chihlasm/resolutionflow
1
0
Fork 0
Code Issues 23 Pull Requests 1 Actions Packages Projects Releases Wiki Activity

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

Browse Source 
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>
This commit is contained in:
chihlasm
2026-03-27 21:02:56 +00:00
parent 3cea949519
commit 11de850054
6 changed files with 324 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
backend/app/api/endpoints/ai_sessions.py
View File

@@ -280,18 +280,28 @@ async def send_chat_message(
user_id = current_user.id
account_id = current_user.account_id
# Fetch attached images from S3 (if any)
# Fetch attached uploads from S3 (if any)
images = None
message = data.message
if data.upload_ids:
from app.services.storage_service import fetch_upload_images
from app.services.storage_service import fetch_upload_images, fetch_upload_documents
images = await fetch_upload_images(data.upload_ids, account_id, db) or None
# Inject document text (PDFs, text files) as context in the message
documents = await fetch_upload_documents(data.upload_ids, account_id, db)
if documents:
doc_parts = []
for doc in documents:
doc_parts.append(f"--- Attached file: {doc['filename']} ---\n{doc['text']}")
doc_context = "\n\n".join(doc_parts)
message = f"{message}\n\n[Attached document content]\n{doc_context}"
try:
ai_content, suggested_flows, session, fork_metadata, actions_data, questions_data = await unified_chat_service.send_chat_message(
session_id=session_id,
user_id=user_id,
account_id=account_id,
message=data.message,
message=message,
db=db,
images=images,
)

34
backend/app/api/endpoints/uploads.py
View File

@@ -61,6 +61,40 @@ async def _generate_ai_description(upload_id: UUID, file_data: bytes, content_ty
max_tokens=100,
)
upload.ai_description = description
elif content_type == "application/pdf":
try:
from pypdf import PdfReader
import io as _io
reader = PdfReader(_io.BytesIO(file_data))
pages_text = []
for page in reader.pages:
page_text = page.extract_text()
if page_text:
pages_text.append(page_text)
text_content = "\n\n".join(pages_text)
except Exception:
logger.warning("PDF text extraction failed for upload %s", upload_id)
text_content = ""
if text_content:
upload.extracted_content = text_content[:10000]
if len(text_content) > 2000:
summary, _, _ = await _call_ai(
system_base="You are a technical document analyst for IT troubleshooting.",
rag_context="",
history=[],
new_message=f"Summarize this PDF content in 2-3 sentences:\n\n{text_content[:5000]}",
max_tokens=200,
)
upload.content_summary = summary
upload.ai_description = summary
else:
upload.ai_description = f"PDF document: {upload.filename}"
else:
upload.ai_description = f"PDF document (no extractable text): {upload.filename}"
elif content_type.startswith("text/") or content_type in (
"application/json", "application/xml", "application/yaml",
):

89
backend/app/services/storage_service.py
View File

@@ -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