OCR

OCR evaluation reports

For testing purposes, let’s load evaluations:

fname_json = '../_data/output/evaluations.json'
evals = load_evals(fname_json)
print(evals[0])

{
    'id': '1a57974ab89d7280988aa6b706147ce1',
    'docs': [
        {
            'Document Subtype': 'Evaluation report',
            'File URL': 
'https://evaluation.iom.int/sites/g/files/tmzbdl151/files/docs/resources/Internal%20Evaluation_NG20P0516_MAY_2023_F
INAL_Abderrahim%20EL%20MOULAT.pdf',
            'File description': 'Evaluation Report'
        },
        {
            'Document Subtype': 'Evaluation brief',
            'File URL': 
'https://evaluation.iom.int/sites/g/files/tmzbdl151/files/docs/resources/RR0163_Evaluation%20Brief_MAY_%202023_Abde
rrahim%20EL%20MOULAT.pdf',
            'File description': 'Evaluation Brief'
        }
    ],
    'meta': {
        'Title': 'EX-POST EVALUATION OF THE PROJECT:  NIGERIA: STRENGTHENING REINTEGRATION FOR RETURNEES (SRARP)  -
PHASE II',
        'Year': 2023,
        'Author': 'Abderrahim El Moulat',
        'Best Practicesor Lessons Learnt': 'Yes',
        'Date of Publication': '2023-05-10',
        'Donor': 'Government of Germany',
        'Evaluation Brief': 'Yes',
        'Evaluation Commissioner': 'Donor, IOM',
        'Evaluation Coverage': 'Country',
        'Evaluation Period From Date': 'nan',
        'Evaluation Period To Date': 'NaT',
        'Executive Summary': 'Yes',
        'External Version of the Report': 'No',
        'Languages': 'English',
        'Migration Thematic Areas': 'Assistance to vulnerable migrants, Migrant training and integration (including
community cohesion), Migration health (assessment, travel, health promotion, crisis-affected), Return and AVRR',
        'Name of Project(s) Being Evaluated': nan,
        'Number of Pages Excluding annexes': nan,
        'Other Documents Included': nan,
        'Project Code': 'RR.0163',
        'Countries Covered': ['Nigeria'],
        'Regions Covered': 'RO Dakar',
        'Relevant Crosscutting Themes': 'Gender, Rights-based approach',
        'Report Published': 'Yes',
        'Terms of Reference': 'No',
        'Type of Evaluation Scope': 'Programme/Project',
        'Type of Evaluation Timing': 'Ex-post (after the end of the project/programme)',
        'Type of Evaluator': 'Internal',
        'Level of Evaluation': 'Decentralized',
        'Document Subtype': 'Evaluation report, Evaluation brief',
        'File URL': 
'https://evaluation.iom.int/sites/g/files/tmzbdl151/files/docs/resources/Internal%20Evaluation_NG20P0516_MAY_2023_F
INAL_Abderrahim%20EL%20MOULAT.pdf,   
https://evaluation.iom.int/sites/g/files/tmzbdl151/files/docs/resources/RR0163_Evaluation%20Brief_MAY_%202023_Abder
rahim%20EL%20MOULAT.pdf',
        'File description': 'Evaluation Report , Evaluation Brief',
        'Management response': 'No',
        'Date added': 'Fri, 07/07/2023 - 15:35',
        'Metaevaluation': '2020-24',
        'exclude': nan,
        'reason': nan
    }
}

To find a particular evaluation per name (e.g “Final Evaluation of the EU-IOM Joint …”), we can simply:

title = 'Final Evaluation of the EU-IOM Joint Initiative for migrant protection and reintegration in the Horn of Africa'
results = [o for o in evals.filter(lambda x: title.lower() in x['meta']['Title'].lower())]; results

[{'id': '49d2fba781b6a7c0d94577479636ee6f',
  'docs': [{'Document Subtype': 'Evaluation report',
    'File URL': 'https://evaluation.iom.int/sites/g/files/tmzbdl151/files/docs/resources/Abridged%20Evaluation%20Report_%20Final_Olta%20NDOJA.pdf',
    'File description': 'Evaluation Report'},
   {'Document Subtype': 'Evaluation brief',
    'File URL': 'https://evaluation.iom.int/sites/g/files/tmzbdl151/files/docs/resources/Evaluation%20Learning%20Brief_Final_Olta%20NDOJA.pdf',
    'File description': 'Evaluation Brief'},
   {'Document Subtype': 'Annexes',
    'File URL': 'https://evaluation.iom.int/sites/g/files/tmzbdl151/files/docs/resources/Final%20Evaluation%20Report%20Final_Olta%20NDOJA.pdf',
    'File description': 'Abridged Report'},
   {'Document Subtype': 'Management response',
    'File URL': 'https://evaluation.iom.int/sites/g/files/tmzbdl151/files/docs/resources/HoA%20EU%20JI%20Final%20Eval%20-%20Management%20Response%20Matrix%20-%20Final.pdf',
    'File description': 'Management Response'}],
  'meta': {'Title': 'Final Evaluation of the EU-IOM Joint Initiative for migrant protection and reintegration in the horn of Africa',
   'Year': 2023,
   'Author': 'PPMI Group',
   'Best Practicesor Lessons Learnt': 'Yes',
   'Date of Publication': '2023-03-17',
   'Donor': 'European Union',
   'Evaluation Brief': 'Yes',
   'Evaluation Commissioner': 'IOM',
   'Evaluation Coverage': 'Regional',
   'Evaluation Period From Date': 'nan',
   'Evaluation Period To Date': 'NaT',
   'Executive Summary': 'Yes',
   'External Version of the Report': 'No',
   'Languages': 'English',
   'Migration Thematic Areas': 'Assistance to vulnerable migrants, Return and AVRR',
   'Name of Project(s) Being Evaluated': nan,
   'Number of Pages Excluding annexes': nan,
   'Other Documents Included': nan,
   'Project Code': 'RT.1354',
   'Countries Covered': ['Djibouti', 'Ethiopia', 'Somalia', 'South Sudan'],
   'Regions Covered': 'RO Nairobi',
   'Relevant Crosscutting Themes': 'Accountability to affected populations, Environment, Gender, Mainstreaming protection into crisis response, Principled humanitarian action, Rights-based approach',
   'Report Published': 'Yes',
   'Terms of Reference': 'No',
   'Type of Evaluation Scope': 'Programme/Project',
   'Type of Evaluation Timing': 'Final (at the end of the project/programme)',
   'Type of Evaluator': 'External',
   'Level of Evaluation': 'Decentralized',
   'Document Subtype': 'Evaluation report, Evaluation brief, Annexes, Management response',
   'File URL': 'https://evaluation.iom.int/sites/g/files/tmzbdl151/files/docs/resources/Abridged%20Evaluation%20Report_%20Final_Olta%20NDOJA.pdf,   https://evaluation.iom.int/sites/g/files/tmzbdl151/files/docs/resources/Evaluation%20Learning%20Brief_Final_Olta%20NDOJA.pdf,   https://evaluation.iom.int/sites/g/files/tmzbdl151/files/docs/resources/Final%20Evaluation%20Report%20Final_Olta%20NDOJA.pdf,   https://evaluation.iom.int/sites/g/files/tmzbdl151/files/docs/resources/HoA%20EU%20JI%20Final%20Eval%20-%20Management%20Response%20Matrix%20-%20Final.pdf',
   'File description': 'Evaluation Report , Evaluation Brief, Abridged Report, Management Response',
   'Management response': 'No',
   'Date added': 'Fri, 05/26/2023 - 12:38',
   'Metaevaluation': '2020-24',
   'exclude': nan,
   'reason': nan}}]

Utils

Given an evaluation id and a pdf file name of one of its supporting doc we’d like to check its subtype:

---

source

get_doc_subtype

 get_doc_subtype (id:str, fname:str, evals)

Get Document Subtype for a given file in the evaluation dataset

	Type	Details
id	str	ID of the evaluation
fname	str	Name of the file
evals		Evaluations data
Returns	str	Document Subtype

def get_doc_subtype(
    id:str, # ID of the evaluation
    fname:str, # Name of the file
    evals # Evaluations data
    )->str: # Document Subtype
    "Get Document Subtype for a given file in the evaluation dataset"
    eval_data = L(evals).filter(lambda x: x['id']==id)
    if not eval_data: return None
    
    docs = L(eval_data[0]['docs'])
    matches = docs.filter(lambda x: Path(x['File URL']).name==fname)
    return matches[0]['Document Subtype'] if matches else None

Now, based on downloaded PDFs on path_pdf, we can get the subtype of each document:

path_pdf = Path('../_data/pdf_library')
id = '49d2fba781b6a7c0d94577479636ee6f'
for o in path_pdf.ls().filter(lambda x: x.name == id)[0].ls():
    print(f'Name: {o.name}\nSubtype: {get_doc_subtype(id, o.name, evals)}')

Name: Evaluation%20Learning%20Brief_Final_Olta%20NDOJA.pdf
Subtype: Evaluation brief

Name: HoA%20EU%20JI%20Final%20Eval%20-%20Management%20Response%20Matrix%20-%20Final.pdf
Subtype: Management response

Name: Abridged%20Evaluation%20Report_%20Final_Olta%20NDOJA.pdf
Subtype: Evaluation report

Name: Final%20Evaluation%20Report%20Final_Olta%20NDOJA.pdf
Subtype: Annexes

---

source

clean_pdf_name

 clean_pdf_name (pdf_name:str)

Clean PDF name to create folder-friendly string. Removes special characters and spaces, replaces with underscores.

# Note: filename will be cleaned upstream in a next version
def clean_pdf_name(pdf_name: str) -> str:
    """
    Clean PDF name to create folder-friendly string.
    Removes special characters and spaces, replaces with underscores.
    """
    # Remove URL encoding
    pdf_name = urllib.parse.unquote(pdf_name)
    
    # Replace spaces and special characters with underscores
    # Replace any character that is not a word character (\w), whitespace (\s), or hyphen (-) with underscore
    cleaned = re.sub(r'[^\w\s-]', '_', pdf_name)
    
    # Replace any sequence of hyphens or whitespace with a single underscore
    cleaned = re.sub(r'[-\s]+', '_', cleaned)
    
    # Replace multiple consecutive underscores with a single underscore
    cleaned = re.sub(r'_+', '_', cleaned)
    cleaned = cleaned.strip('_')  # Remove leading/trailing underscores
    
    return cleaned.lower()

clean_pdf_name("Final%20Evaluation%20Report%20Final_Olta%20NDOJA.pdf")

'final_evaluation_report_final_olta_ndoja_pdf'

Batch processing

---

source

setup_output_dirs

 setup_output_dirs (md_library_path='../_data/md_library')

Set up the output directory structure for markdown files

logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s - %(levelname)s - %(message)s',
    handlers=[
        logging.FileHandler('batch_ocr.log'),
        logging.StreamHandler()  # Also print to console
    ]
)

def setup_output_dirs(md_library_path="../_data/md_library"):
    "Set up the output directory structure for markdown files"
    md_output_dir = Path(md_library_path)
    mkdir(md_output_dir, parents=True, exist_ok=True, overwrite=False)
    return md_output_dir

---

source

get_pdfs_and_dir

 get_pdfs_and_dir (report_path:pathlib.Path, md_output_dir:pathlib.Path)

Get PDFs from report directory and create output directory

	Type	Details
report_path	Path	Path to the report directory
md_output_dir	Path	Path to the output directory
Returns	tuple

def get_pdfs_and_dir(
    report_path:Path, # Path to the report directory
    md_output_dir:Path # Path to the output directory
    ) -> tuple[list[Path], str]:
    "Get PDFs from report directory and create output directory"
    pdfs = report_path.ls(file_exts='.pdf')
    eval_report_path = report_path.name
    mkdir(md_output_dir / eval_report_path, parents=True, exist_ok=True, overwrite=False)
    return pdfs, eval_report_path

Example usage for the “’Final Evaluation of the EU-IOM Joint Initiative …” evaluation:

report_id_test = '49d2fba781b6a7c0d94577479636ee6f'
reports = [p for p in path_pdf.ls() if p.name == report_id_test]
pdfs, eval_report_path = get_pdfs_and_dir(reports[0], md_output_dir)
print(f'Reports: {reports}')
print(f'pdfs[0]: {pdfs[0]}\nEvaluation report path: {eval_report_path}')

Reports: [Path('../_data/pdf_library/49d2fba781b6a7c0d94577479636ee6f')]

pdfs[0]: ../_data/pdf_library/49d2fba781b6a7c0d94577479636ee6f/Evaluation%20Learning%20Brief_Final_Olta%20NDOJA.pdf
Evaluation report path: 49d2fba781b6a7c0d94577479636ee6f

---

source

save_page_images

 save_page_images (page, dest_folder:pathlib.Path)

Save all images from a page to destination folder as PNG

def save_page_images(page, dest_folder: Path):
    "Save all images from a page to destination folder as PNG"
    images = page.images if hasattr(page, 'images') else page.get('images', [])
    
    for img in images:
        img_data = getattr(img, 'image_base64', img.get('image_base64'))
        img_id = getattr(img, 'id', img.get('id'))
        
        if img_data and img_id:
            img_bytes = base64.b64decode(img_data.split(',')[1])
            pil_img = Image.open(BytesIO(img_bytes))
            output_path = dest_folder / img_id
            pil_img.save(output_path)

# def create_batch_ocr_job(
#     pdf_paths: List[Path],
#     model: str = "mistral-ocr-latest",
#     include_images: bool = True,
#     api_key: str = mistral_api_key
# ):
#     "Create a batch job for multiple PDFs"
#     cli = Mistral(api_key=api_key)
    
#     # Upload all PDFs and create batch entries
#     batch_entries = []
#     for i, pdf_path in enumerate(pdf_paths):
#         # Upload PDF
#         uploaded_pdf = cli.files.upload(
#             file={
#                 "file_name": pdf_path.stem,
#                 "content": pdf_path.read_bytes(),
#             },
#             purpose="ocr"
#         )
        
#         signed_url = cli.files.get_signed_url(file_id=uploaded_pdf.id)
        
#         # Create batch entry with custom_id to track which PDF this is
#         entry = {
#             "custom_id": f"{pdf_path.parent.name}_{pdf_path.stem}",  # eval_id_pdfname
#             "body": {
#                 "document": {
#                     "type": "document_url",
#                     "document_url": signed_url.url,
#                 },
#                 "include_image_base64": include_images
#             }
#         }
#         batch_entries.append(entry)
    
#     return batch_entries, cli

---

source

process_batch_results

 process_batch_results (results, md_output_dir)

Process batch OCR results and save to appropriate folders

def process_batch_results(results, md_output_dir):
    "Process batch OCR results and save to appropriate folders"
    for result in results:
        try:
            # Parse custom_id to get eval_id and pdf_name
            eval_id, pdf_name = result['custom_id'].split('_', 1)
            
            # Get OCR response
            ocr_response = result['response']['body']
            
            # Create folder structure
            pdf_clean_name = clean_pdf_name(pdf_name)
            pdf_dir = md_output_dir / eval_id / pdf_clean_name
            pdf_dir.mkdir(parents=True, exist_ok=True)
            
            # Save each page markdown
            for page in ocr_response['pages']:
                page_num = page['index'] + 1
                page_path = pdf_dir / f"page_{page_num}.md"
                page_path.write_text(page['markdown'])
            
            # Save images if they exist
            img_dir = pdf_dir / 'img'
            for page in ocr_response['pages']:
                if page.get('images'):
                    img_dir.mkdir(parents=True, exist_ok=True)
                    save_page_images(page, img_dir)
            
            logging.info(f"Saved {len(ocr_response['pages'])} pages for {pdf_clean_name}")
            
        except Exception as e:
            logging.error(f"Error processing result {result.get('custom_id', 'unknown')}: {e}")

---

source

create_batch_ocr_job

 create_batch_ocr_job (pdf_paths:List[pathlib.Path], eval_report_path:str,
                       model:str='mistral-ocr-latest',
                       include_images:bool=True, api_key:str=None)

Create batch entries for PDFs from one evaluation report

def create_batch_ocr_job(
    pdf_paths: List[Path],
    eval_report_path: str,
    model: str = "mistral-ocr-latest",
    include_images: bool = True,
    api_key: str = mistral_api_key
):
    "Create batch entries for PDFs from one evaluation report"
    cli = Mistral(api_key=api_key)
    
    batch_entries = []
    for pdf_path in pdf_paths:
        uploaded_pdf = cli.files.upload(
            file={
                "file_name": pdf_path.stem,
                "content": pdf_path.read_bytes(),
            },
            purpose="ocr"
        )
        
        signed_url = cli.files.get_signed_url(file_id=uploaded_pdf.id)
        entry = {
            "custom_id": f"{eval_report_path}_{pdf_path.name}",
            "body": {
                "document": {
                    "type": "document_url",
                    "document_url": signed_url.url,
                },
                "include_image_base64": include_images
            }
        }
        batch_entries.append(entry)
        logging.info(f"Added {pdf_path.name} to batch for {eval_report_path}")
        
    return batch_entries, cli

---

source

submit_and_monitor_batch_job

 submit_and_monitor_batch_job (batch_entries, eval_report_path, cli)

Submit batch job and monitor until completion

def submit_and_monitor_batch_job(batch_entries, eval_report_path, cli):
    "Submit batch job and monitor until completion"
    with tempfile.NamedTemporaryFile(mode='w', suffix='.jsonl', delete=True) as temp_file:
        # Write batch entries to temp file
        for entry in batch_entries:
            temp_file.write(json.dumps(entry) + '\n')
        temp_file.flush()
        
        # Upload and create job
        batch_data = cli.files.upload(
            file={"file_name": f"batch_{eval_report_path}.jsonl", 
                  "content": open(temp_file.name, "rb")},
            purpose="batch"
        )
        
        created_job = cli.batch.jobs.create(
            input_files=[batch_data.id],
            model="mistral-ocr-latest",
            endpoint="/v1/ocr"
        )
        
        logging.info(f"Batch job created for {eval_report_path}: {created_job.id}")
        
        # Monitor completion
        while True:
            job = cli.batch.jobs.get(job_id=created_job.id)
            logging.info(f"Job status: {job.status} - {job.succeeded_requests}/{job.total_requests} completed")
            
            if job.status not in ["QUEUED", "RUNNING"]:
                break
            time.sleep(10)
        
        return job

---

source

download_and_parse_results

 download_and_parse_results (job, cli)

Download and parse batch job results

def download_and_parse_results(job, cli):
    "Download and parse batch job results"
    response = cli.files.download(file_id=job.output_file)
    content = response.read().decode('utf-8')
    
    results = []
    for line in content.strip().split('\n'):
        if line:
            results.append(json.loads(line))
    
    logging.info(f"Downloaded and parsed {len(results)} OCR results")
    return results

---

source

process_single_evaluation_batch

 process_single_evaluation_batch (report:pathlib.Path,
                                  md_output_dir:pathlib.Path)

Process one evaluation report using batch OCR

def process_single_evaluation_batch(report: Path, md_output_dir: Path):
    "Process one evaluation report using batch OCR"
    logging.info(f"Processing evaluation: {report.name}")
    pdfs, eval_report_path = get_pdfs_and_dir(report, md_output_dir)
    
    if not pdfs:
        logging.warning(f"No PDFs found in {eval_report_path}")
        return
    
    batch_entries, cli = create_batch_ocr_job(pdfs, eval_report_path)
    
    job = submit_and_monitor_batch_job(batch_entries, eval_report_path, cli)
    
    if job and job.status == "SUCCESS":
        results = download_and_parse_results(job, cli)
        process_batch_results(results, md_output_dir)
        logging.info(f"Completed processing evaluation: {eval_report_path}")
    else:
        logging.error(f"Job failed for {eval_report_path}")

---

source

process_all_reports_batch

 process_all_reports_batch (reports:list[pathlib.Path],
                            md_library_path='../_data/md_library')

Process evaluation reports using batch OCR

def process_all_reports_batch(
    reports: list[Path],
    md_library_path="../_data/md_library"
):
    "Process evaluation reports using batch OCR"
    logging.info(f"Starting batch OCR processing for {len(reports)} reports")
    md_output_dir = setup_output_dirs(md_library_path)
    
    for report in tqdm(reports, desc="Processing reports"):
        process_single_evaluation_batch(report, md_output_dir)
    
    logging.info("Batch OCR processing completed for all reports")

process_all_reports_batch(reports, md_library_path="../_data/md_library")

2025-08-06 17:38:39,461 - INFO - Starting batch OCR processing for 1 reports
Processing reports:   0%|          | 0/1 [00:00<?, ?it/s]2025-08-06 17:38:39,464 - INFO - Processing evaluation: 49d2fba781b6a7c0d94577479636ee6f
2025-08-06 17:38:42,165 - INFO - HTTP Request: POST https://api.mistral.ai/v1/files "HTTP/1.1 200 OK"
2025-08-06 17:38:42,521 - INFO - HTTP Request: GET https://api.mistral.ai/v1/files/20ea2096-b022-492c-84de-4130e24c6188/url?expiry=24 "HTTP/1.1 200 OK"
2025-08-06 17:38:42,522 - INFO - Added Evaluation%20Learning%20Brief_Final_Olta%20NDOJA.pdf to batch for 49d2fba781b6a7c0d94577479636ee6f
2025-08-06 17:38:44,401 - INFO - HTTP Request: POST https://api.mistral.ai/v1/files "HTTP/1.1 200 OK"
2025-08-06 17:38:44,756 - INFO - HTTP Request: GET https://api.mistral.ai/v1/files/67423130-6d16-4f97-960e-35f1e85aa91d/url?expiry=24 "HTTP/1.1 200 OK"
2025-08-06 17:38:44,760 - INFO - Added HoA%20EU%20JI%20Final%20Eval%20-%20Management%20Response%20Matrix%20-%20Final.pdf to batch for 49d2fba781b6a7c0d94577479636ee6f
2025-08-06 17:38:52,790 - INFO - HTTP Request: POST https://api.mistral.ai/v1/files "HTTP/1.1 200 OK"
2025-08-06 17:38:52,986 - INFO - HTTP Request: GET https://api.mistral.ai/v1/files/b3d152f4-a906-4bb3-89e9-b9bb67ee995e/url?expiry=24 "HTTP/1.1 200 OK"
2025-08-06 17:38:52,988 - INFO - Added Abridged%20Evaluation%20Report_%20Final_Olta%20NDOJA.pdf to batch for 49d2fba781b6a7c0d94577479636ee6f
2025-08-06 17:39:18,422 - INFO - HTTP Request: POST https://api.mistral.ai/v1/files "HTTP/1.1 200 OK"
2025-08-06 17:39:18,611 - INFO - HTTP Request: GET https://api.mistral.ai/v1/files/6a41e07f-0e95-4408-81c7-b608d4c6e0c6/url?expiry=24 "HTTP/1.1 200 OK"
2025-08-06 17:39:18,615 - INFO - Added Final%20Evaluation%20Report%20Final_Olta%20NDOJA.pdf to batch for 49d2fba781b6a7c0d94577479636ee6f
2025-08-06 17:39:19,210 - INFO - HTTP Request: POST https://api.mistral.ai/v1/files "HTTP/1.1 200 OK"
2025-08-06 17:39:19,899 - INFO - HTTP Request: POST https://api.mistral.ai/v1/batch/jobs "HTTP/1.1 200 OK"
2025-08-06 17:39:19,905 - INFO - Batch job created for 49d2fba781b6a7c0d94577479636ee6f: 427954fa-adde-448d-ba8e-c7fcd4bd9c40
2025-08-06 17:39:20,110 - INFO - HTTP Request: GET https://api.mistral.ai/v1/batch/jobs/427954fa-adde-448d-ba8e-c7fcd4bd9c40 "HTTP/1.1 200 OK"
2025-08-06 17:39:20,115 - INFO - Job status: QUEUED - 0/4 completed
2025-08-06 17:39:30,467 - INFO - HTTP Request: GET https://api.mistral.ai/v1/batch/jobs/427954fa-adde-448d-ba8e-c7fcd4bd9c40 "HTTP/1.1 200 OK"
2025-08-06 17:39:30,471 - INFO - Job status: RUNNING - 0/4 completed
2025-08-06 17:39:41,172 - INFO - HTTP Request: GET https://api.mistral.ai/v1/batch/jobs/427954fa-adde-448d-ba8e-c7fcd4bd9c40 "HTTP/1.1 200 OK"
2025-08-06 17:39:41,177 - INFO - Job status: RUNNING - 0/4 completed
2025-08-06 17:39:51,469 - INFO - HTTP Request: GET https://api.mistral.ai/v1/batch/jobs/427954fa-adde-448d-ba8e-c7fcd4bd9c40 "HTTP/1.1 200 OK"
2025-08-06 17:39:51,474 - INFO - Job status: RUNNING - 0/4 completed
2025-08-06 17:40:01,881 - INFO - HTTP Request: GET https://api.mistral.ai/v1/batch/jobs/427954fa-adde-448d-ba8e-c7fcd4bd9c40 "HTTP/1.1 200 OK"
2025-08-06 17:40:01,883 - INFO - Job status: SUCCESS - 4/4 completed
2025-08-06 17:40:02,110 - INFO - HTTP Request: GET https://api.mistral.ai/v1/files/0a684deb-2579-43c4-8fd0-477683465972/content "HTTP/1.1 200 OK"
2025-08-06 17:40:03,154 - INFO - Downloaded and parsed 4 OCR results
2025-08-06 17:40:03,156 - INFO - Saved 2 pages for evaluation_learning_brief_final_olta_ndoja_pdf
2025-08-06 17:40:03,159 - INFO - Saved 8 pages for hoa_eu_ji_final_eval_management_response_matrix_final_pdf
2025-08-06 17:40:03,200 - INFO - Saved 31 pages for abridged_evaluation_report_final_olta_ndoja_pdf
2025-08-06 17:40:03,267 - INFO - Saved 142 pages for final_evaluation_report_final_olta_ndoja_pdf
2025-08-06 17:40:03,268 - INFO - Completed processing evaluation: 49d2fba781b6a7c0d94577479636ee6f
Processing reports: 100%|██████████| 1/1 [01:23<00:00, 83.80s/it]
2025-08-06 17:40:03,269 - INFO - Batch OCR processing completed for all reports