Skip to content

Aider Genius Lite - AI Coding Assistant#

Overview#

Aider Genius Lite is a simple Jac-based Streamlit application for AI-powered code generation with task planning and validation. It provides a clean, intuitive interface for generating code from natural language requests with real-time feedback and validation.

Features#

  • Simple, clean interface for code generation requests
  • Pre-built example requests for quick testing
  • Real-time task progress and results display
  • Easy code viewing with syntax highlighting
  • AI feedback and validation results
  • Separate Jac Streamlit frontend and backend

Files#

  • genius_lite.jac - Backend with AI logic, task processing, and API endpoints
  • frontend.jac - Streamlit frontend interface

Complete Code Preview#

Here's what you'll build - a simple AI coding assistant in just two files:

Aider Genius Lite Frontend

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
import from byllm.lib { Model }
import from pathlib { Path }

glob llm = Model(model_name="gemini/gemini-2.5-flash");

# Simple task object with semantic annotations
obj Task {
    has name: str,
        type: str,
        details: str,
        priority: int = 1;
}

sem Task= "A specific development task with clear implementation requirements";
sem Task. name= "Clear, descriptive name for the task";
sem Task. type= "Task category: code, fix, docs, or test";
sem Task. details= "Specific implementation instructions";
sem Task. priority= "Task priority: 1=high, 2=medium, 3=low";

obj CodeResult {
    has task_name: str = "",
        code: str = "",
        status: str = "",
        feedback: str = "";
}

sem CodeResult= "Result of code generation task";
sem CodeResult. task_name= "Name of the completed task";
sem CodeResult. code= "Generated code solution";
sem CodeResult. status= "Task completion status: success or failed";
sem CodeResult. feedback= "Validation feedback and suggestions";

# Core AI functions
def create_plan(request: str) -> list[Task] by llm(method="Reason");
def generate_solution(task: Task) -> str by llm(method="Reason");
def validate_code(code: str, task: Task) -> str by llm(method="Reason");

# Nodes for walker traversal
node TaskNode {
    has task: Task = Task(name="", type="", details="", priority=1),
        code: str = "",
        feedback: str = "",
        status: str = "pending",
        result: CodeResult = CodeResult();

    def process_task {
        print(f"Working on: {self.task.name}");
        self.code = generate_solution(self.task);
        self.feedback = validate_code(self.code, self.task);
        self.status = "success" if self.feedback else "failed";
        print(f"Completed: {self.task.name}");

        # Create result
        self.result.task_name = self.task.name;
        self.result.code = self.code;
        self.result.status = self.status;
        self.result.feedback = self.feedback;
    }
}

node SummaryNode {
    has results: list[CodeResult] = [];

    def show_summary {
        output = f"Summary ({len(self.results)} tasks):\n\n";
        for result in self.results {
            status_icon = "SUCCESS" if result.status == "success" else "FAILED";
            output += f"{status_icon} {result.task_name}\n";
            if result.code and len(result.code) > 0 {
                code_preview = result.code[:300];
                if len(result.code) > 300 {
                    code_preview += "...";
                }
                output += f"   Code: {code_preview}\n";
            }
            if result.feedback {
                output += f"   Feedback: {result.feedback[:100]}...\n";
            }
            output += "\n";
        }
        print(output);
    }
}

# GeniusAgent as a walker
walker GeniusAgent {
    has request: str,
        tasks: list[Task] = [],
        results: list[CodeResult] = [],
        current_task_index: int = 0;

    can start with `root entry {
        print("Genius Lite - AI Coding Assistant");
        print("Simple, structured code generation with validation");
        print("=" * 50);
        self.tasks = create_plan(self.request);
        print(f"Created {len(self.tasks)} tasks");

        if len(self.tasks) > 0 {
            # Create task nodes and connect them
            task_nodes = [];
            for task in self.tasks {
                task_node = TaskNode();
                task_node.task = task;
                task_nodes.append(task_node);
            }
            # Connect nodes in sequence
            for i in range(len(task_nodes) - 1) {
                task_nodes[i] ++> task_nodes[i + 1];
            }
            # Connect last task node to summary
            summary_node = SummaryNode();
            summary_node.results = self.results;
            task_nodes[-1] ++> summary_node;
            # Start traversal from first task
            visit task_nodes[0];
        } else {
            print("No tasks created, ending execution");
        }
    }

    can process_task with TaskNode entry {
        # Let the node handle its own processing
        here.process_task();

        # Collect result from node and add to walker's results
        self.results.append(here.result);

        # Continue to next node
        visit [-->];
    }

    can show_summary with SummaryNode entry {
        # Pass results to the summary node
        here.results = self.results;
        # Let the node handle its own summary display
        here.show_summary();

        # Report results for API endpoint
        report {
            "status": "success",
            "tasks": [
                {
                    "name": result.task_name,
                    "code": result.code,
                    "status": result.status,
                    "feedback": result.feedback
                } for result in self.results
            ],
            "total_tasks": len(self.results)
        } ;
    }
}

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
import streamlit as st;
import requests;

def bootstrap_frontend(token: str) {
    st.set_page_config(
        page_title="Genius Lite - AI Coding Assistant", page_icon="💻", layout="wide"
    );

    st.title("🚀 Genius Lite - AI Coding Assistant");
    st.markdown("✨ Simple, structured code generation with validation");

    # Initialize session state
    if "results" not in st.session_state {
        st.session_state.results = None;
    }
    if "loading" not in st.session_state {
        st.session_state.loading = False;
    }

    # Main interface
    st.markdown("### 💡 What would you like me to code for you?");

    # Examples
    with st.expander("💫 Example Requests") {
        if st.button("🧮 Create a Python calculator") {
            st.session_state.user_request = "Create a Python calculator with basic math operations";
        }
        if st.button("🎮 Make a simple game") {
            st.session_state.user_request = "Create a simple number guessing game in Python";
        }
    }

    # Text input
    user_request = st.text_area(
        "📝 Describe your coding request:",
        value=st.session_state.get("user_request", ""),
        height=100,
        placeholder="e.g., Create a Python script that sorts a list of numbers 📊"
    );

    # Generate button
    if st.button("🚀 Generate Code") {
        if user_request.strip() {
            st.session_state.loading = True;
            st.session_state.results = None;
            st.rerun();
        } else {
            st.warning("⚠️ Please enter a coding request!");
        }
    }

    # Show loading state
    if st.session_state.loading {
        with st.spinner("🧠 AI is thinking and generating code...") {
            try {
                response = requests.post(
                    "http://localhost:8000/walker/GeniusAgent",
                    json={"request": user_request},
                    headers={"Authorization": f"Bearer {token}"}
                );

                if response.status_code == 200 {
                    result = response.json();
                    st.session_state.results = result.get("reports", [{}])[0];
                    st.session_state.loading = False;
                    st.success("✅ Code generation completed!");
                    st.rerun();
                } else {
                    st.error(f"❌ Generation failed: {response.text}");
                    st.session_state.loading = False;
                }
            } except requests.exceptions.Timeout {
                st.error(
                    "⏰ Request timed out. Please try again with a simpler request."
                );
                st.session_state.loading = False;
            } except Exception as e {
                st.error(f"❌ Error: {str(e)}");
                st.session_state.loading = False;
            }
        }
    }

    # Display results
    if st.session_state.results {
        results = st.session_state.results;
        st.markdown("---");
        st.markdown("## 🎯 Generated Code Results");
        if "total_tasks" in results {
            st.info(f"📋 Completed {results['total_tasks']} tasks");
        }
        if "tasks" in results and results["tasks"] {
            i = 1;
            for task in results["tasks"] {
                task_name = task.get('name', 'Unnamed Task');
                with st.expander(f"📝 Task {i}: {task_name}") {
                    # Task status
                    status = task.get("status", "unknown");
                    if status == "success" {
                        st.success("✅ Completed successfully");
                    } else {
                        st.error("❌ Failed");
                    }
                    # Generated code
                    if task.get("code") {
                        st.markdown("**🔧 Generated Code:**");
                        st.code(task["code"]);
                        # Copy button
                        if st.button(f"📋 Copy Code {i}") {
                            st.write(
                                "📋 Code copied to clipboard! (Use Ctrl+C to copy manually)"
                            );
                        }
                    }
                    # Feedback
                    if task.get("feedback") {
                        st.markdown("**💬 AI Feedback:**");
                        st.info(task["feedback"]);
                    }
                }
                i += 1;
            }
        }
        # Clear results button
        if st.button("🗑️ Clear Results") {
            st.session_state.results = None;
            st.session_state.user_request = "";
            st.rerun();
        }
    }

    # Footer
    st.markdown("---");
    st.markdown("**🤖 Powered by Genius Lite AI Assistant**");
    st.markdown("✨ Features: 📋 Task planning • 🔧 Code generation • ✅ Validation");
}

glob INSTANCE_URL = "http://localhost:8000",
     TEST_USER_EMAIL = "test@mail.com",
     TEST_USER_PASSWORD = "password",
     response = requests.post(
         f"{INSTANCE_URL}/user/login",
         json={"email": TEST_USER_EMAIL, "password": TEST_USER_PASSWORD}
     );

with entry {
    if response.status_code != 200 {
        # Try registering the user if login fails
        response = requests.post(
            f"{INSTANCE_URL}/user/register",
            json={"email": TEST_USER_EMAIL, "password": TEST_USER_PASSWORD}
        );
        assert response.status_code == 201;
        response = requests.post(
            f"{INSTANCE_URL}/user/login",
            json={"email": TEST_USER_EMAIL, "password": TEST_USER_PASSWORD}
        );
        assert response.status_code == 200;
    }
}

glob token = response.json()["token"];

with entry {
    bootstrap_frontend(token);
}

Setup#

Installation#

Install dependencies:

pip install jac-streamlit requests jaclang jac-scale byllm
  1. Open your browser to http://localhost:8501

Usage#

  1. Start the backend server:
jac start genius_lite.jac

The backend API will be available at http://localhost:8000

  1. Launch the frontend (in a new terminal):
jac streamlit frontend.jac

The web interface will be available at http://localhost:8501

Usage Examples#

Basic Code Generation#

Try these example requests:

  • "Create a Python calculator with basic math operations"
  • "Make a simple number guessing game in Python"

Looking for the full version? This is a lite version for learning purposes. Check out the full-scale Aider Genius project for a complete implementation with advanced features.

Code Architecture Deep Dive#

Task Planning System#

The core of Genius Lite's intelligence lies in its task planning system, which:

  • Analyzes the complexity and scope of user requests
  • Breaks down large problems into smaller, manageable components
  • Creates dependency graphs for complex multi-step tasks
  • Prioritizes execution order for optimal results

Code Generation Pipeline#

  1. Request Analysis: Understanding the user's intent and requirements
  2. Context Gathering: Identifying relevant libraries, frameworks, and patterns
  3. Code Synthesis: Generating code that meets the specified requirements
  4. Quality Assurance: Validating syntax, logic, and best practices

Validation Framework#

The validation system ensures code quality through:

  • Syntax Checking: Validates generated code for syntactic correctness
  • Logic Analysis: Reviews code flow and potential edge cases
  • Best Practice Compliance: Ensures adherence to coding standards
  • Security Assessment: Identifies potential security vulnerabilities

Integration Options#

Aider Genius Lite demonstrates the power of agentic AI for code generation, showcasing how intelligent systems can understand, plan, and execute complex programming tasks autonomously.