@blockrun/franklin 3.15.48 → 3.15.50

package/dist/tools/imagegen.d.ts

@@ -32,3 +32,32 @@ export interface ImageGenDeps {
 export declare function createImageGenCapability(deps?: ImageGenDeps): CapabilityHandler;
 /** Back-compat static capability for callers that don't want the Content bridge. */
 export declare const imageGenCapability: CapabilityHandler;
+export interface ImagePollBody {
+    data?: {
+        b64_json?: string;
+        url?: string;
+        revised_prompt?: string;
+    }[];
+    error?: unknown;
+    status?: string;
+}
+export type ImagePollOutcome = {
+    kind: 'completed';
+    body: ImagePollBody;
+} | {
+    kind: 'failed';
+    error?: unknown;
+} | {
+    kind: 'timed_out';
+} | {
+    kind: 'poll_http_error';
+    status: number;
+    bodyPreview: string;
+};
+export interface PollImageJobOptions {
+    /** Total wall-clock ceiling. Defaults to 5 min (matches videogen scale). */
+    maxWaitMs?: number;
+    /** Sleep between polls. Defaults to 3 s. */
+    intervalMs?: number;
+}
+export declare function pollImageJob(pollEndpoint: string, headers: Record<string, string>, signal: AbortSignal, options?: PollImageJobOptions): Promise<ImagePollOutcome>;

package/dist/tools/imagegen.js

@@ -264,11 +264,9 @@ function buildExecute(deps) {
             // when the upstream image model takes longer than the inline budget
             // (gpt-image-1/-2 routinely exceed 30s). Verified 2026-05-04 from
             // Cloud Run logs — five back-to-back ImageGen calls that the agent
-            // saw as "No image data returned from API" had all returned 202 +
-            // queued status; payment was settled but the result was still
-            // generating async. Without polling here, the agent burns retries
-            // (each charged) and the image silently completes elsewhere with
-            // no way for Franklin to retrieve it. Mirror videogen.ts's
+            // saw as "No image data returned from API" had all returned 202;
+            // 4 of 5 actually completed in GCS within 41–56s and would have
+            // been retrievable if Franklin had polled. Mirror videogen.ts's
             // pollUntilReady contract: same x-payment header on each poll.
             if (response.status === 202 && result.poll_url) {
                 const origin = new URL(apiUrl).origin;
@@ -280,44 +278,24 @@ function buildExecute(deps) {
                 // generation routinely completes within 1–3 min once queued; the
                 // 5 min ceiling matches videogen's POLL_MAX_WAIT_MS scale.
                 clearTimeout(timeout);
-                const pollDeadline = Date.now() + 5 * 60 * 1000;
-                let polled = null;
-                while (Date.now() < pollDeadline) {
-                    if (controller.signal.aborted)
-                        throw new Error('aborted');
-                    await new Promise(r => setTimeout(r, 3_000));
-                    const pollResp = await fetch(pollEndpoint, {
-                        method: 'GET',
-                        headers: pollHeaders,
-                        signal: controller.signal,
-                    });
-                    if (pollResp.status === 202)
-                        continue; // still queued
-                    if (pollResp.status === 429 || pollResp.status >= 500)
-                        continue; // transient
-                    if (pollResp.ok) {
-                        polled = await pollResp.json().catch(() => null);
-                        if (polled && (polled.status === 'completed' || polled.data?.[0]))
-                            break;
-                        if (polled && polled.status === 'failed') {
-                            return {
-                                output: `Image generation failed upstream: ${JSON.stringify(polled.error ?? polled).slice(0, 240)}`,
-                                isError: true,
-                            };
-                        }
-                        continue;
-                    }
-                    const text = await pollResp.text().catch(() => '');
-                    return { output: `Image poll failed (${pollResp.status}): ${text.slice(0, 200)}`, isError: true };
+                const outcome = await pollImageJob(pollEndpoint, pollHeaders, controller.signal);
+                if (outcome.kind === 'failed') {
+                    return {
+                        output: `Image generation failed upstream: ${JSON.stringify(outcome.error ?? '').slice(0, 240)}`,
+                        isError: true,
+                    };
                 }
-                if (!polled || !polled.data?.[0]) {
+                if (outcome.kind === 'poll_http_error') {
+                    return { output: `Image poll failed (${outcome.status}): ${outcome.bodyPreview}`, isError: true };
+                }
+                if (outcome.kind === 'timed_out') {
                     return {
                         output: `Image generation queued but did not complete within 5 minutes. Payment was settled when the gateway accepted the job (HTTP 202). ` +
                             `If this keeps happening, the upstream image model is overloaded — try a smaller / faster model or retry later.`,
                         isError: true,
                     };
                 }
-                result = polled;
+                result = outcome.body;
             }
             const imageData = result.data?.[0];
             if (!imageData) {
@@ -533,3 +511,48 @@ export function createImageGenCapability(deps = {}) {
 }
 /** Back-compat static capability for callers that don't want the Content bridge. */
 export const imageGenCapability = createImageGenCapability();
+export async function pollImageJob(pollEndpoint, headers, signal, options = {}) {
+    const maxWaitMs = options.maxWaitMs ?? 5 * 60 * 1000;
+    const intervalMs = options.intervalMs ?? 3_000;
+    const deadline = Date.now() + maxWaitMs;
+    while (Date.now() < deadline) {
+        if (signal.aborted)
+            throw new Error('aborted');
+        await sleep(intervalMs, signal);
+        const resp = await fetch(pollEndpoint, { method: 'GET', headers, signal });
+        if (resp.status === 202)
+            continue; // still queued
+        if (resp.status === 429 || resp.status >= 500)
+            continue; // transient
+        if (resp.ok) {
+            const body = (await resp.json().catch(() => null));
+            if (!body)
+                continue;
+            if (body.status === 'failed')
+                return { kind: 'failed', error: body.error };
+            if (body.status === 'completed' || (body.data && body.data[0])) {
+                return { kind: 'completed', body };
+            }
+            // Non-terminal but ok shape (e.g. status: 'in_progress') — wait.
+            continue;
+        }
+        const text = await resp.text().catch(() => '');
+        return { kind: 'poll_http_error', status: resp.status, bodyPreview: text.slice(0, 200) };
+    }
+    return { kind: 'timed_out' };
+}
+function sleep(ms, signal) {
+    return new Promise((resolve, reject) => {
+        if (signal.aborted)
+            return reject(new Error('aborted'));
+        const t = setTimeout(() => {
+            signal.removeEventListener('abort', onAbort);
+            resolve();
+        }, ms);
+        const onAbort = () => {
+            clearTimeout(t);
+            reject(new Error('aborted'));
+        };
+        signal.addEventListener('abort', onAbort, { once: true });
+    });
+}

package/dist/ui/app.js

@@ -14,6 +14,7 @@ import { resolveModel, PICKER_CATEGORIES, PICKER_MODELS_FLAT, } from './model-pi
 import { estimateCost } from '../pricing.js';
 import { formatTokens, shortModelName } from '../stats/format.js';
 import { mouse, forceDisableMouseTracking } from './mouse.js';
+import { resolveAskUserAnswer } from './ask-user-answer.js';
 // ─── Full-width input box ──────────────────────────────────────────────────
 const DISABLE_AUTO_WRAP = '\x1b[?7l';
 const ENABLE_AUTO_WRAP = '\x1b[?7h';
@@ -827,7 +828,12 @@ function RunCodeApp({ initialModel, workDir, walletAddress, walletBalance, chain
                                             ? _jsxs(Text, { color: r.ctxPct >= 80 ? 'red' : r.ctxPct >= 50 ? 'yellow' : undefined, dimColor: r.ctxPct < 50, children: ["  \u00B7  ctx ", r.ctxPct, "%"] })
                                             : ''] }) }))] }, r.key));
                 } }), permissionRequest && (_jsxs(Box, { flexDirection: "column", marginTop: 1, marginLeft: 2, children: [_jsx(Text, { color: "red", bold: true, children: "\u2501\u2501\u2501\u2501\u2501\u2501\u2501\u2501\u2501\u2501 \u26A0  ACTION REQUIRED  \u26A0 \u2501\u2501\u2501\u2501\u2501\u2501\u2501\u2501\u2501\u2501" }), _jsx(Text, { color: "yellow", children: "\u256D\u2500 Permission required \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500" }), _jsxs(Text, { color: "yellow", children: ["\u2502 ", _jsx(Text, { bold: true, children: permissionRequest.toolName })] }), permissionRequest.description.split('\n').map((line, i) => (_jsxs(Text, { dimColor: true, children: ["\u2502 ", line] }, i))), _jsx(Text, { color: "yellow", children: "\u2570\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500" }), _jsx(Box, { marginLeft: 2, children: _jsxs(Text, { children: [_jsx(Text, { bold: true, color: "green", children: "[y]" }), _jsx(Text, { dimColor: true, children: " yes  " }), _jsx(Text, { bold: true, color: "cyan", children: "[a]" }), _jsx(Text, { dimColor: true, children: " always  " }), _jsx(Text, { bold: true, color: "red", children: "[n]" }), _jsx(Text, { dimColor: true, children: " no" })] }) })] })), askUserRequest && (_jsxs(Box, { flexDirection: "column", marginTop: 1, marginLeft: 2, children: [_jsx(Text, { color: "magenta", bold: true, children: "\u2501\u2501\u2501\u2501\u2501\u2501\u2501\u2501\u2501\u2501 \u26A0  ANSWER REQUIRED  \u26A0 \u2501\u2501\u2501\u2501\u2501\u2501\u2501\u2501\u2501\u2501" }), _jsx(Text, { color: "cyan", children: "\u256D\u2500 Question \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500" }), _jsxs(Text, { color: "cyan", children: ["\u2502 ", _jsx(Text, { bold: true, children: askUserRequest.question })] }), askUserRequest.options && askUserRequest.options.length > 0 && (askUserRequest.options.map((opt, i) => (_jsxs(Text, { dimColor: true, children: ["\u2502 ", i + 1, ". ", opt] }, i)))), _jsx(Text, { color: "cyan", children: "\u2570\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500" }), _jsxs(Box, { marginLeft: 2, children: [_jsx(Text, { bold: true, children: "answer> " }), _jsx(TextInput, { value: askUserInput, onChange: setAskUserInput, onSubmit: (val) => {
-                                    const answer = val.trim() || '(no response)';
+                                    // resolveAskUserAnswer translates "1" / "2" / ... into the
+                                    // matching label string when the dialog showed a numbered
+                                    // option list. Without it, every onAskUser caller's
+                                    // exact-label match fails for digit answers and silently
+                                    // falls through to the default branch (typically cancel).
+                                    const answer = resolveAskUserAnswer(val, askUserRequest.options);
                                     const r = askUserRequest.resolve;
                                     setAskUserRequest(null);
                                     setAskUserInput('');

package/dist/ui/ask-user-answer.d.ts

@@ -0,0 +1,24 @@
+/**
+ * Resolve a user-typed AskUser answer against the option list.
+ *
+ * The TUI renders option labels as a numbered list ("1. X", "2. Y", …),
+ * so users naturally type the digit. Every tool-side onAskUser caller
+ * (videogen.ts:113, modal.ts:371, jupiter.ts:368, zerox-base.ts:453,
+ * zerox-gasless.ts:446) does an exact-string match against the full
+ * label, so a bare "1" silently falls through to the caller's default
+ * branch — which is typically "cancel".
+ *
+ * Verified 2026-05-04 in a live session: user typed "1" twice in a
+ * VideoGen flow, both invocations returned "Video generation cancelled
+ * (No USDC was spent)" even though the wallet had $94.72 and the
+ * Content budget had $2.00 untouched.
+ *
+ * Translation rules:
+ *   - "" → "(no response)" (preserve the existing empty-input fallback)
+ *   - "<digit>" with options.length > 0 and 1 ≤ digit ≤ options.length
+ *     → the matching label string
+ *   - anything else → the trimmed input verbatim (callers that match
+ *     against label can still get a literal answer when the user types
+ *     it out, and free-form text questions still work the same way)
+ */
+export declare function resolveAskUserAnswer(raw: string, options: readonly string[] | undefined): string;

package/dist/ui/ask-user-answer.js

@@ -0,0 +1,34 @@
+/**
+ * Resolve a user-typed AskUser answer against the option list.
+ *
+ * The TUI renders option labels as a numbered list ("1. X", "2. Y", …),
+ * so users naturally type the digit. Every tool-side onAskUser caller
+ * (videogen.ts:113, modal.ts:371, jupiter.ts:368, zerox-base.ts:453,
+ * zerox-gasless.ts:446) does an exact-string match against the full
+ * label, so a bare "1" silently falls through to the caller's default
+ * branch — which is typically "cancel".
+ *
+ * Verified 2026-05-04 in a live session: user typed "1" twice in a
+ * VideoGen flow, both invocations returned "Video generation cancelled
+ * (No USDC was spent)" even though the wallet had $94.72 and the
+ * Content budget had $2.00 untouched.
+ *
+ * Translation rules:
+ *   - "" → "(no response)" (preserve the existing empty-input fallback)
+ *   - "<digit>" with options.length > 0 and 1 ≤ digit ≤ options.length
+ *     → the matching label string
+ *   - anything else → the trimmed input verbatim (callers that match
+ *     against label can still get a literal answer when the user types
+ *     it out, and free-form text questions still work the same way)
+ */
+export function resolveAskUserAnswer(raw, options) {
+    const trimmed = raw.trim();
+    if (!trimmed)
+        return '(no response)';
+    if (options && options.length > 0 && /^\d+$/.test(trimmed)) {
+        const idx = parseInt(trimmed, 10) - 1;
+        if (idx >= 0 && idx < options.length)
+            return options[idx];
+    }
+    return trimmed;
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@blockrun/franklin",
-  "version": "3.15.48",
+  "version": "3.15.50",
   "description": "Franklin — The AI agent with a wallet. Spends USDC autonomously to get real work done. Pay per action, no subscriptions.",
   "type": "module",
   "exports": {