AGENT_KNOWLEDGE_BASE.md

Agent Knowledge Base System

Overview

The AI Agent has been enhanced with a knowledge base system that allows it to understand the codebase structure and intelligently handle natural language commands. This enables users to give high-level commands like "Update the Ford resume with this experience" without needing to specify exact file paths or IDs.

Architecture

Knowledge Base File

Location: agent_knowledge_base.json

This JSON file contains:

• Data structure mappings - How resumes, job listings, and other data are organized
• Common operations - Step-by-step guides for frequent tasks
• API endpoints - Available REST API endpoints
• Command patterns - Natural language patterns the agent can recognize
• File locations - Key directories and files in the codebase

Enhanced Agent System Prompt

The agent's system prompt (agent.py) has been updated to:

• Reference the knowledge base structure
• Understand resume naming conventions
• Parse natural language commands
• Map company names to resume files
• Execute appropriate helper scripts

Key Features

1. Natural Language Command Understanding

The agent can parse commands like:

"Update the Ford resume with this experience: data/job_listings/Tailored Experience Summary for Ford.md"

And automatically:

1. Extract the company identifier ("Ford")
2. Search data/resumes/index.json for matching resume
3. Find the resume ID
4. Execute the update operation

2. Resume Discovery

The agent knows how to find resumes by:

• Company name: "Ford", "GM", "Credibly"
• Partial name match: Case-insensitive search
• Resume ID: Direct UUID lookup

Resume naming pattern: {FirstName}_{LastName}_{Role}_{Company}

Example: Sidney_Jones_Senior_Software_Engineer_Ford

3. Helper Scripts

src/update_resume_experience.py

A Python script that automates resume updates:

python src/update_resume_experience.py --resume "Ford" --experience "data/job_listings/Tailored Experience Summary for Ford.md"

Features:

• Finds resume by company name or ID
• Parses experience from markdown files
• Updates resume JSON automatically
• Updates timestamps in index.json
• Supports prepend (default) or replace mode

Options:

• --resume: Company name or resume name to search for
• --resume-id: Direct UUID (alternative to --resume)
• --experience: Path to markdown file with experience
• --replace: Replace all experience (default: prepend)
• --data-dir: Data directory path (default: data)

CRUD Scripts (src/crud/)

Specialized scripts for managing individual resume sections:

Available Scripts:

• basic_info.py - Name, title, location, contact information
• summary.py - Resume summary text
• technical_skills.py - Technical proficiencies by category
• expertise.py - Areas of expertise
• achievements.py - Achievements
• education.py - Education entries
• certifications.py - Certification entries
• experience.py - Work experience and bullets

Example Usage:

# Update title
python src/crud/basic_info.py --resume "Master Resume" --update-title "Principal Architect"

# Add technical skill
python src/crud/technical_skills.py --resume "Ford" --append-to-category "languages" "Python"

# Add certification
python src/crud/certifications.py --resume "Ford" --add --name "AWS Solutions Architect" --issuer "Amazon" --date "Nov 2025"

# List expertise areas
python src/crud/expertise.py --resume "Master Resume" --list

Common Features:

• Natural language friendly (find by resume name)
• Automatic timestamp updates
• Built-in validation
• Consistent error handling
• Comprehensive help (--help)

See CRUD Operations Documentation for complete guide.

Data Structure

Resumes

Location: data/resumes/

Index File: data/resumes/index.json

{
  "resumes": [
    {
      "id": "d474d761-18f2-48ab-99b5-9f30c54f75b2",
      "name": "Sidney_Jones_Senior_Software_Engineer_Ford",
      "created_at": "2025-10-12T08:49:00.020000",
      "updated_at": "2025-10-12T08:51:39.985858",
      "job_listing_id": null,
      "is_master": false,
      "description": "Duplicated from Sidney_Jones_Senior_Software_Engineer_GM"
    }
  ]
}

Resume Files: data/resumes/{UUID}.json

Job Listings

Location: data/job_listings/

File Types:

• JSON files: {UUID}.json - Structured job listing data
• Markdown files:
  • Job descriptions: Sr. Software Engineer - at Credibly.md
  • Tailored experiences: Tailored Experience Summary for Ford.md

Common Operations

Update Resume with Experience

User Command:

Update the Ford resume with this experience: "data/job_listings/Tailored Experience Summary for Ford.md"

Agent Execution:

run: python src/update_resume_experience.py --resume "Ford" --experience "data/job_listings/Tailored Experience Summary for Ford.md"

What Happens:

1. Script searches data/resumes/index.json for resume containing "Ford"
2. Finds: Sidney_Jones_Senior_Software_Engineer_Ford (ID: d474d761-18f2-48ab-99b5-9f30c54f75b2)
3. Parses experience sections from markdown file
4. Updates data/resumes/d474d761-18f2-48ab-99b5-9f30c54f75b2.json
5. Updates timestamp in index.json

List Available Resumes

User Command:

List all resumes

Agent Execution:

run: cat data/resumes/index.json

Or via API:

GET http://localhost:5000/api/resumes

CRUD Operations (Granular Updates)

Update Basic Information

User Command:

Update my title to Principal Software Architect

Agent Execution:

run: python src/crud/basic_info.py --resume "Master Resume" --update-title "Principal Software Architect"

Add Technical Skills

User Command:

Add Python and Rust to my programming languages

Agent Execution:

run: python src/crud/technical_skills.py --resume "Master Resume" --append-to-category "languages" "Python, Rust"

Add Certification

User Command:

Add my AWS certification

Agent Execution:

run: python src/crud/certifications.py --resume "Master Resume" --add --name "AWS Solutions Architect" --issuer "Amazon" --date "November 2025"

List Expertise Areas

User Command:

List my areas of expertise

Agent Execution:

run: python src/crud/expertise.py --resume "Master Resume" --list

Add Experience Bullet

User Command:

Add a bullet to my Microsoft experience about leading a team

Agent Execution:

run: python src/crud/experience.py --resume "Master Resume" --add-bullet --employer "Microsoft" --text "Led team of 10 engineers in developing cloud-native solutions" --tags "Leadership,Cloud"

Tailor Resume to Job

User Command:

Tailor my resume for the Credibly position

Agent Execution:

run: python src/tailor.py --resume data/master_resume.json --jd "data/job_listings/Sr. Software Engineer - at Credibly.md" --out out/credibly_tailored.html --format html --theme professional

Command Whitelisting

For security, the agent can only execute whitelisted commands:

• python src/tailor.py - Resume tailoring
• python src/update_resume_experience.py - Resume updates
• python -m pytest - Running tests
• python -m json.tool - JSON formatting
• git status, git log, git diff - Git operations
• ls, dir, pwd, echo - File system navigation
• cat, type - File reading

Dangerous commands are blocked (e.g., rm -rf, format, sudo).

API Integration

The agent can also use REST API endpoints:

Agent Endpoints

• POST /api/agent/chat - Send messages to agent
• GET /api/agent/memory - Get conversation history
• POST /api/agent/memory/clear - Clear memory
• POST /api/agent/validate-command - Validate command security

Resume Endpoints

• GET /api/resumes - List all resumes
• GET /api/resumes/{id} - Get specific resume
• POST /api/resumes - Create new resume
• PUT /api/resumes/{id} - Update resume
• POST /api/resumes/{id}/duplicate - Duplicate resume
• POST /api/resumes/{id}/tailor - Tailor resume

Job Listing Endpoints

• GET /api/job-listings - List all job listings
• GET /api/job-listings/{id} - Get specific job listing
• POST /api/job-listings - Create new job listing

Example Interactions

Example 1: Update Resume

User:

Update the Ford resume with this experience: "C:\source\repos\bpm\internal\agentic-resume-tailor\data\job_listings\Tailored Experience Summary for Ford.md"

Agent Response:

I'll update the Ford resume with the experience from that file.

run: python src/update_resume_experience.py --resume "Ford" --experience "data/job_listings/Tailored Experience Summary for Ford.md"

✅ Successfully updated resume d474d761-18f2-48ab-99b5-9f30c54f75b2
   Found resume: Sidney_Jones_Senior_Software_Engineer_Ford
   Added 5 experience entries

Example 2: List Resumes

User:

What resumes do I have?

Agent Response:

Let me check your available resumes.

run: cat data/resumes/index.json

You have the following resumes:
1. Master Resume (ID: b3013a2a-349a-4080-802c-3bbe6714b576)
2. Sidney_Jones_Senior_Software_Engineer_Credibly (ID: a041bd2e-d54b-488f-adda-e4c707d5332d)
3. Sidney_Jones_Senior_Software_Engineer_GM (ID: f3bc0d6d-aea7-4e08-96b8-5fc6b581138b)
4. Sidney_Jones_Senior_Software_Engineer_Ford (ID: d474d761-18f2-48ab-99b5-9f30c54f75b2)

Example 3: Tailor Resume

User:

Tailor my resume for the GM position

Agent Response:

I'll tailor your resume for the GM position. Let me use the master resume and the GM job description.

run: python src/tailor.py --resume data/master_resume.json --jd "data/job_listings/Tailored Experience Summary for GM.md" --out out/gm_tailored.html --format html --theme professional

✅ Resume tailored successfully!
   Output: out/gm_tailored.html

Testing

Tests for the knowledge base system are in:

• tests/test_update_resume_experience.py - Tests for the helper script

Run tests:

python -m pytest tests/test_update_resume_experience.py -v

Future Enhancements

Potential improvements to the knowledge base system:

1. Semantic Search - Use embeddings to find similar resumes/experiences
2. Auto-tagging - Automatically tag experiences with relevant skills
3. Resume Comparison - Compare different resume versions
4. Job Matching - Suggest which resume to use for a job posting
5. Experience Library - Maintain a library of reusable experience bullets
6. Version Control - Track changes to resumes over time
7. Analytics - Track which resumes get the most responses

Troubleshooting

Agent Can't Find Resume

Problem: Agent says "No resume found matching 'Ford'"

Solution:

1. Check data/resumes/index.json to see available resumes
2. Verify the company name is in the resume name
3. Try using the full resume name or ID

Experience Not Parsing Correctly

Problem: Experience from markdown file isn't being added

Solution:

1. Verify markdown format uses ### Company — Role (Dates) headers
2. Check that bullets start with *
3. Ensure tags line uses **Tags:** format

Command Not Whitelisted

Problem: Agent says command is not allowed

Solution:

1. Check src/api/app.py for ALLOWED_COMMAND_PREFIXES
2. Use one of the helper scripts instead of direct commands
3. Request the command be added to the whitelist

Contributing

To extend the knowledge base:

1. Update agent_knowledge_base.json with new operations
2. Update agent system prompt in agent.py
3. Add new helper scripts to src/
4. Update command whitelist in src/api/app.py
5. Add tests for new functionality
6. Update this documentation