feat: wire PDF and text file content into AI chat messages

PDF uploads were stored in S3 and had text extracted during upload, but
fetch_upload_images() filtered exclusively for image MIME types, so
document content never reached the AI.

- Add fetch_upload_documents() in storage_service.py to retrieve
  extracted_content for PDFs and text files
- Update ai_sessions.py chat endpoint to call both fetch_upload_images
  and fetch_upload_documents, injecting document text as context
- Add PDF text extraction in _generate_ai_description (pypdf)
- Add pypdf>=4.0.0 to requirements.txt
- Fix test_db teardown to avoid connection pool issues
- Add 5 tests for fetch_upload_documents

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

backend/app/services/storage_service.py

@@ -16,10 +16,12 @@ logger = logging.getLogger(__name__)
 
 ALLOWED_IMAGE_TYPES = {"image/png", "image/jpeg", "image/gif", "image/webp"}
 ALLOWED_TEXT_TYPES = {"text/plain", "text/csv", "application/octet-stream"}
-ALLOWED_TYPES = ALLOWED_IMAGE_TYPES | ALLOWED_TEXT_TYPES
+ALLOWED_DOCUMENT_TYPES = {"application/pdf"}
+ALLOWED_TYPES = ALLOWED_IMAGE_TYPES | ALLOWED_TEXT_TYPES | ALLOWED_DOCUMENT_TYPES
 
-MAX_IMAGE_SIZE = 5 * 1024 * 1024  # 5MB
-MAX_TEXT_SIZE = 1 * 1024 * 1024   # 1MB
+MAX_IMAGE_SIZE = 5 * 1024 * 1024      # 5MB
+MAX_TEXT_SIZE = 1 * 1024 * 1024        # 1MB
+MAX_DOCUMENT_SIZE = 10 * 1024 * 1024   # 10MB
 MAX_FILES_PER_SESSION = 20
 MAX_BYTES_PER_SESSION = 50 * 1024 * 1024  # 50MB
 
@@ -44,7 +46,12 @@ def validate_upload(content_type: str, size_bytes: int) -> str | None:
     """Validate file type and size. Returns error message or None."""
     if content_type not in ALLOWED_TYPES:
         return f"File type {content_type} not allowed"
-    max_size = MAX_IMAGE_SIZE if content_type in ALLOWED_IMAGE_TYPES else MAX_TEXT_SIZE
+    if content_type in ALLOWED_IMAGE_TYPES:
+        max_size = MAX_IMAGE_SIZE
+    elif content_type in ALLOWED_DOCUMENT_TYPES:
+        max_size = MAX_DOCUMENT_SIZE
+    else:
+        max_size = MAX_TEXT_SIZE
     if size_bytes > max_size:
         return f"File too large ({size_bytes} bytes, max {max_size})"
     return None
@@ -199,3 +206,77 @@ async def fetch_upload_images(
         except Exception:
             logger.warning("Failed to fetch upload %s from S3", upload.id)
     return images
+
+
+DOCUMENT_CONTENT_TYPES = ALLOWED_DOCUMENT_TYPES | ALLOWED_TEXT_TYPES
+MAX_DOCUMENT_CONTEXT_CHARS = 10_000  # Cap total injected text to control token cost
+
+
+async def fetch_upload_documents(
+    upload_ids: list[UUID],
+    account_id: UUID,
+    db: Any,
+) -> list[dict[str, str]]:
+    """Fetch extracted text content for non-image uploads (PDFs, text files).
+
+    Returns a list of dicts with 'filename', 'content_type', and 'text' keys.
+    Text is sourced from the FileUpload.extracted_content field (populated
+    during upload by _generate_ai_description). Falls back to downloading
+    and decoding text files from S3 if extracted_content is empty.
+    """
+    if not upload_ids:
+        return []
+
+    from sqlalchemy import select
+    from app.models.file_upload import FileUpload
+
+    result = await db.execute(
+        select(FileUpload).where(
+            FileUpload.id.in_(upload_ids),
+            FileUpload.account_id == account_id,
+            FileUpload.content_type.in_(DOCUMENT_CONTENT_TYPES),
+        )
+    )
+    uploads = result.scalars().all()
+
+    documents: list[dict[str, str]] = []
+    total_chars = 0
+    for upload in uploads:
+        text = upload.extracted_content or ""
+
+        # Fallback: for text files without extracted_content, fetch from S3
+        if not text and upload.content_type in ALLOWED_TEXT_TYPES and settings.STORAGE_ENDPOINT:
+            try:
+                file_data = download_file(upload.storage_key)
+                try:
+                    text = file_data.decode("utf-8")
+                except UnicodeDecodeError:
+                    text = file_data.decode("latin-1")
+                text = text[:MAX_DOCUMENT_CONTEXT_CHARS]
+            except Exception:
+                logger.warning("Failed to fetch text upload %s from S3", upload.id)
+                continue
+
+        if not text:
+            # PDF with no extractable text — include a note so AI knows
+            documents.append({
+                "filename": upload.filename,
+                "content_type": upload.content_type,
+                "text": f"[Attached file: {upload.filename} — no extractable text content]",
+            })
+            continue
+
+        # Cap per-document and total to control token budget
+        remaining = MAX_DOCUMENT_CONTEXT_CHARS - total_chars
+        if remaining <= 0:
+            break
+        truncated = text[:remaining]
+        total_chars += len(truncated)
+
+        documents.append({
+            "filename": upload.filename,
+            "content_type": upload.content_type,
+            "text": truncated,
+        })
+
+    return documents