fix parse model

Use sonnet when testing premium mode in extract e2e
Invalidate cache when running extract tests (#1097 )
2026-07-21 03:55:22 -04:00 · 2026-01-27 16:05:23 -08:00 · 2026-01-27 15:48:14 -08:00 · 2026-01-26 17:33:23 -08:00 · 2026-01-26 17:12:50 -08:00 · 2026-01-26 12:05:16 -08:00
38 changed files with 1597 additions and 3542 deletions
@@ -0,0 +1,162 @@
+name: Hourly Extract E2E Tests
+
+on:
+  schedule:
+    - cron: "18 * * * *"
+  workflow_dispatch:
+    # Allows manual triggering
+    inputs:
+      environment:
+        description: "Environment to run the tests in"
+        required: false
+        default: staging
+        type: choice
+        options:
+          - staging
+          - production
+      notify_slack:
+        description: "Notify Slack"
+        required: false
+        default: false
+        type: boolean
+  workflow_call:
+
+env:
+  UV_VERSION: "0.7.20"
+  PYTHON_VERSION: "3.12"
+  SLACK_CHANNEL_ID: C078PHNTF44 # Extract channel ID
+  API_E2E_LOG_PATH: ${{ github.workspace }}/extract-e2e.log
+
+jobs:
+  extract-e2e:
+    name: "Hourly Extract E2E Tests (${{ matrix.environment }})"
+    runs-on: ubuntu-latest
+    timeout-minutes: 30
+    concurrency:
+      group: ${{ github.workflow }}-${{ github.ref }}-${{ matrix.environment }}
+      cancel-in-progress: true
+    strategy:
+      fail-fast: false
+      matrix:
+        environment: ${{ github.event_name == 'schedule' && fromJson('["staging", "production"]') || fromJson(format('["{0}"]', github.event.inputs.environment || 'staging')) }}
+    steps:
+      - name: Set runtime inputs
+        id: runtime
+        run: |
+          environment=${{ matrix.environment }}
+          notify_slack=${{ github.event.inputs.notify_slack || github.event_name == 'schedule' }}
+          echo "environment=${environment}" >> $GITHUB_OUTPUT
+          echo "notify_slack=${notify_slack}" >> $GITHUB_OUTPUT
+
+          if [ "${environment}" = "production" ]; then
+            echo "LLAMA_CLOUD_BASE_URL=https://api.cloud.llamaindex.ai" >> $GITHUB_ENV
+            api_key_secret="${{ secrets.LLAMA_CLOUD_API_KEY }}"
+            project_id_secret="${{ secrets.LLAMA_CLOUD_PROJECT_ID }}"
+          else
+            echo "LLAMA_CLOUD_BASE_URL=https://api.staging.llamaindex.ai" >> $GITHUB_ENV
+            api_key_secret="${{ secrets.LLAMA_CLOUD_API_KEY_STAGING }}"
+            project_id_secret="${{ secrets.LLAMA_CLOUD_PROJECT_ID_STAGING }}"
+          fi
+
+          if [ -n "$api_key_secret" ]; then
+            echo "LLAMA_CLOUD_API_KEY=$api_key_secret" >> $GITHUB_ENV
+          fi
+
+          if [ -n "$project_id_secret" ]; then
+            echo "LLAMA_CLOUD_PROJECT_ID=$project_id_secret" >> $GITHUB_ENV
+          fi
+
+      - uses: actions/checkout@v5
+        with:
+          fetch-depth: 0
+
+      - name: Install uv
+        uses: astral-sh/setup-uv@v7
+        with:
+          version: ${{ env.UV_VERSION }}
+
+      - name: Set up Python
+        run: uv python install ${{ env.PYTHON_VERSION }} && uv python pin ${{ env.PYTHON_VERSION }}
+
+      - name: Run Extract E2E tests
+        id: extract-tests
+        continue-on-error: true
+        working-directory: py
+        run: |
+          set -o pipefail
+          rm -f "$API_E2E_LOG_PATH"
+          uv run pytest -v -n 8 --timeout=300 --session-timeout=1740 tests/extract/ 2>&1 | tee "$API_E2E_LOG_PATH"
+
+      - name: Extract pytest failure summary
+        id: failed-tests
+        if: steps.extract-tests.outcome == 'failure' || cancelled()
+        run: |
+          summary="$(python3 - <<'PY'
+          import os
+          import re
+          from pathlib import Path
+
+          log_path = Path(os.environ["API_E2E_LOG_PATH"])
+          if not log_path.exists():
+              print("Test log not found.")
+              raise SystemExit(0)
+
+          lines = log_path.read_text(errors="ignore").splitlines()
+
+          # Find the "short test summary info" section
+          start = None
+          for i, line in enumerate(lines):
+              if line.startswith("=") and "short test summary info" in line:
+                  start = i + 1
+                  break
+
+          if start is None:
+              print("No test summary found.")
+              raise SystemExit(0)
+
+          # Extract just the FAILED/ERROR lines (test name + short reason)
+          failed_tests = []
+          for line in lines[start:]:
+              if line.startswith("="):
+                  break  # End of section
+              if line.startswith("FAILED ") or line.startswith("ERROR "):
+                  # Extract test name and truncate the error message
+                  match = re.match(r"(FAILED|ERROR) ([\w/:.\[\]_-]+)", line)
+                  if match:
+                      failed_tests.append(f"{match.group(1)}: {match.group(2)}")
+
+          if failed_tests:
+              print("\n".join(failed_tests[:20]))  # Limit to 20 tests max
+          else:
+              print("No failed tests found in summary.")
+          PY
+          )"
+          if [ -z "$summary" ]; then
+            summary="Failed test summary not available. Review the full run logs."
+          fi
+          {
+            printf 'summary<<EOF\n%s\nEOF\n' "$summary"
+          } >> "$GITHUB_OUTPUT"
+
+      - name: Check test results
+        if: always()
+        run: |
+          if [ "${{ steps.extract-tests.outcome }}" == "failure" ]; then
+            echo "Extract E2E tests failed"
+            exit 1
+          fi
+
+      - name: Post to Extract Slack channel
+        id: slack
+        if: (failure() || cancelled()) && steps.runtime.outputs.notify_slack == 'true'
+        uses: slackapi/slack-github-action@v1.27.0
+        with:
+          channel-id: ${{ env.SLACK_CHANNEL_ID }}
+          slack-message: |
+            :red_circle: *Extract E2E Failed* (${{ steps.runtime.outputs.environment }})
+            ```
+            ${{ steps.failed-tests.outputs.summary }}
+            ```
+            <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|View Run>
+        env:
+          SLACK_BOT_TOKEN: ${{ secrets.SLACK_BOT_TOKEN }}
@@ -0,0 +1,183 @@
+"""
+Example: Batch Processing a Folder of PDFs with LlamaParse
+
+This script demonstrates how to process multiple PDFs from a folder
+using LlamaParse with controlled concurrency using asyncio and semaphores.
+
+Usage:
+    python batch_parse_folder.py --input-dir ./pdfs --max-concurrent 5
+"""
+
+import asyncio
+import argparse
+from pathlib import Path
+from typing import List, Dict, Any
+from datetime import datetime
+from dotenv import load_dotenv
+import os
+
+from llama_cloud_services import LlamaParse
+
+# Load environment variables from .env file
+load_dotenv()
+
+
+async def parse_single_file(
+    parser: LlamaParse,
+    file_path: Path,
+    semaphore: asyncio.Semaphore,
+) -> Dict[str, Any]:
+    """
+    Parse a single PDF file with concurrency control.
+
+    Args:
+        parser: LlamaParse instance
+        file_path: Path to the PDF file
+        semaphore: Semaphore to control concurrent requests
+
+    Returns:
+        Dictionary with file info and parse result
+    """
+    async with semaphore:
+        try:
+            print(f"Starting parse: {file_path.name}")
+
+            result = await parser.aparse(str(file_path))
+
+            print(f"✓ Completed: {file_path.name} ({len(result.pages)} pages)")
+
+            return {
+                "file": file_path.name,
+                "status": "success",
+                "result": result,
+                "pages": len(result.pages) if result.pages else 0,
+            }
+        except Exception as e:
+            print(f"✗ Error parsing {file_path.name}: {str(e)}")
+            return {
+                "file": file_path.name,
+                "status": "error",
+                "error": str(e),
+            }
+
+
+async def parse_folder(
+    input_dir: Path,
+    max_concurrent: int = 5,
+    api_key: str = None,
+) -> List[Dict[str, any]]:
+    """
+    Parse all PDFs in a folder with controlled concurrency.
+
+    Args:
+        input_dir: Directory containing PDF files
+        max_concurrent: Maximum number of concurrent parse operations
+        api_key: LlamaCloud API key (loaded from .env file)
+
+    Returns:
+        List of parse results for each file
+    """
+    # Find all PDF files
+    pdf_files = list(input_dir.glob("*.pdf"))
+
+    if not pdf_files:
+        print(f"No PDF files found in {input_dir}")
+        return []
+
+    print(f"Found {len(pdf_files)} PDF files to parse")
+
+    # Initialize parser
+    parser = LlamaParse(
+        api_key=api_key,
+        num_workers=1,  # We control concurrency with semaphore
+        show_progress=False,  # We'll show our own progress
+    )
+
+    # Create semaphore to limit concurrent requests
+    semaphore = asyncio.Semaphore(max_concurrent)
+
+    # Create tasks for all files
+    tasks = [parse_single_file(parser, pdf_file, semaphore) for pdf_file in pdf_files]
+
+    # Run all tasks concurrently (but limited by semaphore)
+    print(
+        f"Processing {len(tasks)} files with max {max_concurrent} concurrent operations..."
+    )
+    start_time = datetime.now()
+
+    results = await asyncio.gather(*tasks)
+
+    end_time = datetime.now()
+    duration = (end_time - start_time).total_seconds()
+
+    # Process results
+    successful = [
+        r for r in results if isinstance(r, dict) and r.get("status") == "success"
+    ]
+    failed = [r for r in results if isinstance(r, dict) and r.get("status") == "error"]
+
+    # Print summary
+    print("PARSE SUMMARY \n")
+    print(f"Total files: {len(pdf_files)}")
+    print(f"Successful: {len(successful)}")
+    print(f"Failed: {len(failed)}")
+    print(f"Total time: {duration:.2f} seconds")
+    print(f"Average time per file: {duration / len(pdf_files):.2f} seconds")
+
+    if failed:
+        print("\nFailed files:")
+        for result in failed:
+            print(f"  - {result['file']}: {result.get('error', 'Unknown error')}")
+
+    return results
+
+
+def main():
+    """Main entry point for the script."""
+    parser = argparse.ArgumentParser(
+        description="Batch process PDFs in a folder with LlamaParse"
+    )
+    parser.add_argument(
+        "--input-dir",
+        type=str,
+        required=True,
+        help="Directory containing PDF files to parse",
+    )
+    parser.add_argument(
+        "--max-concurrent",
+        type=int,
+        default=5,
+        help="Maximum number of concurrent parse operations (default: 5)",
+    )
+
+    args = parser.parse_args()
+
+    input_dir = Path(args.input_dir)
+
+    # Validate input directory
+    if not input_dir.exists():
+        print(f"Error: Input directory does not exist: {input_dir}")
+        return
+
+    if not input_dir.is_dir():
+        print(f"Error: Input path is not a directory: {input_dir}")
+        return
+
+    # Get API key from environment (loaded from .env file)
+    api_key = os.getenv("LLAMA_CLOUD_API_KEY")
+    if not api_key:
+        print("Error: LLAMA_CLOUD_API_KEY not found. Please set it in your .env file")
+        return
+
+    # Run async function
+    asyncio.run(
+        parse_folder(
+            input_dir=input_dir,
+            max_concurrent=args.max_concurrent,
+            api_key=api_key,
+        )
+    )
+
+
+if __name__ == "__main__":
+    main()
@@ -0,0 +1,540 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Document Splitting with LlamaCloud\n",
+    "\n",
+    "This notebook demonstrates how to use the LlamaCloud **Split** API to automatically segment a concatenated PDF into logical document sections based on content categories.\n",
+    "\n",
+    "## Use Case\n",
+    "\n",
+    "When dealing with large PDFs that contain multiple distinct documents or sections (e.g., a bundle of research papers, a collection of reports), you often need to split them into individual segments. The Split API uses AI to:\n",
+    "\n",
+    "1. Analyze each page's content\n",
+    "2. Classify pages into user-defined categories\n",
+    "3. Group consecutive pages of the same category into segments\n",
+    "\n",
+    "## Example Document\n",
+    "\n",
+    "We'll use a PDF containing three concatenated documents:\n",
+    "- **Alan Turing's essay** \"Intelligent Machinery, A Heretical Theory\" (an essay)\n",
+    "- **ImageNet paper** (a research paper)\n",
+    "- **\"Attention is All You Need\"** paper (a research paper)\n",
+    "\n",
+    "We'll split this into segments categorized as either `essay` or `research_paper`.\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Setup\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Requirement already satisfied: llama-cloud in /Users/javier/llama_cloud_services/.venv/lib/python3.11/site-packages (0.1.44)\n",
+      "Requirement already satisfied: python-dotenv in /Users/javier/llama_cloud_services/.venv/lib/python3.11/site-packages (1.2.1)\n",
+      "Requirement already satisfied: requests in /Users/javier/llama_cloud_services/.venv/lib/python3.11/site-packages (2.32.5)\n",
+      "Requirement already satisfied: certifi>=2024.7.4 in /Users/javier/llama_cloud_services/.venv/lib/python3.11/site-packages (from llama-cloud) (2025.11.12)\n",
+      "Requirement already satisfied: httpx>=0.20.0 in /Users/javier/llama_cloud_services/.venv/lib/python3.11/site-packages (from llama-cloud) (0.28.1)\n",
+      "Requirement already satisfied: pydantic>=1.10 in /Users/javier/llama_cloud_services/.venv/lib/python3.11/site-packages (from llama-cloud) (2.12.5)\n",
+      "Requirement already satisfied: charset_normalizer<4,>=2 in /Users/javier/llama_cloud_services/.venv/lib/python3.11/site-packages (from requests) (3.4.4)\n",
+      "Requirement already satisfied: idna<4,>=2.5 in /Users/javier/llama_cloud_services/.venv/lib/python3.11/site-packages (from requests) (3.11)\n",
+      "Requirement already satisfied: urllib3<3,>=1.21.1 in /Users/javier/llama_cloud_services/.venv/lib/python3.11/site-packages (from requests) (2.5.0)\n",
+      "Requirement already satisfied: anyio in /Users/javier/llama_cloud_services/.venv/lib/python3.11/site-packages (from httpx>=0.20.0->llama-cloud) (4.11.0)\n",
+      "Requirement already satisfied: httpcore==1.* in /Users/javier/llama_cloud_services/.venv/lib/python3.11/site-packages (from httpx>=0.20.0->llama-cloud) (1.0.9)\n",
+      "Requirement already satisfied: h11>=0.16 in /Users/javier/llama_cloud_services/.venv/lib/python3.11/site-packages (from httpcore==1.*->httpx>=0.20.0->llama-cloud) (0.16.0)\n",
+      "Requirement already satisfied: annotated-types>=0.6.0 in /Users/javier/llama_cloud_services/.venv/lib/python3.11/site-packages (from pydantic>=1.10->llama-cloud) (0.7.0)\n",
+      "Requirement already satisfied: pydantic-core==2.41.5 in /Users/javier/llama_cloud_services/.venv/lib/python3.11/site-packages (from pydantic>=1.10->llama-cloud) (2.41.5)\n",
+      "Requirement already satisfied: typing-extensions>=4.14.1 in /Users/javier/llama_cloud_services/.venv/lib/python3.11/site-packages (from pydantic>=1.10->llama-cloud) (4.15.0)\n",
+      "Requirement already satisfied: typing-inspection>=0.4.2 in /Users/javier/llama_cloud_services/.venv/lib/python3.11/site-packages (from pydantic>=1.10->llama-cloud) (0.4.2)\n",
+      "Requirement already satisfied: sniffio>=1.1 in /Users/javier/llama_cloud_services/.venv/lib/python3.11/site-packages (from anyio->httpx>=0.20.0->llama-cloud) (1.3.1)\n",
+      "\n",
+      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m A new release of pip is available: \u001b[0m\u001b[31;49m25.0.1\u001b[0m\u001b[39;49m -> \u001b[0m\u001b[32;49m25.3\u001b[0m\n",
+      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m To update, run: \u001b[0m\u001b[32;49mpip install --upgrade pip\u001b[0m\n",
+      "Note: you may need to restart the kernel to use updated packages.\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Install required packages\n",
+    "%pip install llama-cloud python-dotenv requests"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "✅ API configured with base URL: https://api.cloud.llamaindex.ai\n",
+      "✅ Project ID: using default project\n"
+     ]
+    }
+   ],
+   "source": [
+    "import os\n",
+    "import time\n",
+    "import requests\n",
+    "from dotenv import load_dotenv\n",
+    "\n",
+    "# Load environment variables\n",
+    "load_dotenv()\n",
+    "\n",
+    "# Configuration\n",
+    "LLAMA_CLOUD_API_KEY = os.environ.get(\"LLAMA_CLOUD_API_KEY\", \"llx-...\")\n",
+    "BASE_URL = os.environ.get(\"LLAMA_CLOUD_BASE_URL\", \"https://api.cloud.llamaindex.ai\")\n",
+    "PROJECT_ID = os.environ.get(\"LLAMA_CLOUD_PROJECT_ID\", None)\n",
+    "\n",
+    "# Headers for API requests\n",
+    "headers = {\n",
+    "    \"Authorization\": f\"Bearer {LLAMA_CLOUD_API_KEY}\",\n",
+    "    \"Content-Type\": \"application/json\",\n",
+    "}\n",
+    "\n",
+    "print(f\"✅ API configured with base URL: {BASE_URL}\")\n",
+    "print(f\"✅ Project ID: {PROJECT_ID or 'using default project'}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Step 1: Upload the PDF File\n",
+    "\n",
+    "First, we'll upload our concatenated PDF to LlamaCloud using the Files API. This can be done using the `llama-cloud` SDK.\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "📤 Uploading ./data/turing+imagenet+attention.pdf...\n",
+      "✅ File uploaded successfully!\n",
+      "   File name: turing+imagenet+attention.pdf\n"
+     ]
+    }
+   ],
+   "source": [
+    "from llama_cloud.client import LlamaCloud\n",
+    "\n",
+    "# Initialize the client\n",
+    "client = LlamaCloud(token=LLAMA_CLOUD_API_KEY, base_url=BASE_URL)\n",
+    "\n",
+    "# Path to the PDF file\n",
+    "pdf_path = \"./data/turing+imagenet+attention.pdf\"\n",
+    "\n",
+    "# Upload the file\n",
+    "print(f\"📤 Uploading {pdf_path}...\")\n",
+    "\n",
+    "with open(pdf_path, \"rb\") as f:\n",
+    "    uploaded_file = client.files.upload_file(upload_file=f, project_id=PROJECT_ID)\n",
+    "\n",
+    "file_id = uploaded_file.id\n",
+    "print(f\"✅ File uploaded successfully!\")\n",
+    "print(f\"   File name: {uploaded_file.name}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Step 2: Create a Split Job\n",
+    "\n",
+    "Now we'll create a split job using the Split API. Since the Split API is in beta and not yet available in the SDK, we'll use raw HTTP requests.\n",
+    "\n",
+    "We define two categories:\n",
+    "- **essay**: For philosophical or reflective writing\n",
+    "- **research_paper**: For formal academic documents with methodology and citations\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "🔄 Creating split job...\n",
+      "✅ Split job created!\n",
+      "   Job ID: spl-zsssb632a742aikliu96pqkb56t5\n",
+      "   Status: pending\n",
+      "   Categories: ['essay', 'research_paper']\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Define the split job request\n",
+    "split_request = {\n",
+    "    \"document_input\": {\n",
+    "        \"type\": \"file_id\",  # only file_id is supported for now\n",
+    "        \"value\": file_id,\n",
+    "    },\n",
+    "    \"categories\": [\n",
+    "        {\n",
+    "            \"name\": \"essay\",\n",
+    "            \"description\": \"A philosophical or reflective piece of writing that presents personal viewpoints, arguments, or thoughts on a topic without strict formal structure\",\n",
+    "        },\n",
+    "        {\n",
+    "            \"name\": \"research_paper\",\n",
+    "            \"description\": \"A formal academic document presenting original research, methodology, experiments, results, and conclusions with citations and references\",\n",
+    "        },\n",
+    "    ],\n",
+    "}\n",
+    "\n",
+    "# Create the split job\n",
+    "print(\"🔄 Creating split job...\")\n",
+    "response = requests.post(\n",
+    "    f\"{BASE_URL}/api/v1/beta/split/jobs\",\n",
+    "    params={\"project_id\": PROJECT_ID},\n",
+    "    headers=headers,\n",
+    "    json=split_request,\n",
+    ")\n",
+    "response.raise_for_status()\n",
+    "\n",
+    "split_job = response.json()\n",
+    "job_id = split_job[\"id\"]\n",
+    "\n",
+    "print(f\"✅ Split job created!\")\n",
+    "print(f\"   Job ID: {job_id}\")\n",
+    "print(f\"   Status: {split_job['status']}\")\n",
+    "print(f\"   Categories: {[c['name'] for c in split_job['categories']]}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Step 3: Poll for Job Completion\n",
+    "\n",
+    "The split job runs asynchronously. We'll poll the job status until it completes.\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "⏳ Waiting for split job to complete...\n",
+      "   Status: processing (elapsed: 0s)\n",
+      "   Status: processing (elapsed: 5s)\n",
+      "   Status: processing (elapsed: 11s)\n",
+      "   Status: completed (elapsed: 16s)\n",
+      "\n",
+      "✅ Split job completed successfully!\n"
+     ]
+    }
+   ],
+   "source": [
+    "def poll_split_job(job_id: str, max_wait_seconds: int = 180, poll_interval: int = 5):\n",
+    "    \"\"\"\n",
+    "    Poll a split job until it reaches a terminal state.\n",
+    "\n",
+    "    Args:\n",
+    "        job_id: The split job ID\n",
+    "        max_wait_seconds: Maximum time to wait for completion\n",
+    "        poll_interval: Seconds between poll attempts\n",
+    "\n",
+    "    Returns:\n",
+    "        The completed job response\n",
+    "    \"\"\"\n",
+    "    start_time = time.time()\n",
+    "\n",
+    "    while (time.time() - start_time) < max_wait_seconds:\n",
+    "        response = requests.get(\n",
+    "            f\"{BASE_URL}/api/v1/beta/split/jobs/{job_id}\",\n",
+    "            params={\"project_id\": PROJECT_ID},\n",
+    "            headers=headers,\n",
+    "        )\n",
+    "        response.raise_for_status()\n",
+    "        job = response.json()\n",
+    "\n",
+    "        status = job[\"status\"]\n",
+    "        elapsed = int(time.time() - start_time)\n",
+    "        print(f\"   Status: {status} (elapsed: {elapsed}s)\")\n",
+    "\n",
+    "        if status in [\"completed\", \"failed\"]:\n",
+    "            return job\n",
+    "\n",
+    "        time.sleep(poll_interval)\n",
+    "\n",
+    "    raise TimeoutError(f\"Job did not complete within {max_wait_seconds} seconds\")\n",
+    "\n",
+    "\n",
+    "print(\"⏳ Waiting for split job to complete...\")\n",
+    "completed_job = poll_split_job(job_id)\n",
+    "\n",
+    "if completed_job[\"status\"] == \"completed\":\n",
+    "    print(\"\\n✅ Split job completed successfully!\")\n",
+    "else:\n",
+    "    print(\n",
+    "        f\"\\n❌ Split job failed: {completed_job.get('error_message', 'Unknown error')}\"\n",
+    "    )"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Step 4: Analyze the Results\n",
+    "\n",
+    "Let's examine the split results to see how the document was segmented.\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "📊 Split Results Summary\n",
+      "==================================================\n",
+      "Total segments found: 3\n",
+      "\n",
+      "Segments by category:\n",
+      "   • essay: 1 segment(s)\n",
+      "   • research_paper: 2 segment(s)\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Get the segments from the result\n",
+    "segments = completed_job.get(\"result\", {}).get(\"segments\", [])\n",
+    "\n",
+    "print(f\"📊 Split Results Summary\")\n",
+    "print(f\"=\" * 50)\n",
+    "print(f\"Total segments found: {len(segments)}\")\n",
+    "print()\n",
+    "\n",
+    "# Count by category\n",
+    "category_counts = {}\n",
+    "for segment in segments:\n",
+    "    cat = segment[\"category\"]\n",
+    "    category_counts[cat] = category_counts.get(cat, 0) + 1\n",
+    "\n",
+    "print(\"Segments by category:\")\n",
+    "for cat, count in category_counts.items():\n",
+    "    print(f\"   • {cat}: {count} segment(s)\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "\n",
+      "📄 Segment Details\n",
+      "==================================================\n",
+      "\n",
+      "Segment 1:\n",
+      "   Category: essay\n",
+      "   Pages 1-4 (4 pages)\n",
+      "   Confidence: high\n",
+      "\n",
+      "Segment 2:\n",
+      "   Category: research_paper\n",
+      "   Pages 5-13 (9 pages)\n",
+      "   Confidence: high\n",
+      "\n",
+      "Segment 3:\n",
+      "   Category: research_paper\n",
+      "   Pages 14-24 (11 pages)\n",
+      "   Confidence: high\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Display detailed segment information\n",
+    "print(f\"\\n📄 Segment Details\")\n",
+    "print(f\"=\" * 50)\n",
+    "\n",
+    "for i, segment in enumerate(segments, 1):\n",
+    "    category = segment[\"category\"]\n",
+    "    pages = segment[\"pages\"]\n",
+    "    confidence = segment[\"confidence_category\"]\n",
+    "\n",
+    "    # Format page range\n",
+    "    if len(pages) == 1:\n",
+    "        page_range = f\"Page {pages[0]}\"\n",
+    "    else:\n",
+    "        page_range = f\"Pages {min(pages)}-{max(pages)}\"\n",
+    "\n",
+    "    print(f\"\\nSegment {i}:\")\n",
+    "    print(f\"   Category: {category}\")\n",
+    "    print(f\"   {page_range} ({len(pages)} page{'s' if len(pages) > 1 else ''})\")\n",
+    "    print(f\"   Confidence: {confidence}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Expected Results\n",
+    "\n",
+    "Based on our test document, we expect:\n",
+    "- **1 essay segment**: Alan Turing's \"Intelligent Machinery, A Heretical Theory\"\n",
+    "- **2 research paper segments**: ImageNet paper and \"Attention is All You Need\" paper\n",
+    "\n",
+    "The pages should be grouped consecutively, with no overlap between segments.\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "\n",
+      "✅ Validation\n",
+      "==================================================\n",
+      "Total pages assigned: 24\n",
+      "Unique pages: 24\n",
+      "✅ No page overlap detected - each page belongs to exactly one segment\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Verify no page overlap\n",
+    "all_pages = []\n",
+    "for segment in segments:\n",
+    "    all_pages.extend(segment[\"pages\"])\n",
+    "\n",
+    "unique_pages = set(all_pages)\n",
+    "\n",
+    "print(f\"\\n✅ Validation\")\n",
+    "print(f\"=\" * 50)\n",
+    "print(f\"Total pages assigned: {len(all_pages)}\")\n",
+    "print(f\"Unique pages: {len(unique_pages)}\")\n",
+    "\n",
+    "if len(all_pages) == len(unique_pages):\n",
+    "    print(f\"✅ No page overlap detected - each page belongs to exactly one segment\")\n",
+    "else:\n",
+    "    print(\n",
+    "        f\"⚠️  Page overlap detected - {len(all_pages) - len(unique_pages)} duplicate assignments\"\n",
+    "    )"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Using `allow_uncategorized` Strategy\n",
+    "\n",
+    "You can also use the `allow_uncategorized` splitting strategy. This is useful when you want to capture pages that don't match any defined category.\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "📝 With allow_uncategorized=True and only 'essay' category defined,\n",
+      "   pages that don't match 'essay' will be grouped as 'uncategorized'.\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Example with allow_uncategorized strategy\n",
+    "split_request_uncategorized = {\n",
+    "    \"document_input\": {\"type\": \"file_id\", \"value\": file_id},\n",
+    "    \"categories\": [\n",
+    "        {\n",
+    "            \"name\": \"essay\",\n",
+    "            \"description\": \"A philosophical or reflective piece of writing that presents personal viewpoints, arguments, or thoughts on a topic\",\n",
+    "        }\n",
+    "        # Note: We only define 'essay' category\n",
+    "        # Research papers will be classified as 'uncategorized'\n",
+    "    ],\n",
+    "    \"splitting_strategy\": {\"allow_uncategorized\": True},\n",
+    "}\n",
+    "\n",
+    "print(\"📝 With allow_uncategorized=True and only 'essay' category defined,\")\n",
+    "print(\"   pages that don't match 'essay' will be grouped as 'uncategorized'.\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Conclusion\n",
+    "\n",
+    "The LlamaCloud Split API provides a powerful way to automatically segment concatenated documents based on content categories. This is useful for:\n",
+    "\n",
+    "- **Document processing pipelines**: Automatically separate bundled documents before further processing\n",
+    "- **Content organization**: Categorize and organize mixed document collections\n",
+    "- **Information extraction**: Identify different document types within a single file\n",
+    "\n",
+    "### Key Features\n",
+    "\n",
+    "- **AI-powered classification**: Uses LLMs to understand page content and assign categories\n",
+    "- **Flexible categories**: Define any categories relevant to your use case\n",
+    "- **Confidence scoring**: Each segment includes a confidence level\n",
+    "- **Page-level granularity**: Results include exact page numbers for each segment\n",
+    "\n",
+    "### API Reference\n",
+    "\n",
+    "- **Create Split Job**: `POST /api/v1/beta/split/jobs`\n",
+    "- **Get Split Job**: `GET /api/v1/beta/split/jobs/{job_id}`\n",
+    "- **List Split Jobs**: `GET /api/v1/beta/split/jobs`\n"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": ".venv",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}
@@ -398,6 +398,6 @@ Another option (orthogonal to the above) is to break the document into smaller s
 ## Additional Resources

 - [Extract Documentation](https://docs.cloud.llamaindex.ai/llamaextract/getting_started) - Details on Extract features, API and examples.
- [Example Notebook](docs/examples-py/extract/resume_screening.ipynb) - Detailed walkthrough of resume parsing
+- [Example Notebook](examples/extract/resume_screening.ipynb) - Detailed walkthrough of resume parsing
 - [Example Application with TypeScript](./examples-ts/extract/) - End-to-end examples using LlamaExtract TypeScript client.
 - [Discord Community](https://discord.com/invite/eN6D2HQ4aX) - Get help and share feedback
@@ -97,7 +97,7 @@ for page in result.pages:
    print(page.structuredData)
 ```

-See more details about the result object in the [example notebook](./docs/examples-py/parse/demo_json_tour.ipynb).
+See more details about the result object in the [example notebook](./examples/parse/demo_json_tour.ipynb).

 ### Using with file object / bytes

@@ -153,10 +153,10 @@ Full documentation for `SimpleDirectoryReader` can be found on the [LlamaIndex D

 Several end-to-end indexing examples can be found in the examples folder

- [Getting Started](docs/examples-py/parse/demo_basic.ipynb)
- [Advanced RAG Example](docs/examples-py/parse/demo_advanced.ipynb)
- [Raw API Usage](docs/examples-py/parse/demo_api.ipynb)
- [Result Object Tour](docs/examples-py/parse/demo_json_tour.ipynb)
+- [Getting Started](examples/parse/demo_basic.ipynb)
+- [Advanced RAG Example](examples/parse/demo_advanced.ipynb)
+- [Raw API Usage](examples/parse/demo_api.ipynb)
+- [Result Object Tour](examples/parse/demo_json_tour.ipynb)

 ## Documentation

@@ -1,5 +1,24 @@
 # llama-cloud-services-py

+## 0.6.91
+
+### Patch Changes
+
+- 07ec282: Bump up patch versions for python packages
+- 3040951: Use error description in ExtractedData invalid extraction error
+
+## 0.6.90
+
+### Patch Changes
+
+- 19cbb25: Remove extension filter
+
+## 0.6.89
+
+### Patch Changes
+
+- b9b83c9: Parse bounding boxes from extract jobs results in agent data
+
 ## 0.6.88

 ### Patch Changes
@@ -11,6 +11,9 @@ from .schema import (
    InvalidExtractionData,
    ExtractedFieldMetadata,
    ExtractedFieldMetaDataDict,
+    FieldCitation,
+    BoundingBox,
+    PageDimensions,
 )
 from .client import AsyncAgentDataClient

@@ -28,4 +31,7 @@ __all__ = [
    "InvalidExtractionData",
    "ExtractedFieldMetadata",
    "ExtractedFieldMetaDataDict",
+    "FieldCitation",
+    "BoundingBox",
+    "PageDimensions",
 ]
@@ -174,6 +174,22 @@ class TypedAgentDataItems(BaseModel, Generic[AgentDataT]):
    )


+class BoundingBox(BaseModel):
+    """Bounding box coordinates for a citation location on a page."""
+
+    x: float = Field(description="X coordinate of the bounding box origin")
+    y: float = Field(description="Y coordinate of the bounding box origin")
+    w: float = Field(description="Width of the bounding box")
+    h: float = Field(description="Height of the bounding box")
+
+
+class PageDimensions(BaseModel):
+    """Dimensions of a page in the source document."""
+
+    width: float = Field(description="Width of the page")
+    height: float = Field(description="Height of the page")
+
+
 class FieldCitation(BaseModel):
    page: Optional[int] = Field(
        None, description="The page number that the field occurred on"
@@ -182,6 +198,14 @@ class FieldCitation(BaseModel):
        None,
        description="The original text this field's value was derived from",
    )
+    bounding_boxes: Optional[List[BoundingBox]] = Field(
+        None,
+        description="Bounding boxes indicating where the citation appears on the page",
+    )
+    page_dimensions: Optional[PageDimensions] = Field(
+        None,
+        description="Dimensions of the page containing the citation",
+    )


 class ExtractedFieldMetadata(BaseModel):
@@ -201,6 +225,10 @@ class ExtractedFieldMetadata(BaseModel):
        None,
        description="The confidence score for the field based on the extracted text only",
    )
+    parsing_confidence: Optional[float] = Field(
+        None,
+        description="The confidence score for the field based on the parsing/OCR quality",
+    )
    citation: Optional[List[FieldCitation]] = Field(
        None,
        description="The citation for the field, including page number and matching text",
@@ -447,26 +475,49 @@ class ExtractedData(BaseModel, Generic[ExtractedT]):
                },
            )
        except ValidationError as e:
+            # Capture the job-level error from the extraction run if available
+            job_error = result.error
+
            invalid_item = ExtractedData[Dict[str, Any]].create(
                data=result.data or {},
                status="error",
                field_metadata=field_metadata,
-                metadata={"extraction_error": str(e), **(metadata or {})},
+                metadata={
+                    "extraction_error": str(e),
+                    **({"job_error": job_error} if job_error else {}),
+                    **(metadata or {}),
+                },
                file_id=file_id,
                file_name=file_name,
                file_hash=file_hash,
            )
-            raise InvalidExtractionData(invalid_item) from e
+            raise InvalidExtractionData(invalid_item, extraction_error=job_error) from e


 class InvalidExtractionData(Exception):
    """
    Exception raised when the extracted data does not conform to the schema.
+
+    Attributes:
+        invalid_item: The ExtractedData instance containing the invalid data and metadata
+        extraction_error: The error message from the extraction job, if available
    """

-    def __init__(self, invalid_item: ExtractedData[Dict[str, Any]]):
+    def __init__(
+        self,
+        invalid_item: ExtractedData[Dict[str, Any]],
+        extraction_error: Optional[str] = None,
+    ):
        self.invalid_item = invalid_item
-        super().__init__("Not able to parse the extracted data, parsed invalid format")
+        self.extraction_error = extraction_error
+
+        # Build an informative error message
+        if extraction_error:
+            message = f"Extraction error: {extraction_error}"
+        else:
+            message = "Not able to parse the extracted data, parsed invalid format"
+
+        super().__init__(message)


 def calculate_overall_confidence(
@@ -4,10 +4,11 @@ import os
 import time
 from io import BufferedIOBase, TextIOWrapper
 from pathlib import Path
-from typing import List, Optional, Type, Union, Coroutine, Any, TypeVar
+from typing import Callable, List, Optional, Type, Union, Coroutine, Any, TypeVar
 import warnings
 import httpx
 from pydantic import BaseModel
+from functools import wraps
 from tenacity import (
    retry_if_exception,
    stop_after_attempt,
@@ -54,7 +55,7 @@ DEFAULT_EXTRACT_CONFIG = ExtractConfig(
 def _is_retryable_error(exception: BaseException) -> bool:
    """Check if an exception is retryable."""
    if isinstance(exception, ApiError):
-        return exception.status_code in (502, 503, 504, 425, 408)
+        return exception.status_code in (429, 500, 502, 503, 504, 425, 408)
    elif isinstance(
        exception, (httpx.HTTPStatusError, httpx.RequestError, httpx.TimeoutException)
    ):
@@ -62,6 +63,33 @@ def _is_retryable_error(exception: BaseException) -> bool:
    return False


+def _async_retry(
+    max_attempts: int = 5,
+    initial_wait: float = 1,
+    max_wait: float = 30,
+    jitter: float = 3,
+) -> Callable:
+    """Decorator for async functions with retry logic for rate limiting and transient errors."""
+
+    def decorator(func: Callable) -> Callable:
+        @wraps(func)
+        async def wrapper(*args: Any, **kwargs: Any) -> Any:
+            async for attempt in AsyncRetrying(
+                retry=retry_if_exception(_is_retryable_error),
+                stop=stop_after_attempt(max_attempts),
+                wait=wait_exponential_jitter(
+                    initial=initial_wait, max=max_wait, jitter=jitter
+                ),
+                reraise=True,
+            ):
+                with attempt:
+                    return await func(*args, **kwargs)
+
+        return wrapper
+
+    return decorator
+
+
 async def _validate_schema(
    client: AsyncLlamaCloud, data_schema: SchemaInput
 ) -> JSONObjectType:
@@ -82,50 +110,6 @@ async def _validate_schema(
    return validated_schema.data_schema


-async def _get_job_with_retry(
-    client: AsyncLlamaCloud,
-    job_id: str,
-    max_attempts: int = 5,
-    initial_wait: float = 1,
-    max_wait: float = 60,
-    jitter: float = 5,
-) -> ExtractJob:
-    """Get extraction job with retry logic."""
-    async for attempt in AsyncRetrying(
-        retry=retry_if_exception(_is_retryable_error),
-        stop=stop_after_attempt(max_attempts),
-        wait=wait_exponential_jitter(initial=initial_wait, max=max_wait, jitter=jitter),
-        reraise=True,
-    ):
-        with attempt:
-            return await client.llama_extract.get_job(job_id=job_id)
-
-
-async def _get_run_with_retry(
-    client: AsyncLlamaCloud,
-    job_id: str,
-    project_id: Optional[str] = None,
-    organization_id: Optional[str] = None,
-    max_attempts: int = 3,
-    initial_wait: float = 1,
-    max_wait: float = 20,
-    jitter: float = 3,
-) -> ExtractRun:
-    """Get extraction run with retry logic."""
-    async for attempt in AsyncRetrying(
-        retry=retry_if_exception(_is_retryable_error),
-        stop=stop_after_attempt(max_attempts),
-        wait=wait_exponential_jitter(initial=initial_wait, max=max_wait, jitter=jitter),
-        reraise=True,
-    ):
-        with attempt:
-            return await client.llama_extract.get_run_by_job_id(
-                job_id=job_id,
-                project_id=project_id,
-                organization_id=organization_id,
-            )
-
-
 async def _wait_for_job_result(
    client: AsyncLlamaCloud,
    job_id: str,
@@ -142,30 +126,33 @@ async def _wait_for_job_result(
    run_jitter: float = 3,
 ) -> Optional[ExtractRun]:
    """Wait for and return the results of an extraction job."""
+
+    @_async_retry(
+        max_attempts=job_retry_attempts, max_wait=job_max_wait, jitter=job_jitter
+    )
+    async def _get_job() -> ExtractJob:
+        return await client.llama_extract.get_job(job_id=job_id)
+
+    @_async_retry(
+        max_attempts=run_retry_attempts, max_wait=run_max_wait, jitter=run_jitter
+    )
+    async def _get_run() -> ExtractRun:
+        return await client.llama_extract.get_run_by_job_id(
+            job_id=job_id,
+            project_id=project_id,
+            organization_id=organization_id,
+        )
+
    start = time.perf_counter()
    poll_count = 0

    while True:
        await asyncio.sleep(check_interval)
        poll_count += 1
-        job = await _get_job_with_retry(
-            client,
-            job_id,
-            max_attempts=job_retry_attempts,
-            max_wait=job_max_wait,
-            jitter=job_jitter,
-        )
+        job = await _get_job()

        if job.status == StatusEnum.SUCCESS:
-            return await _get_run_with_retry(
-                client,
-                job_id,
-                project_id,
-                organization_id,
-                max_attempts=run_retry_attempts,
-                max_wait=run_max_wait,
-                jitter=run_jitter,
-            )
+            return await _get_run()
        elif job.status == StatusEnum.PENDING:
            end = time.perf_counter()
            if end - start > max_timeout:
@@ -177,15 +164,7 @@ async def _wait_for_job_result(
            warnings.warn(
                f"Failure in job: {job_id}, status: {job.status}, error: {job.error}"
            )
-            return await _get_run_with_retry(
-                client,
-                job_id,
-                project_id,
-                organization_id,
-                max_attempts=run_retry_attempts,
-                max_wait=run_max_wait,
-                jitter=run_jitter,
-            )
+            return await _get_run()


 def run_in_thread(
@@ -498,9 +477,12 @@ class ExtractionAgent:
        Args:
            run_id (str): The ID of the extraction run to delete
        """
-        self._run_in_thread(
-            self._client.llama_extract.delete_extraction_run(run_id=run_id)
-        )
+
+        @_async_retry()
+        async def _delete() -> None:
+            return await self._client.llama_extract.delete_extraction_run(run_id=run_id)
+
+        self._run_in_thread(_delete())

    def list_extraction_runs(
        self, page: int = 0, limit: int = 100
@@ -510,13 +492,16 @@ class ExtractionAgent:
        Returns:
            PaginatedExtractRunsResponse: Paginated list of extraction runs
        """
-        return self._run_in_thread(
-            self._client.llama_extract.list_extract_runs(
+
+        @_async_retry()
+        async def _list() -> PaginatedExtractRunsResponse:
+            return await self._client.llama_extract.list_extract_runs(
                extraction_agent_id=self.id,
                skip=page * limit,
                limit=limit,
            )
-        )
+
+        return self._run_in_thread(_list())

    def __repr__(self) -> str:
        return f"ExtractionAgent(id={self.id}, name={self.name})"
@@ -658,15 +643,17 @@ class LlamaExtract(BaseComponent):
                "data_schema must be either a dictionary or a Pydantic model"
            )

-        agent = self._run_in_thread(
-            self._async_client.llama_extract.create_extraction_agent(
+        @_async_retry()
+        async def _create() -> CloudExtractAgent:
+            return await self._async_client.llama_extract.create_extraction_agent(
                project_id=self._project_id,
                organization_id=self._organization_id,
                name=name,
                data_schema=data_schema,
                config=config,
            )
-        )
+
+        agent = self._run_in_thread(_create())

        return ExtractionAgent(
            client=self._async_client,
@@ -702,19 +689,27 @@ class LlamaExtract(BaseComponent):
            )

        if id:
-            agent = self._run_in_thread(
-                self._async_client.llama_extract.get_extraction_agent(
+
+            @_async_retry()
+            async def _get_by_id() -> CloudExtractAgent:
+                return await self._async_client.llama_extract.get_extraction_agent(
                    extraction_agent_id=id,
                )
-            )
+
+            agent = self._run_in_thread(_get_by_id())

        elif name:
-            agent = self._run_in_thread(
-                self._async_client.llama_extract.get_extraction_agent_by_name(
-                    name=name,
-                    project_id=self._project_id,
+
+            @_async_retry()
+            async def _get_by_name() -> CloudExtractAgent:
+                return (
+                    await self._async_client.llama_extract.get_extraction_agent_by_name(
+                        name=name,
+                        project_id=self._project_id,
+                    )
                )
-            )
+
+            agent = self._run_in_thread(_get_by_name())
        else:
            raise ValueError("Either name or extraction_agent_id must be provided.")

@@ -734,11 +729,14 @@ class LlamaExtract(BaseComponent):

    def list_agents(self) -> List[ExtractionAgent]:
        """List all available extraction agents."""
-        agents = self._run_in_thread(
-            self._async_client.llama_extract.list_extraction_agents(
+
+        @_async_retry()
+        async def _list() -> List[CloudExtractAgent]:
+            return await self._async_client.llama_extract.list_extraction_agents(
                project_id=self._project_id,
            )
-        )
+
+        agents = self._run_in_thread(_list())

        return [
            ExtractionAgent(
@@ -763,11 +761,14 @@ class LlamaExtract(BaseComponent):
        Args:
            agent_id (str): ID of the extraction agent to delete
        """
-        self._run_in_thread(
-            self._async_client.llama_extract.delete_extraction_agent(
-                extraction_agent_id=agent_id
+
+        @_async_retry()
+        async def _delete() -> None:
+            return await self._async_client.llama_extract.delete_extraction_agent(
+                extraction_agent_id=agent_id,
            )
-        )
+
+        self._run_in_thread(_delete())

    async def _wait_for_job_result(self, job_id: str) -> Optional[ExtractRun]:
        """Wait for and return the results of an extraction job."""
@@ -3,7 +3,7 @@ from typing import BinaryIO
 import os
 from pathlib import Path
 from llama_cloud.client import AsyncLlamaCloud
-from llama_cloud.types import File, FileCreate
+from llama_cloud.types import File
 from typing import Optional
 from llama_cloud_services.utils import SourceText, FileInput

@@ -73,11 +73,9 @@ class FileClient:
            presigned_url = await self.client.files.generate_presigned_url(
                project_id=self.project_id,
                organization_id=self.organization_id,
-                request=FileCreate(
-                    name=name,
-                    external_file_id=external_file_id,
-                    file_size=file_size,
-                ),
+                name=name,
+                external_file_id=external_file_id,
+                file_size=file_size,
            )
            httpx_client = self.client._client_wrapper.httpx_client
            upload_response = await httpx_client.put(
@@ -21,7 +21,6 @@ from llama_cloud import (
    PipelineCreateTransformConfig,
    PipelineFileCreateCustomMetadataValue,
    PipelineType,
-    ProjectCreate,
    ManagedIngestionStatus,
    CloudDocumentCreate,
    CloudDocument,
@@ -507,14 +506,19 @@ class LlamaCloudIndex(BaseManagedIndex):
        client = get_client(api_key, base_url, app_url, timeout)

        if project_id is None:
-            # create project if it doesn't exist
-            project = client.projects.upsert_project(
+            # get project by name
+            projects = client.projects.list_projects(
                organization_id=organization_id,
-                request=ProjectCreate(name=project_name),
+                project_name=project_name,
            )
+            if not projects:
+                raise ValueError(
+                    f"Project '{project_name}' not found. Please create it first in the LlamaCloud UI."
+                )
+            project = projects[0]
            project_id = project.id
            if verbose:
-                print(f"Created project {project_id} with name {project_name}")
+                print(f"Found project {project_id} with name {project_name}")

        # create pipeline
        pipeline_create = PipelineCreate(
@@ -563,15 +567,20 @@ class LlamaCloudIndex(BaseManagedIndex):
        app_url = app_url or os.environ.get("LLAMA_CLOUD_APP_URL", DEFAULT_APP_URL)
        aclient = get_aclient(api_key, base_url, app_url, timeout)

-        # create project if it doesn't exist
-        project = await aclient.projects.upsert_project(
-            organization_id=organization_id, request=ProjectCreate(name=project_name)
+        # get project by name
+        projects = await aclient.projects.list_projects(
+            organization_id=organization_id, project_name=project_name
        )
+        if not projects:
+            raise ValueError(
+                f"Project '{project_name}' not found. Please create it first in the LlamaCloud UI."
+            )
+        project = projects[0]
        if project.id is None:
-            raise ValueError(f"Failed to create/get project {project_name}")
+            raise ValueError(f"Failed to get project {project_name}")

        if verbose:
-            print(f"Created project {project.id} with name {project.name}")
+            print(f"Found project {project.id} with name {project.name}")

        # create pipeline
        pipeline_create = PipelineCreate(
@@ -751,11 +751,9 @@ class LlamaParse(BasePydanticReader):
            file_path = str(file_input)
            file_ext = os.path.splitext(file_path)[1].lower()
            if file_ext not in SUPPORTED_FILE_TYPES:
-                raise Exception(
-                    f"Currently, only the following file types are supported: {SUPPORTED_FILE_TYPES}\n"
-                    f"Current file type: {file_ext}"
-                )
-            mime_type = mimetypes.guess_type(file_path)[0]
+                mime_type = "application/octet-stream"
+            else:
+                mime_type = mimetypes.guess_type(file_path)[0]
            # Open the file here for the duration of the async context
            # load data, set the mime type
            fs = fs or get_default_fs()
@@ -1,5 +1,29 @@
 # llama_parse

+## 0.6.91
+
+### Patch Changes
+
+- 07ec282: Bump up patch versions for python packages
+- Updated dependencies [07ec282]
+- Updated dependencies [3040951]
+  - llama-cloud-services-py@0.6.91
+
+## 0.6.90
+
+### Patch Changes
+
+- 19cbb25: Remove extension filter
+- Updated dependencies [19cbb25]
+  - llama-cloud-services-py@0.6.90
+
+## 0.6.89
+
+### Patch Changes
+
+- Updated dependencies [b9b83c9]
+  - llama-cloud-services-py@0.6.89
+
 ## 0.6.88

 ### Patch Changes
@@ -146,9 +146,9 @@ Full documentation for `SimpleDirectoryReader` can be found on the [LlamaIndex D

 Several end-to-end indexing examples can be found in the examples folder

- [Getting Started](/docs/examples-py/parse/demo_basic.ipynb)
- [Advanced RAG Example](/docs/examples-py/parse/demo_advanced.ipynb)
- [Raw API Usage](/docs/examples-py/parse/demo_api.ipynb)
+- [Getting Started](../../examples/parse/demo_basic.ipynb)
+- [Advanced RAG Example](../../examples/parse/demo_advanced.ipynb)
+- [Raw API Usage](../../examples/parse/demo_api.ipynb)

 ## Documentation

@@ -1,6 +1,6 @@
 {
  "name": "llama_parse",
-  "version": "0.6.88",
+  "version": "0.6.91",
  "description": "",
  "main": "index.js",
  "private": false,
@@ -11,13 +11,13 @@ dev = [

 [project]
 name = "llama-parse"
-version = "0.6.88"
+version = "0.6.91"
 description = "Parse files into RAG-Optimized formats."
 authors = [{name = "Logan Markewich", email = "logan@llamaindex.ai"}]
 requires-python = ">=3.9,<4.0"
 readme = "README.md"
 license = "MIT"
-dependencies = ["llama-cloud-services>=0.6.88"]
+dependencies = ["llama-cloud-services>=0.6.91"]

 [project.scripts]
 llama-parse = "llama_parse.cli.main:parse"
@@ -1,6 +1,6 @@
 {
  "name": "llama-cloud-services-py",
-  "version": "0.6.88",
+  "version": "0.6.91",
  "private": false,
  "license": "MIT",
  "scripts": {},
@@ -23,7 +23,7 @@ dev = [

 [project]
 name = "llama-cloud-services"
-version = "0.6.88"
+version = "0.6.91"
 description = "Tailored SDK clients for LlamaCloud services."
 authors = [{name = "Logan Markewich", email = "logan@runllama.ai"}]
 requires-python = ">=3.9,<4.0"
@@ -31,7 +31,7 @@ readme = "README.md"
 license = "MIT"
 dependencies = [
  "llama-index-core>=0.12.0",
-  "llama-cloud==0.1.45",
+  "llama-cloud==0.1.46",
  "pydantic>=2.8,!=2.10",
  "click>=8.1.7,<9",
  "python-dotenv>=1.0.1,<2",
@@ -1,11 +1,43 @@
 import os
-from typing import List
-from llama_cloud_services.extract import LlamaExtract
+from typing import Any, Dict, List, Optional, Union
+
+from llama_cloud.core.api_error import ApiError
+from llama_cloud.types import ExtractConfig
+from pydantic import BaseModel
+from tenacity import (
+    retry,
+    retry_if_exception,
+    stop_after_attempt,
+    wait_exponential,
+)
+
+from llama_cloud_services.extract import ExtractionAgent, LlamaExtract

 # Global storage for agents to cleanup
 _TEST_AGENTS_TO_CLEANUP: List[str] = []


+def _is_rate_limit_error(exception: BaseException) -> bool:
+    """Check if the exception is a rate limit error (429)."""
+    return isinstance(exception, ApiError) and exception.status_code == 429
+
+
+@retry(
+    retry=retry_if_exception(_is_rate_limit_error),
+    wait=wait_exponential(multiplier=1, min=1, max=30),
+    stop=stop_after_attempt(5),
+    reraise=True,
+)
+def create_agent_with_retry(
+    extractor: LlamaExtract,
+    name: str,
+    data_schema: Union[Dict[str, Any], type[BaseModel]],
+    config: Optional[ExtractConfig] = None,
+) -> ExtractionAgent:
+    """Create an extraction agent with retry logic for rate limiting."""
+    return extractor.create_agent(name=name, data_schema=data_schema, config=config)
+
+
 def pytest_configure(config):
    """Register custom markers for extract tests."""
    config.addinivalue_line("markers", "agent_name: custom agent name for test")
@@ -6,7 +6,7 @@ from pydantic import BaseModel
 from llama_cloud_services.extract import LlamaExtract, ExtractionAgent, SourceText
 from llama_cloud.types import ExtractConfig, ExtractMode, ExtractRun
 from tests.extract.util import load_test_dotenv
-from .conftest import register_agent_for_cleanup
+from .conftest import register_agent_for_cleanup, create_agent_with_retry

 load_test_dotenv()

@@ -87,7 +87,11 @@ def test_agent(llama_extract, test_agent_name, test_schema_dict, request):
    except Exception as e:
        print(f"Warning: Failed to cleanup existing agent: {e}")

-    agent = llama_extract.create_agent(name=name, data_schema=schema)
+    # Use config with cache invalidation to ensure fresh results in tests
+    config = ExtractConfig(invalidate_cache=True)
+    agent = create_agent_with_retry(
+        llama_extract, name=name, data_schema=schema, config=config
+    )

    # Add agent to cleanup list via conftest helper
    register_agent_for_cleanup(agent.id)
@@ -237,7 +241,7 @@ class TestStatelessExtraction:

    @pytest.fixture
    def test_config(self):
-        return ExtractConfig(extraction_mode=ExtractMode.FAST)
+        return ExtractConfig(extraction_mode=ExtractMode.FAST, invalidate_cache=True)

    @pytest.fixture
    def test_schema_dict(self):
@@ -8,7 +8,7 @@ import uuid
 from llama_cloud.types import ExtractConfig, ExtractMode
 from deepdiff import DeepDiff
 from tests.extract.util import json_subset_match_score, load_test_dotenv
-from .conftest import register_agent_for_cleanup
+from .conftest import register_agent_for_cleanup, create_agent_with_retry

 load_test_dotenv()

@@ -56,10 +56,16 @@ def get_test_cases():
            input_files.append(file_path)

        settings = [
-            ExtractConfig(extraction_mode=ExtractMode.FAST),
-            ExtractConfig(extraction_mode=ExtractMode.BALANCED),
-            ExtractConfig(extraction_mode=ExtractMode.MULTIMODAL),
-            ExtractConfig(extraction_mode=ExtractMode.PREMIUM),
+            ExtractConfig(extraction_mode=ExtractMode.FAST, invalidate_cache=True),
+            ExtractConfig(extraction_mode=ExtractMode.BALANCED, invalidate_cache=True),
+            ExtractConfig(
+                extraction_mode=ExtractMode.MULTIMODAL, invalidate_cache=True
+            ),
+            ExtractConfig(
+                extraction_mode=ExtractMode.PREMIUM,
+                invalidate_cache=True,
+                parse_model="anthropic-sonnet-4.5",
+            ),
        ]

        for input_file in sorted(input_files):
@@ -117,8 +123,10 @@ def extraction_agent(test_case: ExtractionTestCase, extractor: LlamaExtract):
    except Exception as e:
        print(f"Warning: Failed to cleanup existing agent: {str(e)}")

-    # Create new agent
-    agent = extractor.create_agent(agent_name, schema, config=test_case.config)
+    # Create new agent with retry logic for rate limiting
+    agent = create_agent_with_retry(
+        extractor, name=agent_name, data_schema=schema, config=test_case.config
+    )

    # Register agent for cleanup at the end of the test session
    register_agent_for_cleanup(agent.id)
@@ -8,7 +8,6 @@ from llama_cloud import (
    AutoTransformConfig,
    PipelineCreate,
    PipelineFileCreate,
-    ProjectCreate,
    CompositeRetrievalMode,
    LlamaParseParameters,
    ReRankConfig,
@@ -60,11 +59,15 @@ def local_figures_file() -> str:
 def _setup_index_with_file(
    client: LlamaCloud, index_name: str, remote_file: Tuple[str, str]
 ) -> LlamaCloudIndex:
-    # create project if it doesn't exist
-    project_create = ProjectCreate(name=project_name)
-    project = client.projects.upsert_project(
-        organization_id=organization_id, request=project_create
+    # get project by name
+    projects = client.projects.list_projects(
+        organization_id=organization_id, project_name=project_name
    )
+    if not projects:
+        raise ValueError(
+            f"Project '{project_name}' not found. Please create it first in the LlamaCloud UI."
+        )
+    project = projects[0]

    # create pipeline
    pipeline_create = PipelineCreate(
@@ -11,10 +11,12 @@ from llama_cloud.types.aggregate_group import AggregateGroup
 from pydantic import BaseModel, Field, ValidationError

 from llama_cloud_services.beta.agent_data.schema import (
+    BoundingBox,
    ExtractedData,
    ExtractedFieldMetadata,
    FieldCitation,
    InvalidExtractionData,
+    PageDimensions,
    TypedAgentData,
    TypedAggregateGroup,
    calculate_overall_confidence,
@@ -421,6 +423,7 @@ def create_extract_run(
    },
    data_schema: Dict[str, Any] = {},
    file: File = create_file(),
+    error: Optional[str] = None,
 ) -> ExtractRun:
    return ExtractRun.parse_obj(
        {
@@ -437,6 +440,7 @@ def create_extract_run(
            "status": "SUCCESS",
            "project_id": str(uuid.uuid4()),
            "from_ui": False,
+            "error": error,
        }
    )

@@ -542,6 +546,46 @@ def test_extracted_data_from_extraction_result_invalid_data():
    assert invalid_data.field_metadata["name"].confidence == 0.9
    assert invalid_data.overall_confidence == 0.9

+    # Verify default error message when no job error present
+    assert exc_info.value.extraction_error is None
+    assert "Not able to parse the extracted data" in str(exc_info.value)
+
+
+def test_extracted_data_from_extraction_result_with_job_error():
+    """Test ExtractedData.from_extraction_result with job-level error prominently displayed."""
+    job_error_message = "Failed to process document: unsupported file format"
+
+    # Create ExtractRun with both invalid data AND a job-level error
+    extract_run = create_extract_run(
+        data={
+            "missing_name": "Valid Name",
+            "age": "not_a_number",
+        },  # Invalid age, missing name
+        extraction_metadata={
+            "name": {"confidence": 0.9},
+        },
+        data_schema={},
+        file=create_file(id="error-file", name="bad_data.pdf"),
+        error=job_error_message,
+    )
+
+    # Should raise InvalidExtractionData with the job error prominently displayed
+    with pytest.raises(InvalidExtractionData) as exc_info:
+        ExtractedData.from_extraction_result(
+            extract_run, Person, metadata={"test": "metadata"}
+        )
+
+    # Verify the exception message prominently shows the job error
+    exception = exc_info.value
+    assert exception.extraction_error == job_error_message
+    assert f"Extraction error: {job_error_message}" == str(exception)
+
+    # Verify the invalid_item contains both errors in metadata
+    invalid_data = exception.invalid_item
+    assert invalid_data.metadata.get("job_error") == job_error_message
+    assert "extraction_error" in invalid_data.metadata  # Validation error still present
+    assert "test" in invalid_data.metadata  # Original metadata preserved
+

 class Dimensions(BaseModel):
    length: Optional[str] = Field(
@@ -663,3 +707,69 @@ def test_field_conflict_in_schema():
    assert isinstance(
        extracted["majority_opinion"]["reasoning"], ExtractedFieldMetadata
    )
+
+
+def test_parse_extracted_field_metadata_with_bounding_boxes():
+    """Test parse_extracted_field_metadata with bounding boxes and page dimensions."""
+    raw_metadata = {
+        "document_type": {
+            "citation": [
+                {
+                    "page": 1,
+                    "matching_text": "FACTURE ORIGINALE",
+                    "bounding_boxes": [{"x": 77.28, "y": 615.12, "w": 70.6, "h": 7.2}],
+                    "page_dimensions": {"width": 222.24, "height": 736.56},
+                }
+            ],
+            "parsing_confidence": 1.0,
+            "extraction_confidence": 0.7252506422636493,
+            "confidence": 0.7252506422636493,
+        },
+        "summary": {
+            "citation": [
+                {
+                    "page": 1,
+                    "matching_text": "FACTURE ORIGINALE",
+                    "bounding_boxes": [{"x": 77.28, "y": 615.12, "w": 70.6, "h": 7.2}],
+                    "page_dimensions": {"width": 222.24, "height": 736.56},
+                },
+                {
+                    "page": 1,
+                    "matching_text": "Café filtre assiette — $1.90",
+                    "bounding_boxes": [
+                        {"x": 10.56, "y": 172.83, "w": 171.85, "h": 497.01}
+                    ],
+                    "page_dimensions": {"width": 222.24, "height": 736.56},
+                },
+            ],
+            "parsing_confidence": 1.0,
+            "extraction_confidence": 0.5700013128334419,
+            "confidence": 0.5700013128334419,
+        },
+    }
+
+    result = parse_extracted_field_metadata(raw_metadata)
+
+    # Verify document_type citation with bounding boxes
+    assert isinstance(result["document_type"], ExtractedFieldMetadata)
+    assert result["document_type"].parsing_confidence == 1.0
+    assert result["document_type"].extraction_confidence == 0.7252506422636493
+    assert result["document_type"].confidence == 0.7252506422636493
+    assert len(result["document_type"].citation) == 1
+
+    citation = result["document_type"].citation[0]
+    assert citation.page == 1
+    assert citation.matching_text == "FACTURE ORIGINALE"
+    assert len(citation.bounding_boxes) == 1
+    assert citation.bounding_boxes[0] == BoundingBox(x=77.28, y=615.12, w=70.6, h=7.2)
+    assert citation.page_dimensions == PageDimensions(width=222.24, height=736.56)
+
+    # Verify summary citation with multiple bounding boxes
+    assert isinstance(result["summary"], ExtractedFieldMetadata)
+    assert len(result["summary"].citation) == 2
+    assert result["summary"].citation[0].bounding_boxes[0].x == 77.28
+    assert result["summary"].citation[1].bounding_boxes[0].x == 10.56
+
+    # Verify round-trip serialization
+    result2 = parse_extracted_field_metadata(result)
+    assert result2 == result
@@ -34,9 +34,10 @@ TEST_PIPELINE = Pipeline(
 def mock_client() -> MagicMock:
    """Mock client with sensible defaults."""
    client = MagicMock()
-    client.projects.upsert_project.return_value = Project(
+    default_project = Project(
        id="default-proj", name=DEFAULT_PROJECT_NAME, organization_id="default-org"
    )
+    client.projects.list_projects.return_value = [default_project]
    client.pipelines.upsert_pipeline.return_value = Pipeline(
        id="default-pipe",
        name="default",
@@ -100,8 +101,8 @@ def test_from_documents_uses_provided_project_id(mock_client: MagicMock) -> None
            project_id=provided_project_id,
        )

-    # Assert - project upsert not called; pipeline uses provided project_id
-    mock_client.projects.upsert_project.assert_not_called()
+    # Assert - project list not called (project_id provided); pipeline uses provided project_id
+    mock_client.projects.list_projects.assert_not_called()
    assert mock_client.pipelines.upsert_pipeline.call_count == 1
    assert (
        mock_client.pipelines.upsert_pipeline.call_args.kwargs["project_id"]
@@ -110,29 +111,29 @@ def test_from_documents_uses_provided_project_id(mock_client: MagicMock) -> None
    assert index.project.id == provided_project_id


-def test_from_documents_upserts_project_when_project_id_missing(
+def test_from_documents_lists_project_when_project_id_missing(
    mock_client: MagicMock,
 ) -> None:
    organization_id = "org-xyz"
    index_name = "my_new_index"

-    # Project is created when project_id is not provided
-    upserted_project = Project(
+    # Project is found by name when project_id is not provided
+    found_project = Project(
        id="proj-999", name=DEFAULT_PROJECT_NAME, organization_id=organization_id
    )
-    mock_client.projects.upsert_project.return_value = upserted_project
+    mock_client.projects.list_projects.return_value = [found_project]

    test_pipeline = Pipeline(
        id="pipe-xyz",
        name=index_name,
-        project_id=upserted_project.id,
+        project_id=found_project.id,
        embedding_config=EMBEDDING_CONFIG,
    )

    with patch.object(
        base,
        "resolve_project_and_pipeline",
-        return_value=(upserted_project, test_pipeline),
+        return_value=(found_project, test_pipeline),
    ):
        docs = [Document(text="world")]
        index = LlamaCloudIndex.from_documents(
@@ -141,15 +142,15 @@ def test_from_documents_upserts_project_when_project_id_missing(
            organization_id=organization_id,
        )

-    # Assert - project was upserted with org id and default project name
-    mock_client.projects.upsert_project.assert_called_once()
-    kwargs = mock_client.projects.upsert_project.call_args.kwargs
+    # Assert - project was listed with org id and default project name
+    mock_client.projects.list_projects.assert_called_once()
+    kwargs = mock_client.projects.list_projects.call_args.kwargs
    assert kwargs["organization_id"] == organization_id
-    assert kwargs["request"].name == DEFAULT_PROJECT_NAME
+    assert kwargs["project_name"] == DEFAULT_PROJECT_NAME

-    # Pipeline created under the upserted project id
+    # Pipeline created under the found project id
    assert (
        mock_client.pipelines.upsert_pipeline.call_args.kwargs["project_id"]
-        == upserted_project.id
+        == found_project.id
    )
-    assert index.project.id == upserted_project.id
+    assert index.project.id == found_project.id
@@ -1,5 +1,5 @@
 version = 1
-revision = 2
+revision = 3
 requires-python = ">=3.9, <4.0"
 resolution-markers = [
    "python_full_version >= '3.14'",
@@ -1595,21 +1595,21 @@ wheels = [

 [[package]]
 name = "llama-cloud"
-version = "0.1.45"
+version = "0.1.46"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
    { name = "certifi" },
    { name = "httpx" },
    { name = "pydantic" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/e0/b7/3a2a209f1c3fa516de172cb13e03f5a897adea5523f2ee0f544d035e3704/llama_cloud-0.1.45.tar.gz", hash = "sha256:140244008cc5710e31ae97c6043973a3a9969a51b0f38155fa33a8434078e8aa", size = 140968, upload-time = "2025-12-03T02:22:49.484Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/40/f3/f4d6520f8d546e6c5a02f6ebeed5c09774a074b8d2c24ad559ace97a56a6/llama_cloud-0.1.46.tar.gz", hash = "sha256:e86f8791c053590d70cc59e0fc13ce72f9b681a8e658bc61df86d0285288d8ee", size = 127752, upload-time = "2026-01-21T18:40:57.103Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/62/1d/466b0df69b81ce9410ad6ec7229a1e6601ff69640f02f246e06cfcc7428c/llama_cloud-0.1.45-py3-none-any.whl", hash = "sha256:500299a6d3f25f97bcf6755d6338523023564fa8f376955c2cf299bbc9561cc2", size = 397184, upload-time = "2025-12-03T02:22:48.335Z" },
+    { url = "https://files.pythonhosted.org/packages/c4/3a/6caaea28c8c804add33c91d356ed7d5a5412d6c9598e1450af95a15e0bcd/llama_cloud-0.1.46-py3-none-any.whl", hash = "sha256:6c6546c09c04a038c86d84d42f00eae8fd3bff49991ad3aab844bd866ecdf352", size = 361989, upload-time = "2026-01-21T18:40:54.863Z" },
 ]

 [[package]]
 name = "llama-cloud-services"
-version = "0.6.85"
+version = "0.6.90"
 source = { editable = "." }
 dependencies = [
    { name = "click", version = "8.1.8", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.10'" },
@@ -1649,7 +1649,7 @@ dev = [
 requires-dist = [
    { name = "click", specifier = ">=8.1.7,<9" },
    { name = "eval-type-backport", marker = "python_full_version < '3.10'", specifier = ">=0.2.0,<0.3" },
-    { name = "llama-cloud", specifier = "==0.1.45" },
+    { name = "llama-cloud", specifier = "==0.1.46" },
    { name = "llama-index-core", specifier = ">=0.12.0" },
    { name = "packaging", specifier = ">=23.0" },
    { name = "platformdirs", specifier = ">=4.3.7,<5" },
@@ -1,5 +1,38 @@
 # llama-cloud-services

+## 0.5.3
+
+### Patch Changes
+
+- d7864af: bugfixes in retry logic for LlamaExtract and LlamaClassify
+
+## 0.5.2
+
+### Patch Changes
+
+- 997bcc8: Add types for bounding boxes
+
+## 0.5.1
+
+### Patch Changes
+
+- d5b18a0: Fix publishing
+
+## 0.5.0
+
+### Minor Changes
+
+- 576c3d9: feat: support zod v4 & v3
+
+  Adds support for zod v4 while maintaining backward compatibility with v3.
+  - Updated zod peer dependency to accept both v3 and v4: `^3.25.76 || ^4.0.0`
+  - Migrated all import statements to use `zod/v4` import path for compatibility
+
+### Patch Changes
+
+- c8321d2: Improve parse results polling
+- 576c3d9: Support zod v3 an v4
+
 ## 0.4.3

 ### Patch Changes
@@ -1,12 +1,12 @@
 {
  "name": "llama-cloud-services",
-  "version": "0.4.3",
+  "version": "0.5.3",
  "type": "module",
  "license": "MIT",
  "scripts": {
    "get-openapi": "node ./scripts/get-openapi.js",
    "generate": "./node_modules/.bin/openapi-ts",
-    "build": "pnpm run generate && bunchee",
+    "build": "bunchee",
    "dev": "bunchee --watch",
    "lint": "eslint src/ --ignore-pattern client/*.ts --no-warn-ignored",
    "format": "prettier --write ./src/ tests/",
@@ -116,9 +116,9 @@
    "@eslint/js": "^9.32.0",
    "@hey-api/client-fetch": "^0.10.1",
    "@hey-api/openapi-ts": "^0.67.5",
-    "@llamaindex/core": "^0.6.19",
+    "@llamaindex/core": "^0.6.22",
    "@llamaindex/env": "^0.1.30",
-    "@llamaindex/workflow-core": "^0.4.1",
+    "@llamaindex/workflow-core": "^1.3.3",
    "@types/node": "^20.19.9",
    "@typescript-eslint/eslint-plugin": "^8.38.0",
    "@typescript-eslint/parser": "^8.38.0",
@@ -131,18 +131,19 @@
    "turbo": "^2.5.5",
    "typescript": "^5.8.3",
    "typescript-eslint": "^8.38.0",
-    "vitest": "^2.0.0"
+    "vitest": "^2.0.0",
+    "zod": "^4.1.13"
  },
  "peerDependencies": {
    "@llamaindex/core": "^0.6.19",
    "@llamaindex/env": "^0.1.30",
-    "@llamaindex/workflow-core": "^0.4.1"
+    "@llamaindex/workflow-core": "^1.3.3",
+    "zod": "^3.25.0 || ^4.0.0"
  },
  "dependencies": {
    "ajv": "^8.17.1",
    "file-type": "^21.0.0",
-    "p-retry": "^6.2.1",
-    "zod": "^3.25.76"
+    "p-retry": "^6.2.1"
  },
  "packageManager": "pnpm@10.8.1"
 }
@@ -2,11 +2,14 @@ export { AgentClient, createAgentDataClient } from "./client";

 export type {
  AggregateAgentDataOptions,
+  BoundingBox,
  ComparisonOperator,
  ExtractedData,
  ExtractedFieldMetadata,
  ExtractedFieldMetadataDict,
+  FieldCitation,
  FilterOperation,
+  PageDimensions,
  SearchAgentDataOptions,
  StatusType,
  TypedAgentData,
@@ -28,6 +28,44 @@ export type ComparisonOperator =
 */
 export type FilterOperation = RawFilterOperation;

+/**
+ * Bounding box coordinates for a citation location on a page
+ */
+export interface BoundingBox {
+  /** X coordinate of the bounding box origin */
+  x: number;
+  /** Y coordinate of the bounding box origin */
+  y: number;
+  /** Width of the bounding box */
+  w: number;
+  /** Height of the bounding box */
+  h: number;
+}
+
+/**
+ * Dimensions of a page in the source document
+ */
+export interface PageDimensions {
+  /** Width of the page */
+  width: number;
+  /** Height of the page */
+  height: number;
+}
+
+/**
+ * Citation information for an extracted field
+ */
+export interface FieldCitation {
+  /** The page number that the field occurred on */
+  page?: number;
+  /** The original text this field's value was derived from */
+  matching_text?: string;
+  /** Bounding boxes indicating where the citation appears on the page */
+  bounding_boxes?: BoundingBox[];
+  /** Dimensions of the page containing the citation */
+  page_dimensions?: PageDimensions;
+}
+
 /**
 * Metadata for an extracted field, including confidence and citation information
 */
@@ -38,16 +76,11 @@ export interface ExtractedFieldMetadata {
  confidence?: number;
  /** The confidence score for the field based on the extracted text only */
  extraction_confidence?: number;
+  /** The confidence score for the field based on the parsing/OCR quality */
+  parsing_confidence?: number;
  citation?: FieldCitation[];
 }

-export interface FieldCitation {
-  /** The page number that the field occurred on */
-  page?: number;
-  /** The original text this field's value was derived from */
-  matching_text?: string;
-}
-
 /**
 * Dictionary mapping field names to their metadata
 * Values can be ExtractedFieldMetadata objects, nested dictionaries, or arrays
@@ -108,20 +108,19 @@ async function pollForJobCompletion({
    }
    const response =
      await getClassifyJobApiV1ClassifierJobsClassifyJobIdGet(jobOptions);
-    if (!response.response.ok) {
-      numIterations++;
-    }
    if (typeof response.data != "undefined") {
      status = response.data.status as StatusEnum;
      if (status == StatusEnum.CANCELLED || status == StatusEnum.ERROR) {
-        throw new Error("There was an error during the classification job.");
-      } else if (status == StatusEnum.SUCCESS) {
+        throw new Error("There was an error extracting data from your file.");
+      } else if (
+        status == StatusEnum.SUCCESS ||
+        status == StatusEnum.PARTIAL_SUCCESS
+      ) {
        return true;
-      } else {
-        numIterations++;
-        await sleep(interval * 1000);
      }
    }
+    numIterations++;
+    await sleep(interval * 1000);
  }
 }

@@ -169,7 +168,7 @@ async function getJobResult({
      retries++;
      await sleep(retryInterval * 1000);
    }
-    if (typeof response.data != "undefined") {
+    if (response.response.ok && typeof response.data != "undefined") {
      return response.data as ClassifyJobResults;
    } else {
      throw new Error(
@@ -1,6 +1,6 @@
 // This file is auto-generated by @hey-api/openapi-ts

-import { z } from "zod";
+import { z } from "zod/v4";

 export const zNoneSegmentationConfig = z.object({
  mode: z.literal("none").optional().default("none"),
@@ -770,7 +770,7 @@ export const zMetadataFilter = z.object({

 export const zFilterCondition: z.ZodTypeAny = z.enum(["and", "or", "not"]);

-export const zMetadataFilters: z.AnyZodObject = z.object({
+export const zMetadataFilters: z.ZodObject<z.ZodRawShape> = z.object({
  filters: z.array(z.unknown()),
  condition: z.union([zFilterCondition, z.null()]).optional(),
 });
@@ -782,7 +782,7 @@ export const zRetrievalMode: z.ZodTypeAny = z.enum([
  "auto_routed",
 ]);

-export const zPresetRetrievalParams: z.AnyZodObject = z.object({
+export const zPresetRetrievalParams: z.ZodObject<z.ZodRawShape> = z.object({
  dense_similarity_top_k: z
    .union([z.number().int().gte(1).lte(100), z.null()])
    .optional(),
@@ -825,7 +825,7 @@ export const zSupportedLlmModelNames: z.ZodTypeAny = z.enum([
  "VERTEX_AI_CLAUDE_3_5_SONNET_V2",
 ]);

-export const zLlmParameters: z.AnyZodObject = z.object({
+export const zLlmParameters: z.ZodObject<z.ZodRawShape> = z.object({
  model_name: zSupportedLlmModelNames.optional(),
  system_prompt: z.union([z.string().max(3000), z.null()]).optional(),
  temperature: z.union([z.number(), z.null()]).optional(),
@@ -834,7 +834,7 @@ export const zLlmParameters: z.AnyZodObject = z.object({
  class_name: z.string().optional().default("base_component"),
 });

-export const zChatData: z.AnyZodObject = z.object({
+export const zChatData: z.ZodObject<z.ZodRawShape> = z.object({
  retrieval_parameters: zPresetRetrievalParams.optional(),
  llm_parameters: z.union([zLlmParameters, z.null()]).optional(),
  class_name: z.string().optional().default("base_component"),
@@ -2141,7 +2141,7 @@ export const zTextItem = z.object({
  value: z.string(),
 });

-export const zListItem: z.AnyZodObject = z.object({
+export const zListItem: z.ZodObject<z.ZodRawShape> = z.object({
  type: z.literal("list").optional().default("list"),
  bBox: z.union([z.unknown(), z.null()]).optional(),
  items: z.array(z.unknown()),
@@ -1,6 +1,6 @@
 import { workflowEvent } from "@llamaindex/workflow-core";
 import { zodEvent } from "@llamaindex/workflow-core/util/zod";
-import { z } from "zod";
+import { z } from "zod/v4";
 import { parseFormSchema } from "./schema";

 export const uploadEvent = zodEvent(
@@ -296,20 +296,16 @@ async function pollForJobCompletion(
      return false;
    }
    const response = await getJobApiV1ExtractionJobsJobIdGet(jobOptions);
-    if (!response.response.ok) {
-      numIterations++;
-    }
    if (typeof response.data != "undefined") {
      status = response.data.status as StatusEnum;
      if (status == StatusEnum.CANCELLED || status == StatusEnum.ERROR) {
        throw new Error("There was an error extracting data from your file.");
      } else if (status == StatusEnum.SUCCESS) {
        return true;
-      } else {
-        numIterations++;
-        await sleep(interval * 1000);
      }
    }
+    numIterations++;
+    await sleep(interval * 1000);
  }
 }

@@ -350,7 +346,7 @@ async function getJobResult(
      retries++;
      await sleep(retryInterval * 1000);
    }
-    if (typeof response.data != "undefined") {
+    if (response.response.ok && typeof response.data != "undefined") {
      return {
        data: response.data.data,
        extractionMetadata: response.data.extraction_metadata,
@@ -2,7 +2,7 @@ import type { JSONValue } from "@llamaindex/core/global";
 import type { ToolMetadata } from "@llamaindex/core/llms";
 import type { BaseQueryEngine } from "@llamaindex/core/query-engine";
 import { tool } from "@llamaindex/core/tools";
-import { z } from "zod";
+import { z } from "zod/v4";

 const DEFAULT_NAME = "llama_cloud_index_tool";
 const DEFAULT_DESCRIPTION =
@@ -21,9 +21,7 @@ export function createQueryEngineTool(
    name: metadata?.name ?? DEFAULT_NAME,
    description: metadata?.description ?? DEFAULT_DESCRIPTION,
    parameters: z.object({
-      query: z.string({
-        description: "The query to search for",
-      }),
+      query: z.string().describe("The query to search for"),
    }),
    execute: async ({ query }) => {
      const response = await queryEngine.query({ query });
@@ -1,9 +1,10 @@
 /* eslint-disable @typescript-eslint/no-explicit-any */
 import { type Client, createClient, createConfig } from "@hey-api/client-fetch";
+import { type FailedAttemptError } from "p-retry";
 import { Document, FileReader } from "@llamaindex/core/schema";
 import { fs, getEnv, path } from "@llamaindex/env";
 import {
-  type BodyUploadFileApiParsingUploadPost,
+  type BodyUploadFileApiV1ParsingUploadPost,
  type FailPageMode,
  type ParserLanguages,
  type ParsingMode,
@@ -32,6 +33,33 @@ type WriteStream = {
 // eslint-disable-next-line no-var
 var process: any;

+function handleFailedAttempt(
+  error: FailedAttemptError,
+  jobId: string,
+  verbose: boolean,
+) {
+  // Retry only on 5XX or socket errors.
+  const status = (error.cause as any)?.response?.status;
+  if (
+    !(
+      (status && status >= 500 && status < 600) ||
+      ((error.cause as any)?.code &&
+        ((error.cause as any).code === "ECONNRESET" ||
+          (error.cause as any).code === "ETIMEDOUT" ||
+          (error.cause as any).code === "ECONNREFUSED")) ||
+      (status && status === 404)
+    )
+  ) {
+    throw error;
+  }
+
+  if (verbose) {
+    console.warn(
+      `Attempting to get job ${jobId} result (attempt ${error.attemptNumber}) failed. Retrying...`,
+    );
+  }
+}
+
 /**
 * Represents a reader for parsing files using the LlamaParse API.
 * See https://github.com/run-llama/llama_parse
@@ -188,6 +216,16 @@ export class LlamaParseReader extends FileReader {
  extract_printed_page_number?: boolean | undefined;
  tier?: string | undefined;
  version?: string | undefined;
+  layout_aware?: boolean | undefined;
+  line_level_bounding_box?: boolean | undefined;
+  specialized_image_parsing?: boolean | undefined;
+  aggressive_table_extraction?: boolean | undefined;
+  preserve_very_small_text?: boolean | undefined;
+  spreadsheet_force_formula_computation?: boolean | undefined;
+  inline_images_in_markdown?: boolean | undefined;
+  keep_page_separator_when_merging_tables?: boolean | undefined;
+  remove_hidden_text?: boolean | undefined;
+  presentation_out_of_bounds_content?: boolean | undefined;

  constructor(
    params: Partial<Omit<LlamaParseReader, "language" | "apiKey">> & {
@@ -387,11 +425,25 @@ export class LlamaParseReader extends FileReader {
      extract_printed_page_number: this.extract_printed_page_number,
      tier: this.tier,
      version: this.version,
+      layout_aware: this.layout_aware,
+      line_level_bounding_box: this.line_level_bounding_box,
+      specialized_image_parsing: this.specialized_image_parsing,
+      aggressive_table_extraction: this.aggressive_table_extraction,
+      preserve_very_small_text: this.preserve_very_small_text,
+      spreadsheet_force_formula_computation:
+        this.spreadsheet_force_formula_computation,
+      inline_images_in_markdown: this.inline_images_in_markdown,
+      webhook_configurations: undefined,
+      keep_page_separator_when_merging_tables:
+        this.keep_page_separator_when_merging_tables,
+      remove_hidden_text: this.remove_hidden_text,
+      presentation_out_of_bounds_content:
+        this.presentation_out_of_bounds_content,
    } satisfies {
-      [Key in keyof BodyUploadFileApiParsingUploadPost]-?:
-        | BodyUploadFileApiParsingUploadPost[Key]
+      [Key in keyof BodyUploadFileApiV1ParsingUploadPost]-?:
+        | BodyUploadFileApiV1ParsingUploadPost[Key]
        | undefined;
-    } as unknown as BodyUploadFileApiParsingUploadPost;
+    } as unknown as BodyUploadFileApiV1ParsingUploadPost;

    const response = await uploadFileApiV1ParsingUploadPost({
      client: this.#client,
@@ -443,26 +495,8 @@ export class LlamaParseReader extends FileReader {
            }),
          {
            retries: this.maxErrorCount,
-            onFailedAttempt: (error) => {
-              // Retry only on 5XX or socket errors.
-              const status = (error.cause as any)?.response?.status;
-              if (
-                !(
-                  (status && status >= 500 && status < 600) ||
-                  ((error.cause as any)?.code &&
-                    ((error.cause as any).code === "ECONNRESET" ||
-                      (error.cause as any).code === "ETIMEDOUT" ||
-                      (error.cause as any).code === "ECONNREFUSED"))
-                )
-              ) {
-                throw error;
-              }
-              if (this.verbose) {
-                console.warn(
-                  `Attempting to get job ${jobId} result (attempt ${error.attemptNumber}) failed. Retrying...`,
-                );
-              }
-            },
+            onFailedAttempt: (error) =>
+              handleFailedAttempt(error, jobId, this.verbose),
          },
        );
      } catch (e: any) {
@@ -475,49 +509,69 @@ export class LlamaParseReader extends FileReader {
      const status = (data as Record<string, unknown>)["status"];

      if (status === "SUCCESS") {
-        let resultData;
        switch (resultType) {
          case "json": {
-            resultData =
-              await getJobJsonResultApiV1ParsingJobJobIdResultJsonGet({
-                client: this.#client,
-                throwOnError: true,
-                path: { job_id: jobId },
-                query: {
-                  organization_id: this.organization_id ?? null,
-                },
-                signal: AbortSignal.timeout(this.maxTimeout * 1000),
-              });
-            break;
+            const resultData = await pRetry(
+              () =>
+                getJobJsonResultApiV1ParsingJobJobIdResultJsonGet({
+                  client: this.#client,
+                  throwOnError: true,
+                  path: { job_id: jobId },
+                  query: {
+                    organization_id: this.organization_id ?? null,
+                  },
+                  signal: AbortSignal.timeout(this.maxTimeout * 1000),
+                }),
+              {
+                retries: this.maxErrorCount,
+                onFailedAttempt: (error) =>
+                  handleFailedAttempt(error, jobId, this.verbose),
+              },
+            );
+            return resultData.data;
          }
          case "markdown": {
-            resultData =
-              await getJobResultApiV1ParsingJobJobIdResultMarkdownGet({
-                client: this.#client,
-                throwOnError: true,
-                path: { job_id: jobId },
-                query: {
-                  organization_id: this.organization_id ?? null,
-                },
-                signal: AbortSignal.timeout(this.maxTimeout * 1000),
-              });
-            break;
+            const resultData = await pRetry(
+              () =>
+                getJobResultApiV1ParsingJobJobIdResultMarkdownGet({
+                  client: this.#client,
+                  throwOnError: true,
+                  path: { job_id: jobId },
+                  query: {
+                    organization_id: this.organization_id ?? null,
+                  },
+                  signal: AbortSignal.timeout(this.maxTimeout * 1000),
+                }),
+              {
+                retries: this.maxErrorCount,
+                onFailedAttempt: (error) =>
+                  handleFailedAttempt(error, jobId, this.verbose),
+              },
+            );
+            return resultData.data;
          }
          case "text": {
-            resultData =
-              await getJobTextResultApiV1ParsingJobJobIdResultTextGet({
-                client: this.#client,
-                throwOnError: true,
-                path: { job_id: jobId },
-                query: {
-                  organization_id: this.organization_id ?? null,
-                },
-                signal: AbortSignal.timeout(this.maxTimeout * 1000),
-              });
-            break;
+            const resultData = await pRetry(
+              () =>
+                getJobTextResultApiV1ParsingJobJobIdResultTextGet({
+                  client: this.#client,
+                  throwOnError: true,
+                  path: { job_id: jobId },
+                  query: {
+                    organization_id: this.organization_id ?? null,
+                  },
+                  signal: AbortSignal.timeout(this.maxTimeout * 1000),
+                }),
+              {
+                retries: this.maxErrorCount,
+                onFailedAttempt: (error) =>
+                  handleFailedAttempt(error, jobId, this.verbose),
+              },
+            );
+
+            return resultData.data;
          }
        }
-        return resultData.data;
      } else if (status === "PENDING") {
        if (this.verbose && tries % 10 === 0) {
          this.stdout?.write(".");
@@ -1,6 +1,6 @@
 import { FailPageMode, ParserLanguages, ParsingMode } from "./client";

-import { z } from "zod";
+import { z } from "zod/v4";

 type Language = ParserLanguages;
 const VALUES: [Language, ...Language[]] = [
@@ -52,9 +52,10 @@ export const parseFormSchema = z.object({
  html_remove_navigation_elements: z.boolean().optional(),
  http_proxy: z
    .string()
-    .url(
-      'Set a valid URL for the HTTP proxy, e.g., "http://proxy.example.com:8080"',
-    )
+    .url({
+      error:
+        'Set a valid URL for the HTTP proxy, e.g., "http://proxy.example.com:8080"',
+    })
    .refine(
      (url) => {
        try {
@@ -67,7 +68,7 @@ export const parseFormSchema = z.object({
        }
      },
      {
-        message: "Invalid HTTP proxy URL",
+        error: "Invalid HTTP proxy URL",
      },
    )
    .optional(),
@@ -100,7 +101,7 @@ export const parseFormSchema = z.object({
  vendor_multimodal_model_name: z.string().optional(),
  model: z.string().optional(),
  webhook_url: z.string().url().optional(),
-  parse_mode: z.nativeEnum(ParsingMode).nullable().optional(),
+  parse_mode: z.enum(ParsingMode).nullable().optional(),
  system_prompt: z.string().optional(),
  system_prompt_append: z.string().optional(),
  user_prompt: z.string().optional(),
@@ -129,7 +130,7 @@ export const parseFormSchema = z.object({
  compact_markdown_table: z.boolean().optional(),
  markdown_table_multiline_header_separator: z.string().optional(),
  page_error_tolerance: z.number().min(0).max(1).optional(),
-  replace_failed_page_mode: z.nativeEnum(FailPageMode).nullable().optional(),
+  replace_failed_page_mode: z.enum(FailPageMode).nullable().optional(),
  replace_failed_page_with_error_message_prefix: z.string().optional(),
  replace_failed_page_with_error_message_suffix: z.string().optional(),
 });
Author	SHA1	Message	Date
Neeraj Pradhan	39a93d602b	fix parse model	2026-01-27 16:05:23 -08:00
Neeraj Pradhan	0c04955c10	Use sonnet when testing premium mode in extract e2e	2026-01-27 15:48:14 -08:00
Neeraj Pradhan	38da9a52d7	Invalidate cache when running extract tests (#1097 )	2026-01-26 17:33:23 -08:00
Neeraj Pradhan	1e7ec40ee7	Fix verbose logging on slack channel (#1096 )	2026-01-26 17:12:50 -08:00
Neeraj Pradhan	dd83c1a9d0	Add retries to all extract sdk functions uniformly (#1095 )	2026-01-26 12:05:16 -08:00
Neeraj Pradhan	7cb83f5cd3	Change cron schedule for hourly extract tests (#1094 )	2026-01-26 10:15:34 -08:00
Neeraj Pradhan	b05266be6d	Try to reparse scheduled workflow (#1093 )	2026-01-26 09:56:22 -08:00
Neeraj Pradhan	eab4798165	Force github reparse of the workflow (#1090 )	2026-01-23 11:36:28 -08:00
Neeraj Pradhan	b174fa8fab	Run hourly extract tests to catch SDK schema drifts (#1089 ) * Run hourly extract tests to catch SDK schema drifts * fix url * fix prod/staging env	2026-01-22 18:18:45 -08:00
github-actions[bot]	b12ffef916	chore: version packages (#1087 ) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>	2026-01-21 12:44:43 -08:00
Neeraj Pradhan	07ec282257	Bump up patch version for python packages (#1086 )	2026-01-21 12:30:23 -08:00
Neeraj Pradhan	013b689812	Bump up minor version for python packages (#1085 )	2026-01-21 12:13:13 -08:00
Adrian Lyjak	3040951cb8	Use error description in invalid extraction error (#1081 ) * fix: display extraction job error in InvalidExtractionData exception Refactored InvalidExtractionData to read the `error` field from ExtractRun and prominently display it in the exception message. The job-level error is now stored in the `extraction_error` attribute and included in the invalid_item's metadata as `job_error`. * Create three-yaks-beg.md --------- Co-authored-by: Claude <noreply@anthropic.com>	2026-01-18 17:43:21 -05:00
github-actions[bot]	9239498945	chore: version packages (#1076 ) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>	2026-01-14 19:15:05 +01:00
Pierre-Loic Doulcet	19cbb25631	remove extension filter (#1075 ) * remove extension filter * changeset * Update ninety-goats-look.md Make it a patch version * Update package.json back out of version bump * Update pyproject.toml back out of version bump * Update package.json back out of version bump * Update pyproject.toml back out of version bump --------- Co-authored-by: Adrian Lyjak <adrianlyjak@gmail.com>	2026-01-14 19:13:39 +01:00
github-actions[bot]	812e2f7d72	chore: version packages (#1073 ) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>	2026-01-12 19:03:13 +01:00
Clelia (Astra) Bertelli	d7864afe3f	fix: bug fix retry logic in Classify and Extract (#1066 ) * fix: bug fix retry logic in Classify and Extract * chore: apply suggestion * chore: add PARTIAL_SUCCESS to classify	2026-01-12 18:57:40 +01:00
github-actions[bot]	ade8d027a5	chore: version packages (#1071 ) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>	2026-01-09 20:29:00 -05:00
Adrian Lyjak	997bcc8531	forgot ts changeset (#1070 )	2026-01-09 20:23:29 -05:00
github-actions[bot]	8be554c234	chore: version packages (#1068 ) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>	2026-01-09 18:56:51 -05:00
Adrian Lyjak	f777cab0c5	Add bounding box type support to TS too (#1069 ) ts too	2026-01-09 18:55:16 -05:00
Adrian Lyjak	b9b83c953d	Parse bounding boxes from extract jobs results in agent data (#1067 )	2026-01-09 18:47:57 -05:00
github-actions[bot]	3ec7024626	chore: version packages (#1058 )	2025-12-10 11:53:30 -06:00
Logan	d5b18a03fa	Remove `generate` from build path to fix publishing (#1057 )	2025-12-10 11:52:43 -06:00
Clelia (Astra) Bertelli	18dd04b6de	docs: correct links in readme (#1056 )	2025-12-10 17:08:58 +01:00
github-actions[bot]	685a5e6ccc	chore: version packages (#1054 )	2025-12-09 15:30:13 -06:00
Jim Geurts	576c3d9076	feat: support zod v4 & v3 (#1052 )	2025-12-09 15:29:23 -06:00
Logan	c8321d2bc5	improve parse ts polling (#1053 )	2025-12-09 15:21:19 -06:00
Tuana Çelik	131bbed7aa	batch parse sctript with asyncio (#1051 ) * batch parse sctript with asyncio * lint --------- Co-authored-by: Logan Markewich <logan.markewich@live.com>	2025-12-08 18:50:11 +01:00
Javier Torres	41c8ac2348	docs: Split Example Notebook (#1044 ) * split notebook * Lint	2025-12-08 13:57:20 +01:00