Skip to content

Latest commit

 

History

History
173 lines (132 loc) · 9.83 KB

README.md

File metadata and controls

173 lines (132 loc) · 9.83 KB

LLM Context Generator Streamlit App

Overview

This Streamlit application provides a user-friendly interface to aggregate content from various source files (including PDFs, plain text, Markdown, and source code) into a single, structured context file suitable for Large Language Models (LLMs). It leverages the capabilities of two external command-line tools:

  1. llama-parse: Used to extract text content from PDF documents into Markdown format via the LlamaCloud API.
  2. files-to-prompt: Used to gather text-based files from specified directories and format the combined content recursively.

Additionally, it includes a feature to suggest an expert persona system prompt based on the generated context, utilizing Google's Gemini AI model.

The primary goal is to simplify preparing large, diverse context sets for LLMs, outputting a single file where each source document is wrapped in Claude XML-style <document> tags (files-to-prompt --cxml).

Key External Dependencies:

  • PDF Parsing (llama-parse): Requires internet and a one-time manual authentication (llama-parse auth) using a LlamaCloud API key.
  • Prompt Suggestion (Gemini): Requires internet and a Google AI API key set as the GEMINI_API_KEY environment variable.

This app checks for necessary prerequisites but does not store or directly handle API keys itself.

Features

  • Multi-Source Processing: Handles PDF documents and various text-based files (.txt, .md, .py, .js, .json, .xml, etc.).
  • PDF Parsing Integration: Uses llama-parse to convert PDF content to Markdown via LlamaCloud. Parsed files are stored in a configurable sub-directory.
  • LlamaParse Authentication Check: Verifies if llama-parse auth has been completed before attempting PDF parsing.
  • Flexible Processing Modes:
    • Both: Parses PDFs, then combines all files recursively from the main TXT directory (including parsed PDFs in the subfolder).
    • PDF only: Parses PDFs, then combines only the results from the parsed PDF sub-directory.
    • TXT only: Skips PDF parsing, clears old parsed results, then combines only files recursively from the main TXT directory.
  • File Management UI: Upload, view, and delete PDF and Plaintext files in their respective input directories directly through the app.
  • Configurable Paths: Set input/output directories and filenames easily via the sidebar.
  • ✨ System Prompt Suggestion: After generating context, uses Google's Gemini model (gemini-2.5-pro-exp-03-25 as of 2025-03-27) to analyze a snippet and suggest a system prompt instructing an AI to act as a relevant expert using that context. Requires GEMINI_API_KEY environment variable.
  • Structured Output Format: Generates context using Claude XML tags (<document path="...">...</document>).
  • In-App Previews: Displays generated context and suggested system prompts.
  • Status & Error Feedback: Provides messages, progress indicators, and toasts.

Screenshots

Mode Selection

Mode Selection

Progress

Progress

Generated Context & Suggested System Prompt

Generated Context Preview

Suggested System Prompt

File Management

PDF file management

Plaintext file management

Prerequisites (Linux Host Recommended)

  1. Python: Version 3.8+ recommended (python3 --version).
  2. uv (Recommended Installer/Venv): A fast, modern Python package installer and virtual environment manager. While standard pip/venv works, uv significantly speeds up installation.
    • Install uv: 
      # Linux/macOS/WSL
curl -LsSf [https://astral.sh/uv/install.sh](https://astral.sh/uv/install.sh) | sh
    • Verify: uv --version (You might need to restart your terminal after installation).
  3. Node.js & npm: Required only for installing llama-parse-cli.
    • Check: node -v and npm -v
    • Install (e.g., Ubuntu/Debian): sudo apt update && sudo apt install nodejs npm -y
  4. LlamaCloud API Key: Needed only for the one-time llama-parse auth setup in your terminal. Get from LlamaCloud.
  5. Google AI API Key: Needed for the "Suggest System Prompt" feature. Get from Google AI Studio. Must be set as the GEMINI_API_KEY environment variable.

Setup Instructions (Linux Host using uv)

  1. Get the Code:

    • Save the Streamlit Python scripts (context_generator_app.py, gemini_interface.py) and requirements.txt into a dedicated directory.
    • Navigate there: cd /path/to/your/app_directory

  2. Create Virtual Environment & Install Dependencies (using uv):

    • uv combines environment creation and package installation. 
      # Create a virtual environment named .venv in the current directory
uv venv

# Activate the environment (syntax depends on shell)
# Linux/macOS (bash/zsh):
source .venv/bin/activate
# Windows (Command Prompt):
# .venv\Scripts\activate.bat
# Windows (PowerShell):
# .venv\Scripts\Activate.ps1

# Install dependencies from requirements.txt using uv
uv pip install -r requirements.txt
    • (Alternatively, without requirements.txt: uv pip install streamlit files-to-prompt google-generativeai)
    • (Using standard pip/venv: python3 -m venv venv && source venv/bin/activate && pip install -r requirements.txt)

  3. Install llama-parse CLI Tool:

    • npm install -g llama-parse-cli (You might need sudo or specific npm configurations).
    • Verify: llama-parse --version
    • If not found, ensure npm global bin directory is in your system PATH and restart your terminal.

  4. Set Google AI API Key Environment Variable:

    • Set GEMINI_API_KEY before launching Streamlit. Choose the command for your terminal session:

      Linux / macOS (.bashrc, .zshrc, etc.):

      export GEMINI_API_KEY="YOUR_API_KEY_HERE"

      Windows (Command Prompt - current session only):

      set GEMINI_API_KEY=YOUR_API_KEY_HERE

      Windows (PowerShell - current session only):

      $env:GEMINI_API_KEY = 'YOUR_API_KEY_HERE'

    • Replace YOUR_API_KEY_HERE with your key. For persistence, add to your shell profile or system variables.

  5. Authenticate llama-parse (CRITICAL ONE-TIME STEP):

    • Links the CLI tool to your LlamaCloud account.
    • Run manually in your terminal: 
      llama-parse auth
    • Enter your LlamaCloud API key (starts with llx-). Creates ~/.llama-parse/config.json.

  6. Delete GEMINI_API_KEY.py (Cleanup):

    • If you previously had this file, delete it.

Usage Guide

  1. Prepare Input Folders: Create/populate directories (e.g., pdfs_to_parse, txt_files) or configure paths in sidebar.
  2. Activate Environment & Run App:
    • Open terminal, navigate to the app directory.
    • Activate the virtual environment: (e.g., source .venv/bin/activate)
    • Set Environment Variable: Ensure GEMINI_API_KEY is set for the current session (Setup Step 4).
    • Launch: streamlit run context_generator_app.py
    • Open the provided local URL in your browser.
  3. Configure Settings (Sidebar): Adjust paths if needed. Verify directory status.
  4. Manage Files (Tabs 2 & 3 - Optional): Upload/view/delete source files.
  5. Select Processing Mode (Tab 1): Choose TXT only, PDF only, or Both.
  6. Generate Context (Tab 1): Click "Generate Context File". Monitor progress.
  7. Review Context Output (Tab 1): View generated file path and preview content.
  8. Suggest System Prompt (Tab 1 - Optional): Click "✨ Suggest System Prompt". View suggested persona prompt in the text area.
  9. Use Output: Copy prompt/context from UI or the generated output file.

Troubleshooting

  • Error: LlamaParse Auth Missing...: Run llama-parse auth manually (Setup Step 5). Check ~/.llama-parse/config.json.
  • Error: 'llama-parse'/'files-to-prompt' command not found: Check tool installation and system PATH. Restart terminal. Verify with --version.
  • Error: uv: command not found: Ensure uv is installed and its location is in PATH (Setup Step 2 / Prerequisites).
  • Error: Configuration Error: Environment variable 'GEMINI_API_KEY' not found...: Ensure variable set before streamlit run. Check name/value. Check session persistence.
  • Error: API Error: Invalid Google AI API Key...: Verify key value in environment variable.
  • Error: API Error: Quota exceeded...: Check Google Cloud Gemini API usage/limits.
  • Error: API Error: Model '...' not found or permission denied...: Check model name (gemini-1.5-pro-latest) and API key permissions.
  • Error: Content generation blocked due to safety filters...: Google safety filters triggered. Review input/output content.
  • PDF Parsing Failures: Check internet, LlamaCloud status/key, PDF file validity/complexity. Check UI error expander.
  • Permission Errors: Check read/write permissions for input/output directories for the user running Streamlit.

Dependencies

  • streamlit: (Python - via uv pip or pip) Web application framework.
  • files-to-prompt: (Python - via uv pip or pip) File gathering and formatting.
  • google-generativeai: (Python - via uv pip or pip) Google AI SDK for Python (Gemini).
  • uv: (Installer - via curl/pipx/etc.) Recommended package installer/venv manager.
  • llama-parse-cli: (Node.js - via npm install -g) CLI for LlamaParse PDF processing. Requires manual llama-parse auth.