2.5-flash";

#[tokio::test]
#[ignore = "requires MINDFORK_GEMINI_KEY"]
async fn cache_probe_gemini_live() {
    let Ok(key) = std::env::var("MINDFORK_GEMINI_KEY") else {
        eprintln!("skip: MINDFORK_GEMINI_KEY not set");
        return;
    };
    let http = reqwest::Client::new();
    let models = models("MINDFORK_CACHE_PROBE_GEMINI_MODELS", MODELS);
    probe_models("gemini", &models, |model, req| {
        http.post(format!(
            "https://generativelanguage.googleapis.com/v1beta/models/{model}:streamGenerateContent?alt=sse"
        ))
        .header("x-goog-api-key", &key)
        .json(&build_request(req, model))
    })
    .await;
}