ChromaView Master is a Streamlit-based tool designed to help you understand, visualize, and manipulate Chroma database collections. I created this tool for situations where I frequently create ad-hoc RAG (Retrieval Augmented Generation) Chroma databases, and I needed an easier way to inspect their structure and content without relying solely on an LLM. I just needed a simple tool to quickly analyze chroma databases.
This is actively being revised still....
- Collection Overview: Get a quick overview of all the collections stored within your Chroma database, along with basic statistics like the number of documents in each collection.
- Topic Modeling: Employ Latent Dirichlet Allocation (LDA) to uncover the main themes and topics discussed within your document collections. Understand the distribution of topics across documents and identify the most representative words for each topic. Includes an interactive pie chart to visualize topic distribution.
- Similarity Search: Enter a query and find the most similar documents within your collection based on semantic similarity. This can be helpful for retrieving relevant information or identifying duplicates.
- Knowledge Graph Visualization: Extract entities (people, places, organizations, etc.) and their relationships from your documents and visualize them as an interactive network graph using Plotly and NetworkX. Gain insights into the connections between different concepts in your data.
- Tag Cloud: Visualize the most frequent words in your collection as a word cloud, where the size of each word corresponds to its frequency. Quickly grasp the dominant themes and vocabulary used in your documents.
- Treemap Chart: Visualize hierarchical data with nested rectangles using a treemap chart. Each rectangle represents a category, and its size is proportional to its value. This replaces the previous sunburst chart for a more intuitive hierarchical representation.
- Document Length Analysis: Analyze the distribution of document lengths within your collection. Identify outliers and understand the general characteristics of your text data. Includes detailed statistics such as mean, median, mode, min, max, standard deviation, and percentiles.
- Entity Co-occurrence Matrix: Explore how often different entities co-occur within the same documents. This can reveal patterns and associations between entities. Visualize the matrix using heatmaps and network graphs to identify strong co-occurrences and relationships dynamically.
- Sentiment Analysis: Perform sentiment analysis on your documents to determine whether the overall tone is positive, negative, or neutral. Understand the emotional content of your text data with enhanced visualizations such as histograms, density plots, pie charts, and word clouds for different sentiment categories.
- Knowledge Graph (PyVis): Dynamically explore relationships between entities by entering a topic or entity of interest. The graph will display nodes and edges related to the entered topic, helping to uncover hidden connections within your data.
- Python 3.7 or above
- Operating System: Windows, macOS, or Linux
- Conda (recommended for creating a clean environment)
-
Prerequisites:
- Ensure you have Python and
pipinstalled. (I recommend using a clean conda environment.) - Install required libraries:
pip install -r requirements.txt
- Download the spaCy model:
python -m spacy download en_core_web_sm
- Ensure you have Python and
-
Run ChromaView Master:
- Locate the path to your Chroma database.
- Execute the following command in your terminal:
(Replace
streamlit run chromaViewMaster.py /path_to_your_database/testdb
/path_to_your_database/testdbwith your actual path)
Note: This tool currently assumes your database has no password protection.
If your Chroma database is located at /Users/yourname/documents/mychromadb, you would run:
streamlit run chromaMaster.py /Users/yourname/documents/mychromadb
## Troubleshooting
- Large datastores may take awhile to generate. This is a limitation of browser & streamlit.
You can find me on discord Preston McCauley