vision-streamlit

This is a test of the llama3.2-vision model.

Prereq

I used ollama for this. So I did

ollama pull llama3.2-vision

Virtual environment

python -m venv .venv

python pip install streamlit ollama

How to run

streamlit run app.py

Network diagram

flowchart LR
  %% Direction
  %% =========================
  %% Groups (by color via classDefs at bottom)
  %% browser = blue, mcp = gray, app = green, service = yellow, storage = red

  %% Nodes
  enduser[End User]
  browser[Client Browser]
  claude[Claude Desktop]
  streamlit[Streamlit App]
  data[(Marine Cadastre)]
  cache[(Local Cache)]
  ollama[Ollama Daemon]
  model[[llama3.2-vision]]
  ext[[Optional External APIs*]]

  %% Edges
  enduser --> |Gathers historical intel from| browser
  enduser --> |Gather imagery information from| streamlit
  
  browser -->|MCP| claude
  claude -->|Requests data from| data

  streamlit -->|read vessel metadata| data
  streamlit -->|write results| cache

  streamlit -->|vision inference| ollama
  ollama -->|Load/Infer| model
  ollama -->|predictions / text| streamlit

  streamlit -->|optional queries*| ext
  streamlit -->|render dashboards| browser

  %% Classes
  class enduser,browser browser;
  class claude, mcp;
  class streamlit app;
  class ollama,model,ext service;
  class data,cache storage;

  %% Styles
  classDef browser fill:#CFE2FF,stroke:#004085,color:#001933,stroke-width:1px;
  classDef mcp fill:#E2E3E5,stroke:#383D41,color:#1d1f22,stroke-width:1px;
  classDef app fill:#D4EDDA,stroke:#155724,color:#0d2b18,stroke-width:1px;
  classDef service fill:#FFF3CD,stroke:#856404,color:#3d2d00,stroke-width:1px;
  classDef storage fill:#F8D7DA,stroke:#721C24,color:#3a0c10,stroke-width:1px;

Loading