Files
Loki/rust/loki-rs/tests/imagetools.rs
T
Claude db1f20c0a4 Lot 45 : outils d'image (see_image, web_screenshot) tenus par la façade
La façade exécute les deux outils qui montrent une image au modèle
(src/imagetools.rs, port de chat_vision_tool.go, chat_screenshot.go et
prepareImageForModel) : préparation de l'image comme les pièces jointes
(orientation EXIF, grand côté sous 1568 px, mêmes pixels), sonde du moteur
(/props authentifié, gardée 10 s ; preset externe : sa case), see_image aux
textes du Go, web_screenshot par Playwright (adresses locales quand internet
est coupé, mêmes arguments, erreur résumée, capture rangée dans le dossier
de la discussion, ménage à 20 captures et 40 Mo, relayée au modèle s'il
voit les images). Le Go garde les siens pour ses sous-agents et tâches IA.

Parité : journal tiré du Go avec ses fichiers, un faux moteur et un faux
Playwright (tests/parity_images/images.json), rejoué par
tests/imagetools.rs ; un tour de la façade sans passer par le Go
(tests/lot6.rs) ; tests unitaires.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016tcGfJWefW5kUGoC2wL8cv
2026-10-16 06:36:08 +00:00

287 lines
11 KiB
Rust

//! Lot 45 : outils d'image de l'agent (`see_image`, `web_screenshot`)
//! tenus par la façade, rejoués sur le journal du Go
//! (`internal/loki/parity_image_tools_test.go` →
//! `tests/parity_images/images.json`) : mêmes fichiers, même faux moteur
//! (`/props`), même faux Playwright. Une image laissée telle quelle se
//! compare à l'octet ; une image réencodée par ses dimensions et son type,
//! et une PNG réduite par ses pixels.
#![cfg(target_os = "linux")]
mod common;
use std::convert::Infallible;
use std::sync::{Arc, Mutex};
use base64::Engine as _;
use bytes::Bytes;
use common::fakego;
use http_body_util::Full;
use hyper::service::service_fn;
use hyper::{Request, Response};
use hyper_util::rt::TokioIo;
use loki_rs::imagetools::{self, VisionProbe};
use loki_rs::store::Store;
use regex::Regex;
use serde_json::{Map, Value, json};
use tokio::net::TcpListener;
const FAKE_PLAYWRIGHT: &str = r#"#!/bin/sh
d=$(dirname "$0")
printf '%s\n' "$@" > "$d/args"
for a in "$@"; do out="$a"; done
mode=$(cat "$d/mode" 2>/dev/null)
case "$mode" in
echec) echo "Launching chromium" ; echo "" ; echo "page.goto: net::ERR_NAME_NOT_RESOLVED at https://x" >&2 ; echo "Call log:" >&2 ; echo " - navigating" >&2 ; exit 1 ;;
vide) : > "$out" ;;
rien) ;;
*) cat "$d/image" > "$out" ;;
esac
"#;
/// Le faux moteur : `/props` authentifié, la vision réglée par le test.
async fn fake_engine(sees: Arc<Mutex<(bool, u32)>>) -> u16 {
let l = TcpListener::bind("127.0.0.1:0").await.unwrap();
let port = l.local_addr().unwrap().port();
tokio::spawn(async move {
while let Ok((st, _)) = l.accept().await {
let sees = sees.clone();
tokio::spawn(async move {
let svc = service_fn(move |req: Request<hyper::body::Incoming>| {
let sees = sees.clone();
async move {
let ok = req.uri().path() == "/props"
&& req
.headers()
.get("authorization")
.and_then(|h| h.to_str().ok())
== Some("Bearer cle-moteur");
let mut r = Response::new(Full::new(Bytes::new()));
if ok {
let mut s = sees.lock().unwrap();
s.1 += 1;
r = Response::new(Full::new(Bytes::from(
json!({ "modalities": { "vision": s.0 } }).to_string(),
)));
} else {
*r.status_mut() = hyper::StatusCode::UNAUTHORIZED;
}
Ok::<_, Infallible>(r)
}
});
let _ = hyper::server::conn::http1::Builder::new()
.serve_connection(TokioIo::new(st), svc)
.await;
});
}
});
port
}
/// `parityPartFacts`.
fn part_facts(part: &Value, raw: &Map<String, Value>) -> Value {
let u = part["image_url"]["url"].as_str().unwrap();
let i = u.find(";base64,").unwrap();
let mime = u[..i].trim_start_matches("data:");
let b = base64::engine::general_purpose::STANDARD
.decode(&u[i + 8..])
.unwrap();
let mut out = Map::new();
out.insert("mime".into(), json!(mime));
for (name, r) in raw {
let rb = base64::engine::general_purpose::STANDARD
.decode(r.as_str().unwrap_or(""))
.unwrap();
if rb == b {
out.insert("same_as".into(), json!(name));
return Value::Object(out);
}
}
let Ok(img) = image::load_from_memory(&b) else {
panic!("image relayée illisible");
};
let img = img.to_rgba8();
let (w, h) = img.dimensions();
out.insert("w".into(), json!(w));
out.insert("h".into(), json!(h));
if mime == "image/png" {
let mut px = Map::new();
for (x, y) in [(0, 0), (1, 1), (100, 50), (w - 1, h - 1), (w / 2, h / 3)] {
let p = img.get_pixel(x, y).0;
px.insert(
format!("{x},{y}"),
json!(format!("{},{},{},{}", p[0], p[1], p[2], p[3])),
);
}
out.insert("px".into(), Value::Object(px));
}
Value::Object(out)
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn outils_image_comme_le_go() {
let fx: Value =
serde_json::from_str(include_str!("parity_images/images.json")).expect("journal");
let tmp = tempfile::tempdir().unwrap();
let files = tmp.path().join("fichiers");
let raw = fx["files"].as_object().unwrap();
for (name, b) in raw {
let p = files.join(name);
std::fs::create_dir_all(p.parent().unwrap()).unwrap();
std::fs::write(
&p,
base64::engine::general_purpose::STANDARD
.decode(b.as_str().unwrap_or(""))
.unwrap(),
)
.unwrap();
}
std::fs::create_dir_all(files.join("dossier.png")).unwrap();
std::fs::File::create(files.join("lourd.png"))
.unwrap()
.set_len(13 << 20)
.unwrap();
let files_s = files.to_string_lossy().into_owned();
let sees = Arc::new(Mutex::new((false, 0u32)));
let port = fake_engine(sees.clone()).await;
let (_, go, _stop) = fakego::start(tmp.path().join("home")).await;
let backend = go.backend.lock().unwrap().unwrap();
let store = Store::new(backend, fakego::JETON.into());
go.set("state", "api_key", "cle-moteur");
let mut cfg = std::collections::HashMap::new();
cfg.insert("PORT".to_string(), port.to_string());
let probe = VisionProbe::default();
let pw_dir = tmp.path().join("pw");
std::fs::create_dir_all(&pw_dir).unwrap();
let pw = pw_dir.join("playwright");
std::fs::write(&pw, FAKE_PLAYWRIGHT).unwrap();
use std::os::unix::fs::PermissionsExt as _;
std::fs::set_permissions(&pw, std::fs::Permissions::from_mode(0o755)).unwrap();
let conv_dir = tmp.path().join("conv");
let cap_dir = conv_dir.join("captures");
let stamp_re = Regex::new(r"[0-9]{8}-[0-9]{6}").unwrap();
for (n, st) in fx["steps"].as_array().unwrap().iter().enumerate() {
let name = st["name"].as_str().unwrap();
*sees.lock().unwrap() = (st["sees"] == true, 0);
probe.reset().await;
std::fs::write(pw_dir.join("mode"), st["mode"].as_str().unwrap_or("")).unwrap();
if let Some(img) = st["image"].as_str() {
let b = base64::engine::general_purpose::STANDARD
.decode(raw[img].as_str().unwrap())
.unwrap();
std::fs::write(pw_dir.join("image"), b).unwrap();
}
let _ = std::fs::remove_file(pw_dir.join("args"));
if name == "ménage des captures" {
let old = std::time::SystemTime::now() - std::time::Duration::from_secs(3600);
for i in 0..22 {
let p = cap_dir.join(format!("vieille-{i:02}.jpg"));
std::fs::write(&p, "x").unwrap();
let f = std::fs::File::options().write(true).open(&p).unwrap();
f.set_modified(old + std::time::Duration::from_secs(i))
.unwrap();
}
}
let args: Map<String, Value> =
serde_json::from_str(&st["args"].to_string().replace("{{FICHIERS}}", &files_s))
.unwrap();
let label = st["label"]
.as_str()
.unwrap()
.replace("{{FICHIERS}}", &files_s);
let vision = st["vision"] == true;
let (mut result, part, shown) = match st["tool"].as_str().unwrap() {
"see_image" => {
let (r, img) =
imagetools::see_image(&label, &files, "", vision, probe.sees(&store, &cfg))
.await;
let relay = img.map(|i| loki_rs::chat::host::see_image_message(&label, i));
(r, relay, String::new())
}
_ => {
let stamp = format!("20260101-1200{:02}", n % 60);
let shot = imagetools::web_screenshot(
&args,
&pw.to_string_lossy(),
&conv_dir,
st["internet"] == true,
&stamp,
)
.await;
let relay = if !shot.rel.is_empty() && vision && probe.sees(&store, &cfg).await {
imagetools::screenshot_message(&shot.abs)
.and_then(|m| serde_json::from_value(m).ok())
} else {
None
};
(shot.result, relay, shot.rel)
}
};
let mut facts = None;
if let Some(m) = &relay_parts(&part) {
for p in m {
if p["type"] == "image_url" {
facts = Some(part_facts(p, raw));
}
if p["type"] == "text" {
result += &format!("\n[relais] {}", p["text"].as_str().unwrap());
}
}
}
let mask = |s: &str| {
stamp_re
.replace_all(&s.replace(&files_s, "{{FICHIERS}}"), "{{T}}")
.into_owned()
};
assert_eq!(mask(&result), st["result"].as_str().unwrap(), "{name}");
assert_eq!(facts, st.get("part").cloned(), "{name} : image relayée");
assert_eq!(
mask(&shown),
st["shown"].as_str().unwrap_or(""),
"{name} : capture montrée"
);
assert_eq!(
sees.lock().unwrap().1 as u64,
st["probes"].as_u64().unwrap(),
"{name} : sondes du moteur"
);
let pw_args: Vec<String> = std::fs::read_to_string(pw_dir.join("args"))
.map(|b| {
b.trim_end_matches('\n')
.split('\n')
.map(|l| {
mask(&l.replace(&cap_dir.to_string_lossy().into_owned(), "{{CAPTURES}}"))
})
.collect()
})
.unwrap_or_default();
assert_eq!(
json!(pw_args),
st.get("pw_args").cloned().unwrap_or(json!([])),
"{name} : Playwright"
);
if st["tool"] == "web_screenshot" {
let mut caps: Vec<String> = std::fs::read_dir(&cap_dir)
.into_iter()
.flatten()
.flatten()
.map(|e| mask(&e.file_name().to_string_lossy()))
.collect();
caps.sort();
assert_eq!(
json!(caps),
st.get("captures").cloned().unwrap_or(json!([])),
"{name} : dossier des captures"
);
}
}
}
/// Les parties d'un message relayé.
fn relay_parts(m: &Option<loki_rs::chat::types::Message>) -> Option<Vec<Value>> {
let m = m.as_ref()?;
m.content.as_ref()?.as_array().cloned()
}