Add PageIndexClient with agent-based retrieval via OpenAI Agents SDK

README.md

@@ -147,12 +147,14 @@ You can follow these steps to generate a PageIndex tree from a PDF document.
 pip3 install --upgrade -r requirements.txt
 ```
 
-### 2. Set your OpenAI API key
+### 2. Set your API key
 
 Create a `.env` file in the root directory and add your API key:
 
 ```bash
-CHATGPT_API_KEY=your_openai_key_here
+OPENAI_API_KEY=your_openai_key_here
+# or
+CHATGPT_API_KEY=your_openai_key_here  # legacy, still supported
 ```
 
 ### 3. Run PageIndex on your PDF
@@ -189,7 +191,41 @@ python3 run_pageindex.py --md_path /path/to/your/document.md
 > Note: in this function, we use "#" to determine node heading and their levels. For example, "##" is level 2, "###" is level 3, etc. Make sure your markdown file is formatted correctly. If your Markdown file was converted from a PDF or HTML, we don't recommend using this function, since most existing conversion tools cannot preserve the original hierarchy. Instead, use our [PageIndex OCR](https://pageindex.ai/blog/ocr), which is designed to preserve the original hierarchy, to convert the PDF to a markdown file and then use this function.
 </details>
 
-<!-- 
+---
+
+# 🐍 Python API
+
+### Index & Retrieve
+
+```python
+from pageindex import PageIndexClient
+
+client = PageIndexClient(workspace="~/.pageindex")
+
+# Index a document (PDF or Markdown)
+doc_id = client.index("path/to/document.pdf")
+
+# Retrieve
+client.get_document(doc_id)            # metadata: name, type, page count
+client.get_document_structure(doc_id)  # full tree structure
+client.get_page_content(doc_id, pages="5-7")  # page content
+```
+
+### Agent-based QA (OpenAI Agents)
+
+For a complete agent QA example using the [OpenAI Agents SDK](https://github.com/openai/openai-agents-python), see [`examples/openai_agents_demo.py`](examples/openai_agents_demo.py).
+
+```bash
+# Install optional dependency
+pip install openai-agents
+
+# Run the demo
+python examples/openai_agents_demo.py
+```
+
+---
+
+<!--
 # ☁️ Improved Tree Generation with PageIndex OCR
 
 This repo is designed for generating PageIndex tree structure for simple PDFs, but many real-world use cases involve complex PDFs that are hard to parse by classic Python tools. However, extracting high-quality text from PDF documents remains a non-trivial challenge. Most OCR tools only extract page-level content, losing the broader document context and hierarchy.

examples/openai_agents_demo.py

@@ -0,0 +1,165 @@
+"""
+PageIndex x OpenAI Agents Demo
+
+Demonstrates how to use PageIndexClient with the OpenAI Agents SDK
+to build a document QA agent with 3 tools:
+  - get_document()
+  - get_document_structure()
+  - get_page_content()
+
+Requirements:
+    pip install openai-agents
+
+Steps:
+  1 — Index PDF and inspect tree structure
+  2 — Inspect document metadata
+  3 — Ask a question (agent auto-calls tools)
+  4 — Reload from workspace and verify persistence
+"""
+import os
+import sys
+import asyncio
+import concurrent.futures
+import requests
+from pathlib import Path
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+from agents import Agent, Runner, function_tool
+from agents.stream_events import RawResponsesStreamEvent, RunItemStreamEvent
+from openai.types.responses import ResponseTextDeltaEvent
+
+from pageindex import PageIndexClient
+import pageindex.utils as utils
+
+PDF_URL = "https://arxiv.org/pdf/2501.12948.pdf"
+PDF_PATH = "tests/pdfs/deepseek-r1.pdf"
+WORKSPACE = "./pageindex_workspace"
+
+AGENT_SYSTEM_PROMPT = """
+You are PageIndex, a document QA assistant.
+TOOL USE:
+- Call get_document() first to confirm status and page/line count.
+- Call get_document_structure() to find relevant page ranges (use node summaries and start_index/end_index).
+- Call get_page_content(pages="5-7") with tight ranges. Never fetch the whole doc.
+- When calling tool call, output one short sentence explaining reason.
+ANSWERING: Answer based only on tool output. Be concise.
+"""
+
+
+def query_agent(
+    client: PageIndexClient,
+    doc_id: str,
+    prompt: str,
+    verbose: bool = False,
+) -> str:
+    """Run a document QA agent using the OpenAI Agents SDK.
+
+    Streams text output token-by-token and returns the full answer string.
+    Tool calls are always printed; verbose=True also prints arguments and output previews.
+    """
+
+    @function_tool
+    def get_document() -> str:
+        """Get document metadata: status, page count, name, and description."""
+        return client.get_document(doc_id)
+
+    @function_tool
+    def get_document_structure() -> str:
+        """Get the document's full tree structure (without text) to find relevant sections."""
+        return client.get_document_structure(doc_id)
+
+    @function_tool
+    def get_page_content(pages: str) -> str:
+        """
+        Get the text content of specific pages or line numbers.
+        Use tight ranges: e.g. '5-7' for pages 5 to 7, '3,8' for pages 3 and 8, '12' for page 12.
+        For Markdown documents, use line numbers from the structure's line_num field.
+        """
+        return client.get_page_content(doc_id, pages)
+
+    agent = Agent(
+        name="PageIndex",
+        instructions=AGENT_SYSTEM_PROMPT,
+        tools=[get_document, get_document_structure, get_page_content],
+        model=client.model,
+    )
+
+    async def _run():
+        turn = 0
+        collected = []
+        streamed_run = Runner.run_streamed(agent, prompt)
+        async for event in streamed_run.stream_events():
+            if isinstance(event, RawResponsesStreamEvent):
+                if isinstance(event.data, ResponseTextDeltaEvent):
+                    delta = event.data.delta
+                    print(delta, end="", flush=True)
+                    collected.append(delta)
+            elif isinstance(event, RunItemStreamEvent):
+                if event.name == "tool_called":
+                    if collected:
+                        print()
+                        collected.clear()
+                    raw = event.item.raw_item
+                    args = getattr(raw, "arguments", "{}")
+                    args_str = f"({args})" if verbose else "()"
+                    print(f"\nCalling tool: {raw.name}{args_str}")
+                elif event.name == "tool_output" and verbose:
+                    output = str(event.item.output)
+                    preview = output[:200] + "..." if len(output) > 200 else output
+                    print(f"Tool output: {preview}")
+        if collected:
+            print()
+        return "".join(collected)
+
+    try:
+        asyncio.get_running_loop()
+        with concurrent.futures.ThreadPoolExecutor(max_workers=1) as pool:
+            return pool.submit(asyncio.run, _run()).result()
+    except RuntimeError:
+        return asyncio.run(_run())
+
+
+# ── Download PDF if needed ─────────────────────────────────────────────────────
+if not os.path.exists(PDF_PATH):
+    print(f"Downloading {PDF_URL} ...")
+    os.makedirs(os.path.dirname(PDF_PATH), exist_ok=True)
+    with requests.get(PDF_URL, stream=True, timeout=30) as r:
+        r.raise_for_status()
+        with open(PDF_PATH, "wb") as f:
+            for chunk in r.iter_content(chunk_size=8192):
+                if chunk:
+                    f.write(chunk)
+    print("Download complete.\n")
+
+# ── Setup ──────────────────────────────────────────────────────────────────────
+client = PageIndexClient(workspace=WORKSPACE)
+
+# ── Step 1: Index + Tree ───────────────────────────────────────────────────────
+print("=" * 60)
+print("Step 1: Indexing PDF and inspecting tree structure")
+print("=" * 60)
+_id_cache = Path(WORKSPACE).expanduser() / "demo_doc_id.txt"
+if _id_cache.exists() and (doc_id := _id_cache.read_text().strip()) in client.documents:
+    print(f"\nLoaded cached doc_id: {doc_id}")
+else:
+    doc_id = client.index(PDF_PATH)
+    _id_cache.parent.mkdir(parents=True, exist_ok=True)
+    _id_cache.write_text(doc_id)
+    print(f"\nIndexed. doc_id: {doc_id}")
+print("\nTree Structure (top-level sections):")
+utils.print_tree(client.documents[doc_id]["structure"])
+
+# ── Step 2: Document Metadata ──────────────────────────────────────────────────
+print("\n" + "=" * 60)
+print("Step 2: Document Metadata (get_document)")
+print("=" * 60)
+print(client.get_document(doc_id))
+
+# ── Step 3: Agent Query ────────────────────────────────────────────────────────
+print("\n" + "=" * 60)
+print("Step 3: Agent Query (auto tool-use)")
+print("=" * 60)
+question = "What reward design does DeepSeek-R1-Zero use, and why was it chosen over supervised fine-tuning?"
+print(f"\nQuestion: '{question}'\n")
+query_agent(client, doc_id, question, verbose=True)

pageindex/__init__.py

@@ -1,2 +1,4 @@
 from .page_index import *
-from .page_index_md import md_to_tree
+from .page_index_md import md_to_tree
+from .retrieve import get_document, get_document_structure, get_page_content
+from .client import PageIndexClient

pageindex/client.py

@@ -0,0 +1,118 @@
+import os
+import uuid
+import json
+import asyncio
+from pathlib import Path
+
+from .page_index import page_index
+from .page_index_md import md_to_tree
+from .retrieve import get_document, get_document_structure, get_page_content
+
+
+class PageIndexClient:
+    """
+    A client for indexing and retrieving document content.
+    Flow: index() -> get_document() / get_document_structure() / get_page_content()
+
+    For agent-based QA, see examples/openai_agents_demo.py.
+    """
+    def __init__(self, api_key: str = None, model: str = "gpt-4o-2024-11-20", workspace: str = None):
+        if api_key:
+            os.environ["OPENAI_API_KEY"] = api_key
+        elif not os.getenv("OPENAI_API_KEY") and os.getenv("CHATGPT_API_KEY"):
+            os.environ["OPENAI_API_KEY"] = os.getenv("CHATGPT_API_KEY")
+        self.model = model
+        self.workspace = Path(workspace).expanduser() if workspace else None
+        if self.workspace:
+            self.workspace.mkdir(parents=True, exist_ok=True)
+        self.documents = {}
+        if self.workspace:
+            self._load_workspace()
+
+    def index(self, file_path: str, mode: str = "auto") -> str:
+        """Index a document. Returns a document_id."""
+        if not os.path.exists(file_path):
+            raise FileNotFoundError(f"File not found: {file_path}")
+
+        doc_id = str(uuid.uuid4())
+        ext = os.path.splitext(file_path)[1].lower()
+
+        is_pdf = ext == '.pdf'
+        is_md = ext in ['.md', '.markdown']
+
+        if mode == "pdf" or (mode == "auto" and is_pdf):
+            print(f"Indexing PDF: {file_path}")
+            result = page_index(
+                doc=file_path,
+                model=self.model,
+                if_add_node_summary='yes',
+                if_add_node_text='yes',
+                if_add_node_id='yes',
+                if_add_doc_description='yes'
+            )
+            self.documents[doc_id] = {
+                'id': doc_id,
+                'path': file_path,
+                'type': 'pdf',
+                'structure': result['structure'],
+                'doc_name': result.get('doc_name', ''),
+                'doc_description': result.get('doc_description', '')
+            }
+
+        elif mode == "md" or (mode == "auto" and is_md):
+            print(f"Indexing Markdown: {file_path}")
+            result = asyncio.run(md_to_tree(
+                md_path=file_path,
+                if_thinning=False,
+                if_add_node_summary='yes',
+                summary_token_threshold=200,
+                model=self.model,
+                if_add_doc_description='yes',
+                if_add_node_text='yes',
+                if_add_node_id='yes'
+            ))
+            self.documents[doc_id] = {
+                'id': doc_id,
+                'path': file_path,
+                'type': 'md',
+                'structure': result['structure'],
+                'doc_name': result.get('doc_name', ''),
+                'doc_description': result.get('doc_description', '')
+            }
+        else:
+            raise ValueError(f"Unsupported file format for: {file_path}")
+
+        print(f"Indexing complete. Document ID: {doc_id}")
+        if self.workspace:
+            self._save_doc(doc_id)
+        return doc_id
+
+    def _save_doc(self, doc_id: str):
+        path = self.workspace / f"{doc_id}.json"
+        with open(path, "w", encoding="utf-8") as f:
+            json.dump(self.documents[doc_id], f, ensure_ascii=False, indent=2)
+
+    def _load_workspace(self):
+        loaded = 0
+        for path in self.workspace.glob("*.json"):
+            try:
+                with open(path, "r", encoding="utf-8") as f:
+                    doc = json.load(f)
+                self.documents[path.stem] = doc
+                loaded += 1
+            except (json.JSONDecodeError, OSError) as e:
+                print(f"Warning: skipping corrupt workspace file {path.name}: {e}")
+        if loaded:
+            print(f"Loaded {loaded} document(s) from workspace.")
+
+    def get_document(self, doc_id: str) -> str:
+        """Return document metadata JSON."""
+        return get_document(self.documents, doc_id)
+
+    def get_document_structure(self, doc_id: str) -> str:
+        """Return document tree structure JSON (without text fields)."""
+        return get_document_structure(self.documents, doc_id)
+
+    def get_page_content(self, doc_id: str, pages: str) -> str:
+        """Return page content for the given pages string (e.g. '5-7', '3,8', '12')."""
+        return get_page_content(self.documents, doc_id, pages)