Fast API update

app.py

@@ -10,7 +10,7 @@ from typing import List, Optional
 # Alternatively you can do this in a "startup" event handler
 os.system("python download_models_hf.py")
 
-from parallel_multiproc import process_batch_in_parallel
+from mineru_single import to_markdown
 # Or if you want single-file approach, from miner_single import to_markdown
 
 app = FastAPI()
@@ -22,67 +22,19 @@ os.makedirs(INBOX_DIR, exist_ok=True)
 os.makedirs(OUTPUT_DIR, exist_ok=True)
 
 @app.post("/process")
-async def process_pdfs_in_parallel(
-    files: List[UploadFile] = File(...),
-    background_tasks: BackgroundTasks = None,
-    num_workers: int = 2,
-    num_gpus: int = 1
-):
-    """
-    POST multiple PDFs via multipart/form-data.
-    We store them in ./inbox, then process in the background using parallel_processor.
-    """
-    pdf_paths = []
-    for f in files:
-        file_path = os.path.join(INBOX_DIR, f.filename)
-        with open(file_path, "wb") as out_file:
-            shutil.copyfileobj(f.file, out_file)
-        pdf_paths.append(file_path)
+async def process_pdf(file: UploadFile = File(...)):
 
-    # Launch parallel processing in background (so we can return immediately)
-    background_tasks.add_task(
-        process_batch_in_parallel,
-        pdf_paths,
-        OUTPUT_DIR,
-        num_workers,
-        num_gpus
-    )
+    file_path = os.path.join(INBOX_DIR, file.filename)
+    with open(file_path, "wb") as out_file:
+        shutil.copyfileobj(file.file, out_file)
 
-    return {
-        "message": "Processing started",
-        "files_received": [f.filename for f in files],
-        "workers": num_workers,
-        "gpus": num_gpus
-    }
-
-@app.get("/check_output")
-def check_output(filename: str):
-    """
-    Simple endpoint: provide a PDF filename, returns the final .md (if ready).
-    We assume the PDF was named e.g. 'paper.pdf'. The final output is `paper.md`
-    in `./output/paper/paper_<timestamp>.md`.
-    Because of how we rename outputs in `miner_single.py`, you may need to locate them by pattern.
-    """
-    name_wo_ext = os.path.splitext(filename)[0]
-    # Because we appended timestamp, let's see if we can locate a .md in the folder
-    subdir = os.path.join(OUTPUT_DIR, name_wo_ext)
-    if not os.path.exists(subdir):
-        return {"status": "not_found"}
-
-    # Try to find a .md in the subdir
-    found_md = [f for f in os.listdir(subdir) if f.endswith(".md")]
-    if not found_md:
-        return {"status": "incomplete"}
-
-    # If we do find it:
-    md_path = os.path.join(subdir, found_md[0])
-    with open(md_path, "r", encoding="utf-8") as f:
-        content = f.read()
+    # Process the file and wait for completion
+    markdown_text = to_markdown(file_path)
 
     return {
-        "status": "complete",
-        "markdown_file": found_md[0],
-        "content": content
+        "message": "Processing completed",
+        "code": 200,
+        "content": markdown_text
     }
 
 # If you want to run locally or for debug:

mineru_single.py

@@ -135,8 +135,4 @@ def to_markdown(
 
     md_content_with_embeds = replace_image_with_base64(original_md_content, local_md_dir)
 
-    # Overwrite the original Markdown with the embedded one
-    with open(md_path, "w", encoding="utf-8") as fw:
-        fw.write(md_content_with_embeds)
-
-    return md_path
+    return md_content_with_embeds