consciousness/poc-agent/src/tools/vision.rs

150 lines
4.6 KiB
Rust
Raw Normal View History

// tools/vision.rs — Image viewing tool
//
// Reads image files from disk and returns them as base64 data URIs
// for multimodal models. Also supports capturing tmux pane contents
// as screenshots.
use anyhow::{Context, Result};
use base64::Engine;
use serde::Deserialize;
use super::ToolOutput;
use crate::types::ToolDef;
#[derive(Deserialize)]
struct Args {
file_path: Option<String>,
pane_id: Option<String>,
#[serde(default = "default_lines")]
lines: usize,
}
fn default_lines() -> usize { 50 }
pub fn definition() -> ToolDef {
ToolDef::new(
"view_image",
"View an image file or capture a tmux pane screenshot. \
Returns the image to your visual input so you can see it. \
Supports PNG, JPEG, GIF, WebP files. \
Use pane_id (e.g. '0:1.0') to capture a tmux pane instead.",
serde_json::json!({
"type": "object",
"properties": {
"file_path": {
"type": "string",
"description": "Path to an image file (PNG, JPEG, GIF, WebP)"
},
"pane_id": {
"type": "string",
"description": "Tmux pane ID to capture (e.g. '0:1.0'). Alternative to file_path."
},
"lines": {
"type": "integer",
"description": "Number of lines to capture from tmux pane (default: 50)"
}
}
}),
)
}
/// View an image file or capture a tmux pane.
pub fn view_image(args: &serde_json::Value) -> Result<ToolOutput> {
let a: Args = serde_json::from_value(args.clone())
.context("invalid view_image arguments")?;
if let Some(ref pane_id) = a.pane_id {
return capture_tmux_pane(pane_id, a.lines);
}
let file_path = a.file_path
.as_deref()
.context("view_image requires either file_path or pane_id")?;
let path = std::path::Path::new(file_path);
if !path.exists() {
anyhow::bail!("File not found: {}", file_path);
}
let data = std::fs::read(path).with_context(|| format!("Failed to read {}", file_path))?;
// Sanity check file size (don't send huge images)
const MAX_SIZE: usize = 20 * 1024 * 1024; // 20 MB
if data.len() > MAX_SIZE {
anyhow::bail!(
"Image too large: {} bytes (max {} MB)",
data.len(),
MAX_SIZE / (1024 * 1024)
);
}
let mime = mime_from_extension(path);
let b64 = base64::engine::general_purpose::STANDARD.encode(&data);
let data_uri = format!("data:{};base64,{}", mime, b64);
Ok(ToolOutput {
text: format!(
"Image loaded: {} ({}, {} bytes)",
file_path,
mime,
data.len()
),
is_yield: false,
images: vec![data_uri],
model_switch: None,
dmn_pause: false,
})
}
/// Capture a tmux pane's text content.
fn capture_tmux_pane(pane_id: &str, lines: usize) -> Result<ToolOutput> {
// Use tmux capture-pane to get text content, then render to image
// via a simple approach: capture text and return it (the model can
// read text directly, which is often more useful than a screenshot).
//
// For actual pixel-level screenshots we'd need a terminal renderer,
// but text capture covers 95% of use cases.
let output = std::process::Command::new("tmux")
.args(["capture-pane", "-t", pane_id, "-p", "-S", &format!("-{}", lines)])
.output()
.context("Failed to run tmux capture-pane")?;
if !output.status.success() {
let stderr = String::from_utf8_lossy(&output.stderr);
anyhow::bail!("tmux capture-pane failed: {}", stderr.trim());
}
let text = String::from_utf8_lossy(&output.stdout).to_string();
// Return as text — the model can read terminal output directly.
// This is actually more useful than a screenshot for most tasks.
Ok(ToolOutput {
text: format!(
"Tmux pane {} (last {} lines):\n```\n{}\n```",
pane_id, lines, text.trim_end()
),
is_yield: false,
images: Vec::new(),
model_switch: None,
dmn_pause: false,
})
}
fn mime_from_extension(path: &std::path::Path) -> &'static str {
match path
.extension()
.and_then(|e| e.to_str())
.map(|e| e.to_lowercase())
.as_deref()
{
Some("png") => "image/png",
Some("jpg" | "jpeg") => "image/jpeg",
Some("gif") => "image/gif",
Some("webp") => "image/webp",
Some("svg") => "image/svg+xml",
Some("bmp") => "image/bmp",
_ => "image/png", // default assumption
}
}