How anurag07-crypto/ai_pokedex works
AI Pokedex identifies an uploaded Pokémon image, displays its predicted label and confidence, then retrieves structured Pokémon details for presentation. The frontend performs image cleanup and classification, calls the FastAPI service for search-backed details, and formats the returned information with a separate Groq service. The backend and Work_dir/main_pipeline.py also expose agent workflows, though the sampled frontend calls prediction and backend endpoints directly rather than using that pipeline. Source excerpts do not establish model-weight loading or saved prediction behavior.
Main entry points: front_server.py Streamlit app · pokedex_agent.py Pokédex agent · back_server.py FastAPI endpoint · main_pipeline.py Workflow pipeline
Components 12
User interface
- Streamlit app
Frontend/front_server.py - Background removal
Frontend/services.py - Response formatter
Frontend/services.py
Image inference
- CNN classifier
Work_dir/Cnn_model.py - Workflow pipeline
Work_dir/main_pipeline.py
Details API
- FastAPI endpoint
Backend/back_server.py
Knowledge agent
- Pokédex agent
Work_dir/pokedex_agent.py - Pokémon schema
Work_dir/pokedex_agent.py
Other components
- Image uploader
- Tavily search
- Groq LLM
- Application logs
logs
Connections 13
- Image uploader → Streamlit app: uploads image
- Streamlit app → Background removal: cleans image
- Streamlit app → CNN classifier: classifies image
- Streamlit app → FastAPI endpoint: posts label
- FastAPI endpoint → Pokédex agent: requests details
- Pokédex agent → Tavily search: searches context
- Pokédex agent → Groq LLM: structures results
- Pokédex agent → Pokémon schema: formats fields
- FastAPI endpoint → Streamlit app: returns details
- Streamlit app → Response formatter: formats details
- Response formatter → Groq LLM: invokes model
- Workflow pipeline → CNN classifier: classifies image
- Workflow pipeline → Pokédex agent: requests details