chat-with-pdf 0.3.3

Chat with your PDFs using local embedding search and OpenAI.

📄 Chat with PDF

Chat with your PDF documents easily using local embeddings and powerful LLMs through a unified SDK. Upload any PDF and ask natural language questions about its content — powered by semantic search and AI.

---

🛠️ Installation

pip install chat-with-pdf

Or using Poetry:

poetry add chat-with-pdf

---

✨ Quickstart Example

1. Set your credentials and optionally choose a model/provider:

# Default provider key
export OPENAI_API_KEY="sk-your-openai-key"

# The model for the provider
export OPENAI_MODEL="gpt-4"

# Switch to another provider (e.g., perplexity, openai or deepseek)
export LLM_PROVIDER="perplexity"

2. Use the SDK to chat with any PDF:

from chat_with_pdf import PDFChat

# Local PDF file
chat = PDFChat("path/to/your/document.pdf")
print(chat.ask("Summarize the introduction section."))

# Remote URL
chat = PDFChat("https://example.com/sample.pdf")
print(chat.ask("What is the main point of this document?"))

# PDF in memory
with open("path/to/your/document.pdf", "rb") as f:
    data = f.read()
chat = PDFChat(data)
print(chat.ask("Give me a brief overview."))

---

⚙️ Configuration Options

Configure via environment variables (in order of precedence):

Variable	Purpose	Default
LLM_PROVIDER	Provider to use (openai, perplexity, deepseek)	openai
OPENAI_API_KEY	Your OpenAI API key	—
OPENAI_MODEL	GPT model name (used for all providers)	gpt-3.5-turbo
EMBEDDING_MODEL	Embedding model	all-MiniLM-L6-v2
DEFAULT_CHUNK_SIZE	Characters per text chunk	500
TOP_K_RETRIEVAL	Number of chunks to retrieve per query	5

💡 For local development, you can also create a .env file with these variables and the SDK will load it automatically.

--------------------- | :------------------------------------------------ | :------------------------ | | LLM_PROVIDER | Provider to use (openai, perplexity, deepseek) | openai | | OPENAI_API_KEY | Your OpenAI API key | — | | OPENAI_MODEL | GPT model name (OpenAI) | gpt-3.5-turbo | | PERPLEXITY_API_KEY | Your Perplexity API key | — | | PERPLEXITY_MODEL | Model name (Perplexity) | perplexity-v1 | | DEEPSEEK_API_KEY | Your DeepSeek API key | — | | DEEPSEEK_MODEL | Model name (DeepSeek) | deepseek-v1 | | EMBEDDING_MODEL | Embedding model | all-MiniLM-L6-v2 | | DEFAULT_CHUNK_SIZE | Characters per text chunk | 500 | | TOP_K_RETRIEVAL | Number of chunks to retrieve per query | 5 |

💡 For local development, you can also create a .env file with these variables and the SDK will load it automatically.

---

🔥 Advanced Usage

Override provider/model at runtime:

from chat_with_pdf import PDFChat

# Use GPT-4 on OpenAI
chat = PDFChat("doc.pdf")
print(chat.ask("What are the key findings?", provider="openai", model="gpt-4"))

# Use DeepSeek
print(chat.ask("Summarize", provider="deepseek"))

---

📝 License

This project is licensed under the MIT License.

---

🌟 Acknowledgements

• OpenAI
• HuggingFace Sentence Transformers
• PyMuPDF