bloby-bot 0.70.12 → 0.71.0

package/supervisor/harnesses/codex.ts

@@ -44,6 +44,15 @@ import type { SavedFile } from '../file-saver.js';
 import { getCodexAccessToken } from '../../worker/codex-auth.js';
 import { assembleSystemPrompt } from '../../worker/prompts/prompt-assembler.js';
 import { mirrorSkillsInto } from './skills.js';
+import {
+  routeAttachment,
+  normalizeImageMediaType,
+  approxBase64Bytes,
+  buildSavedFilesNote,
+  INLINE_TEXT_PER_FILE_CHARS,
+  INLINE_TEXT_TOTAL_CHARS,
+  MAX_INLINE_IMAGE_BYTES,
+} from './attachment-policy.js';
 import type { OnAgentMessage, RecentMessage, AgentAttachment, AgentQueryRequest, AgentQueryResult } from './types.js';
 export type { RecentMessage, AgentAttachment };
 
@@ -555,42 +564,24 @@ function emitDone(conv: CodexConversation): void {
 
 /* ── Input building ────────────────────────────────────────────────────── */
 
-/** mediaTypes whose content we inline into the prompt — codex has no document
- *  input type (verified against 0.138 UserInput), so this is the closest we get
- *  to claude's native base64 document ingestion for text-like files. */
-const INLINE_TEXT_MEDIA = /^(text\/|application\/(json|xml|yaml|x-yaml|toml|csv|javascript|typescript))/;
-const INLINE_TEXT_MAX_BYTES = 48_000;
-const INLINE_TEXT_TOTAL_BUDGET = 96_000;
-
+/**
+ * Build codex `UserInput` blocks from the user text + saved files + raw
+ * attachments. Routing is delegated to the shared attachment-policy so codex
+ * stays byte-for-byte consistent with the Claude/PI harnesses. Codex's UserInput
+ * has NO native document type (verified against 0.138), so canNativeDocument is
+ * FALSE: PDFs/binaries become a disk-pointer note and the agent opens them with
+ * its file tools.
+ *
+ * Block order is MEDIA-first then TEXT (matching Claude/PI): the inline-text
+ * notes and the saved-files pointer are folded into the trailing text block.
+ */
 function buildUserInput(text: string, savedFiles?: SavedFile[], attachments?: AgentAttachment[]): Array<Record<string, any>> {
   const input: Array<Record<string, any>> = [];
 
-  let promptText = text || '(attached files)';
-  if (savedFiles?.length) {
-    const lines = savedFiles.map((f) => `- ${f.name} -> ${f.relPath}`);
-    promptText += `\n\n[Attached files saved to disk]\n${lines.join('\n')}\nYou can read or reference these files using the paths above (relative to your cwd).`;
-  }
-
-  // Inline text-like attachments (size-capped) so the model can answer about
-  // their contents immediately instead of shelling out to read them from disk.
-  if (attachments?.length) {
-    let budget = INLINE_TEXT_TOTAL_BUDGET;
-    for (const att of attachments) {
-      if (att.type !== 'file' || !INLINE_TEXT_MEDIA.test(att.mediaType || '')) continue;
-      const approxBytes = Math.floor((att.data?.length || 0) * 0.75);
-      if (approxBytes === 0 || approxBytes > INLINE_TEXT_MAX_BYTES || approxBytes > budget) continue;
-      try {
-        const decoded = Buffer.from(att.data, 'base64').toString('utf-8');
-        budget -= approxBytes;
-        promptText += `\n\n[Attached file content: ${att.name}]\n\`\`\`\n${decoded}\n\`\`\``;
-      } catch {}
-    }
-  }
-
-  input.push({ type: 'text', text: promptText });
-
   // Codex understands `localImage` (path on disk) — Bloby's file-saver already
-  // wrote attachments to disk, so we just point at the absolute path.
+  // wrote image attachments to disk, so we point at the absolute path. Track a
+  // per-name COUNT (not presence): WhatsApp multi-image pushes share one
+  // attachment name and each saved file covers exactly one of them.
   const savedImageCounts = new Map<string, number>();
   if (savedFiles?.length) {
     for (const f of savedFiles) {
@@ -601,23 +592,58 @@ function buildUserInput(text: string, savedFiles?: SavedFile[], attachments?: Ag
     }
   }
 
-  // Fallback for image attachments that never made it to disk (file-saver
-  // failure): inline as a data URL so the agent still SEES the image — claude
-  // always gets the base64 inline, codex shouldn't silently go blind.
-  // Match by COUNT, not name presence: WhatsApp multi-image pushes share one
-  // attachment name, and each saved file covers exactly one of them.
+  // Route every attachment through the shared policy. Inline-text notes are
+  // accumulated into `inlineNotes` (appended to the trailing text block);
+  // images become data-URL blocks (with the localImage path already covering
+  // disk-saved copies); everything else falls back to the saved-files pointer.
+  let promptText = text || '(attached files)';
+  const inlineNotes: string[] = [];
+  let inlineBudget = INLINE_TEXT_TOTAL_CHARS;
   if (attachments?.length) {
     for (const att of attachments) {
-      if (att.type !== 'image' || !att.data) continue;
-      const remaining = savedImageCounts.get(att.name) || 0;
-      if (remaining > 0) {
-        savedImageCounts.set(att.name, remaining - 1);
-        continue;
+      switch (routeAttachment(att, { canNativeDocument: false })) {
+        case 'image': {
+          if (!att.data) break;
+          // Skip data-URL inlining when a disk copy exists (localImage already
+          // points codex at it) or when the payload is too big to resend on
+          // every stateless turn — the saved-files pointer covers it instead.
+          const remaining = savedImageCounts.get(att.name) || 0;
+          if (remaining > 0) {
+            savedImageCounts.set(att.name, remaining - 1);
+            break;
+          }
+          if (approxBase64Bytes(att.data) > MAX_INLINE_IMAGE_BYTES) break;
+          const mediaType = normalizeImageMediaType(att.mediaType);
+          input.push({ type: 'image', url: `data:${mediaType};base64,${att.data}` });
+          break;
+        }
+        case 'inline-text': {
+          if (!att.data) break;
+          try {
+            let decoded = Buffer.from(att.data, 'base64').toString('utf-8');
+            if (decoded.length > INLINE_TEXT_PER_FILE_CHARS) decoded = decoded.slice(0, INLINE_TEXT_PER_FILE_CHARS);
+            if (decoded.length > inlineBudget) decoded = decoded.slice(0, inlineBudget);
+            if (!decoded.length) break;
+            inlineBudget -= decoded.length;
+            inlineNotes.push(`\n\n[Attached file content: ${att.name}]\n\`\`\`\n${decoded}\n\`\`\``);
+          } catch {}
+          break;
+        }
+        // 'native-document' cannot occur (canNativeDocument:false); it and
+        // 'reference-only' both rely on the saved-files disk pointer below.
+        default:
+          break;
       }
-      input.push({ type: 'image', url: `data:${att.mediaType};base64,${att.data}` });
     }
   }
 
+  for (const note of inlineNotes) promptText += note;
+
+  const savedNote = buildSavedFilesNote(savedFiles || []);
+  if (savedNote) promptText += `\n\n${savedNote}`;
+
+  input.push({ type: 'text', text: promptText });
+
   return input;
 }