teapot-coding-agent 0.37.68 → 0.37.70

package/dist/agent/agent.js

@@ -1850,7 +1850,26 @@ export class Agent {
                     durationMs: Date.now() - t0,
                     result: capped,
                 });
-                this.messages.push({ role: "tool", tool_call_id: call.id, content: capped });
+                /**
+                 * A tool may hand the model an IMAGE (read_file on a picture).
+                 *
+                 * It rides on the tool message as `content_parts`, which `sanitize()` already knows how to send
+                 * to a vision model. The text stays in `content` so the log, the timeline and a non-vision model
+                 * all still see something meaningful.
+                 */
+                this.messages.push({
+                    role: "tool",
+                    tool_call_id: call.id,
+                    content: capped,
+                    ...(result.image
+                        ? {
+                            content_parts: [
+                                { type: "text", text: capped },
+                                { type: "image_url", image_url: { url: result.image.url } },
+                            ],
+                        }
+                        : {}),
+                });
                 this.consecutiveToolErrors = result.ok ? 0 : this.consecutiveToolErrors + 1;
                 if (this.consecutiveToolErrors >= this.opts.maxConsecutiveToolErrors) {
                     await this.log.append("system_note", this.currentSession, this.currentBranch, {

package/dist/agent/tools.js

@@ -32,6 +32,13 @@ export async function looksBinary(p) {
         await fh.close();
     }
 }
+/** the `looksBinary()` kinds that are pictures a vision model can actually receive */
+const IMAGE_MIME = {
+    "PNG image": "image/png",
+    "JPEG image": "image/jpeg",
+    "GIF image": "image/gif",
+    "WEBP image": "image/webp",
+};
 function sniff(head) {
     const hex = (n) => head.subarray(0, n).toString("hex");
     if (hex(8).startsWith("89504e47"))
@@ -40,6 +47,9 @@ function sniff(head) {
         return "JPEG image";
     if (head.subarray(0, 4).toString("latin1") === "GIF8")
         return "GIF image";
+    // RIFF....WEBP — the four bytes after "RIFF" are the size, so "WEBP" sits at offset 8
+    if (head.subarray(0, 4).toString("latin1") === "RIFF" && head.subarray(8, 12).toString("latin1") === "WEBP")
+        return "WEBP image";
     if (head.subarray(0, 4).toString("latin1") === "PK\x03\x04")
         return "zip archive";
     if (hex(4) === "25504446")
@@ -979,7 +989,8 @@ export function lineIdMode(args) {
 export const TOOLS = [
     {
         name: "read_file",
-        description: "Read a text file from the workspace. Returns numbered lines (`N| ` prefixes are display-only — never copy them into edit_file). " +
+        description: "Read a file from the workspace. Returns numbered lines (`N| ` prefixes are display-only — never copy them into edit_file). " +
+            "On an IMAGE (png/jpeg/gif/webp) the picture itself is attached to the result, so you can see it — do not decode bytes. " +
             "With `line_ids:\"hash\"` it instead returns `LINE:HASH| content` anchors (the hashline format used by other coding harnesses), " +
             "which edit_file accepts VERBATIM and re-verifies against the file — so prefer it when you intend to edit what you just read. " +
             "Pass `paths` (max 10) to read SEVERAL files in one call instead of one call per file; " +
@@ -1045,6 +1056,27 @@ export const TOOLS = [
             try {
                 const probe = await looksBinary(p);
                 if (probe.binary) {
+                    /**
+                     * An IMAGE is handed to the model as an image; anything else is still refused.
+                     *
+                     * Refusing images was never a feature — it was the fix for returning raw bytes as mangled
+                     * text (`d907af3`). But the operator's point is the asymmetry: a user can attach a picture
+                     * and the agent cannot, so the agent is asked to work on files it is not allowed to see.
+                     * MEASURED: `read_file` had never been able to return an image at all; the multimodal commit
+                     * (`257333a`) was the USER path (composer → `/prompt`), not a tool.
+                     */
+                    const mime = IMAGE_MIME[probe.kind];
+                    if (mime) {
+                        const buf = await fs.readFile(p);
+                        const name = path.basename(p);
+                        return {
+                            ok: true,
+                            result: `${str(args.path)} is a ${probe.kind} (${probe.bytes} bytes) — the image itself is ` +
+                                `attached to this message, so you are looking at it. Describe or reason about it ` +
+                                `directly; there is no need to decode bytes.`,
+                            image: { url: `data:${mime};base64,${buf.toString("base64")}`, name },
+                        };
+                    }
                     return {
                         ok: false,
                         result: `${str(args.path)} is a ${probe.kind} (${probe.bytes} bytes), not text — read_file returns ` +

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "teapot-coding-agent",
-  "version": "0.37.68",
+  "version": "0.37.70",
   "description": "A lightweight, always-on multi-agent harness for AI coding agents",
   "type": "module",
   "license": "AGPL-3.0-or-later",