Skip to content

feat: support custom OpenAI base URL via environment variable #35

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
+35 −14
Open

feat: support custom OpenAI base URL via environment variable #35

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b194783
feat: support custom OpenAI base URL via environment variable
Aug 31, 2025
af26ba3
docs: add custom OpenAI API base URL configuration option
Aug 31, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
35 changes: 24 additions & 11 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,9 +1,9 @@
<div align="center">

<a href="https://vectify.ai/pageindex" target="_blank">
<img src="https://github.com/user-attachments/assets/46201e72-675b-43bc-bfbd-081cc6b65a1d" alt="PageIndex Banner" />
</a>

<br/>
<br/>

Expand All @@ -16,16 +16,16 @@
<a href="https://discord.com/invite/VuXuf29EUj">💬 Discord</a>&nbsp; • &nbsp;
<a href="https://ii2abc2jejf.typeform.com/to/tK3AXl8T">✉️ Contact</a>&nbsp;
</p>

</div>

---

# 📄 Introduction to PageIndex
# 📄 Introduction to PageIndex

Are you frustrated with vector database retrieval accuracy for long professional documents? Traditional vector-based RAG relies on semantic *similarity* rather than true *relevance*. But **similarity ≠ relevance** — what we truly need in retrieval is **relevance**, and that requires **reasoning**. When working with professional documents that demand domain expertise and multi-step reasoning, similarity search often falls short.
Are you frustrated with vector database retrieval accuracy for long professional documents? Traditional vector-based RAG relies on semantic _similarity_ rather than true _relevance_. But **similarity ≠ relevance** — what we truly need in retrieval is **relevance**, and that requires **reasoning**. When working with professional documents that demand domain expertise and multi-step reasoning, similarity search often falls short.

Inspired by AlphaGo, we propose [PageIndex](https://vectify.ai/pageindex), a **reasoning-based RAG** system that simulates how **human experts** navigate and extract knowledge from long documents through **tree search**, enabling LLMs to *think* and *reason* their way to the most relevant document sections. It performs retrieval in two steps:
Inspired by AlphaGo, we propose [PageIndex](https://vectify.ai/pageindex), a **reasoning-based RAG** system that simulates how **human experts** navigate and extract knowledge from long documents through **tree search**, enabling LLMs to _think_ and _reason_ their way to the most relevant document sections. It performs retrieval in two steps:

1. Generate a "Table-of-Contents" **tree structure index** of documents
2. Perform reasoning-based retrieval through **tree search**
Expand All @@ -34,9 +34,10 @@ Inspired by AlphaGo, we propose [PageIndex](https://vectify.ai/pageindex), a *
<img src="https://docs.pageindex.ai/images/cookbook/vectorless-rag.png" width="90%">
</div>

### 💡 Features
### 💡 Features

Compared to traditional vector-based RAG, PageIndex features:

- **No Vectors Needed**: Uses document structure and LLM reasoning for retrieval.
- **No Chunking Needed**: Documents are organized into natural sections, not artificial chunks.
- **Human-like Retrieval**: Simulates how human experts navigate and extract knowledge from complex documents.
Expand All @@ -45,12 +46,14 @@ Compared to traditional vector-based RAG, PageIndex features:
PageIndex powers a reasoning-based RAG system that achieved [98.7% accuracy](https://github.com/VectifyAI/Mafin2.5-FinanceBench) on FinanceBench, showing state-of-the-art performance in professional document analysis (see our [blog post](https://vectify.ai/blog/Mafin2.5) for details).

### 🚀 Deployment Options

- 🛠️ Self-host — run locally with this open-source repo
- ☁️ **[Cloud Service](https://dash.pageindex.ai/)** — try instantly with our 🖥️ [Dashboard](https://dash.pageindex.ai/) or 🔌 [API](https://docs.pageindex.ai/quickstart), no setup required

### ⚡ Quick Hands-on

Check out this simple [*Vectorless RAG Notebook*](https://github.com/VectifyAI/PageIndex/blob/main/cookbook/pageindex_RAG_simple.ipynb) — a minimal, hands-on, reasoning-based RAG pipeline using **PageIndex**.
Check out this simple [_Vectorless RAG Notebook_](https://github.com/VectifyAI/PageIndex/blob/main/cookbook/pageindex_RAG_simple.ipynb) — a minimal, hands-on, reasoning-based RAG pipeline using **PageIndex**.

<p align="center">
<a href="https://colab.research.google.com/github/VectifyAI/PageIndex/blob/main/cookbook/pageindex_RAG_simple.ipynb">
<img src="https://img.shields.io/badge/Open_In_Colab-Vectorless_RAG_With_PageIndex-orange?style=for-the-badge&logo=googlecolab" alt="Open in Colab"/>
Expand All @@ -60,7 +63,8 @@ Check out this simple [*Vectorless RAG Notebook*](https://github.com/VectifyAI/P
---

# 📦 PageIndex Tree Structure
PageIndex can transform lengthy PDF documents into a semantic **tree structure**, similar to a _"table of contents"_ but optimized for use with Large Language Models (LLMs). It's ideal for: financial reports, regulatory filings, academic textbooks, legal or technical manuals, and any document that exceeds LLM context limits.

PageIndex can transform lengthy PDF documents into a semantic **tree structure**, similar to a *"table of contents"* but optimized for use with Large Language Models (LLMs). It's ideal for: financial reports, regulatory filings, academic textbooks, legal or technical manuals, and any document that exceeds LLM context limits.

Here is an example output. See more [example documents](https://github.com/VectifyAI/PageIndex/tree/main/tests/pdfs) and [generated trees](https://github.com/VectifyAI/PageIndex/tree/main/tests/results).

Expand Down Expand Up @@ -92,7 +96,7 @@ Here is an example output. See more [example documents](https://github.com/Vecti
...
```

You can either generate the PageIndex tree structure with this open-source repo or try our ☁️ **[Cloud Service](https://dash.pageindex.ai/)** — instantly accessible via our 🖥️ [Dashboard](https://dash.pageindex.ai/) or 🔌 [API](https://docs.pageindex.ai/quickstart), with no setup required.
You can either generate the PageIndex tree structure with this open-source repo or try our ☁️ **[Cloud Service](https://dash.pageindex.ai/)** — instantly accessible via our 🖥️ [Dashboard](https://dash.pageindex.ai/) or 🔌 [API](https://docs.pageindex.ai/quickstart), with no setup required.

---

Expand All @@ -114,6 +118,14 @@ Create a `.env` file in the root directory and add your API key:
CHATGPT_API_KEY=your_openai_key_here
```

Optional: if you're using an OpenAI-compatible endpoint (self-hosted proxy, gateway, or vendor), set a custom API base URL:

```bash
CHATGPT_API_BASE_URL=https://your-openai-compatible-endpoint/v1
```

When set, PageIndex will automatically use this base URL for all Chat Completions requests.

### 3. Run PageIndex on your PDF

```bash
Expand All @@ -135,6 +147,7 @@ You can customize the processing with additional optional arguments:
--if-add-node-summary Add node summary (yes/no, default: no)
--if-add-doc-description Add doc description (yes/no, default: yes)
```

</details>

---
Expand Down Expand Up @@ -180,7 +193,7 @@ Refer to the [API Documentation](https://docs.pageindex.ai/quickstart) for integ

# ⭐ Support Us

Leave a star if you like our project — thank you!
Leave a star if you like our project — thank you!

<p align="center">
<img src="https://github.com/user-attachments/assets/eae4ff38-48ae-4a7c-b19f-eab81201d794" width="75%">
Expand Down
14 changes: 11 additions & 3 deletions pageindex/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,14 @@
from types import SimpleNamespace as config

CHATGPT_API_KEY = os.getenv("CHATGPT_API_KEY")
CHATGPT_API_BASE_URL = os.getenv("CHATGPT_API_BASE_URL")

def _openai_kwargs(api_key):
"""Build kwargs for OpenAI clients with optional base_url from env."""
kwargs = {"api_key": api_key}
if CHATGPT_API_BASE_URL:
kwargs["base_url"] = CHATGPT_API_BASE_URL
return kwargs

def count_tokens(text, model=None):
if not text:
Expand All @@ -28,7 +36,7 @@ def count_tokens(text, model=None):

def ChatGPT_API_with_finish_reason(model, prompt, api_key=CHATGPT_API_KEY, chat_history=None):
max_retries = 10
client = openai.OpenAI(api_key=api_key)
client = openai.OpenAI(**_openai_kwargs(api_key))
for i in range(max_retries):
try:
if chat_history:
Expand Down Expand Up @@ -60,7 +68,7 @@ def ChatGPT_API_with_finish_reason(model, prompt, api_key=CHATGPT_API_KEY, chat_

def ChatGPT_API(model, prompt, api_key=CHATGPT_API_KEY, chat_history=None):
max_retries = 10
client = openai.OpenAI(api_key=api_key)
client = openai.OpenAI(**_openai_kwargs(api_key))
for i in range(max_retries):
try:
if chat_history:
Expand Down Expand Up @@ -91,7 +99,7 @@ async def ChatGPT_API_async(model, prompt, api_key=CHATGPT_API_KEY):
messages = [{"role": "user", "content": prompt}]
for i in range(max_retries):
try:
async with openai.AsyncOpenAI(api_key=api_key) as client:
async with openai.AsyncOpenAI(**_openai_kwargs(api_key)) as client:
response = await client.chat.completions.create(
model=model,
messages=messages,
Expand Down