How kashish1530/agentic-ai works
This repository is a five-day collection of agentic-AI lessons and runnable demos, culminating in a persistent, tool-using assistant with Streamlit and FastAPI interfaces. The principal workflow is a user request entering the capstone, which loads and updates user memory, calls the model, dispatches requested tools, applies safety checks, and returns an answer with source information. Separate lessons demonstrate alternative agent patterns, multi-agent routing, context management, reliability, evaluation, and model routing. Some branches are documented only by filenames; their internal wiring is therefore omitted.
Main entry points: day5_s1_capstone.py Capstone Agent · day5_s2_streamlit.py Streamlit Chat · day5_s2_fastapi.py FastAPI Interface · memory_store User Memory
Components 24
User Interfaces
- Streamlit Chat
DAY5/day5_s2_streamlit.py - FastAPI Interface
DAY5/day5_s2_fastapi.py
Capstone Runtime
- Capstone Agent
DAY5/day5_s1_capstone.py - User Memory
DAY5/memory_store - Tool Dispatch
DAY5/day5_s1_capstone.py - Safety Checks
DAY5/day5_s1_capstone.py - Cited Sources
DAY5/day5_s1_capstone.py
Agent Patterns
- Tool Calling
DAY1/s1_tool_calling.py - Tool Examples
DAY1/s2_tools.py - Multi-step Agent
DAY1/s3_multistep.py - MCP Lesson
DAY1/s4_mcp.py - ReAct Loop
DAY2/day2_s1_react.py - LangGraph Agent
DAY2/day2_s2_langgraph.py - Multi-agent Routing
DAY3/day3_s1_multiagent.py - Pipeline Comparison
DAY3/day3_s2_pipeline_vs_agent.py
Trust And Operations
- Context Management
DAY3/day3_s3_contextengineering.py - Run Observability
DAY3/day3_s4_observability.py - Reliability Controls
DAY2/day2_s4_reliability.py - Prompt Guardrails
DAY4/day4_s2_guardrails.py - Agent Evaluation
DAY4/DAY4_S3.PY - Model Routing
DAY4/DAY4_S4.PY
Other components
- User
- Groq Model API
- Tool Services
Connections 24
- User → Streamlit Chat: submits chat
- Streamlit Chat → Capstone Agent: invokes agent
- User → FastAPI Interface: sends request
- FastAPI Interface → Capstone Agent: invokes agent
- Capstone Agent → User Memory: loads and saves
- Capstone Agent → Groq Model API: requests completion
- Capstone Agent → Tool Dispatch: dispatches calls
- Tool Dispatch → Tool Services: runs tools
- Capstone Agent → Safety Checks: checks response
- Capstone Agent → Cited Sources: records citations
- Streamlit Chat → User Memory: loads preferences
- Tool Calling → Groq Model API: requests tool calls
- Tool Examples → Groq Model API: requests completion
- Multi-step Agent → Groq Model API: requests tool calls
- ReAct Loop → Groq Model API: requests next step
- LangGraph Agent → Groq Model API: runs agent
- Multi-agent Routing → Groq Model API: routes work
- Pipeline Comparison → Groq Model API: compares strategies
- Context Management → Groq Model API: compares prompts
- Run Observability → Groq Model API: invokes model
- Reliability Controls → Groq Model API: requests tool calls
- Prompt Guardrails → Groq Model API: sends guarded prompt
- Agent Evaluation → Groq Model API: judges responses
- Model Routing → Groq Model API: selects model tier