llm-chatbots

A Python chat program supporting multiple LLM providers (Google Gemini, OpenAI, Anthropic) with runtime model switching, streaming responses, and a web UI that supports document uploads (PDF/txt/md) with retrieval-augmented question answering.

Supported Providers

Provider	Models	Free tier
Google Gemini	gemini-3.1-flash-lite, gemini-2.5-flash, ...	aistudio.google.com
OpenAI	gpt-4o, gpt-4o-mini, ...	platform.openai.com
Anthropic	claude-opus-4-7, claude-sonnet-4-6, ...	console.anthropic.com

Requirements

Python 3.10+

Setup

1. Create and activate a virtual environment:

   python3.10 -m venv venv
   source venv/bin/activate

2. Install dependencies:

   pip install -r requirements.txt

3. Copy .env.example to .env and fill in your keys:

   cp .env.example .env

   PROVIDER=gemini              # gemini | openai | anthropic
   GEMINI_API_KEY=your_key
   OPENAI_API_KEY=your_key
   ANTHROPIC_API_KEY=your_key
   MODEL=                       # leave blank to use the default for the provider
   SYSTEM_PROMPT=You are a helpful assistant.
   

   You only need the API key for the provider(s) you plan to use.

Usage

# activate the venv first if not already active
source venv/bin/activate

python main.py                # CLI mode (default)
python main.py --mode web     # web UI in your browser (Streamlit)

CLI mode

Type your message and press Enter. Use Ctrl+C or Ctrl+D to quit.

Switch models at runtime:

/model                              # show current provider and model
/model openai/gpt-4o                # switch to OpenAI GPT-4o
/model anthropic/claude-opus-4-7    # switch to Anthropic Claude
/model gemini/gemini-2.5-flash      # switch back to Gemini

Switching starts a fresh conversation (history is cleared).

Web mode

python main.py --mode web launches a Streamlit app in your browser. The sidebar lets you pick a provider, set the model, edit the system prompt, apply changes, or clear the conversation. Responses stream as they are generated.

Document uploads (web only)

Attach .pdf, .txt, or .md files from the sidebar to ask questions about them. Documents are chunked and embedded in memory using Gemini's gemini-embedding-001; on each message, the top excerpts most relevant to your question are retrieved and sent along with it.

• Requires GEMINI_API_KEY in .env for embeddings, regardless of which chat provider you're using.
• Uncheck "Use uploaded documents in answers" to skip retrieval without removing the indexed files.
• Click ✕ next to a filename to remove it from the index.
• Indexed documents do not persist across app restarts.
• Removing a file or unchecking the box after the model has already answered with documents will clear the conversation, so the model can't continue referencing them from earlier turns. A toast notifies you when this happens.

Running Tests

python -m pytest        # run all tests
python -m pytest tests/test_foo.py::test_bar   # run a single test

Project Structure

main.py                  # entry point; --mode cli|web dispatch
app.py                   # Streamlit web UI
chat.py                  # ChatSession: thin wrapper over provider
config.py                # shared constants (providers, defaults, RAG knobs)
rag.py                   # file extraction, chunking, embeddings, retrieval
providers/
  base.py                # BaseProvider ABC (send + send_stream)
  gemini.py              # GeminiProvider
  openai.py              # OpenAIProvider
  anthropic.py           # AnthropicProvider
  __init__.py            # create_provider() factory
tests/
  test_chat.py
  test_main.py
  test_rag.py
  test_providers_*.py