Praxis

---

Praxis is a localized, multi-agent pipeline designed to solve the critical failures of using standard LLMs (like ChatGPT) to write resumes. By employing a strict Zero-Loss Relational Database, Adversarial Drafting Loops, and Automated Deep Data Harvesting, Praxis generates highly targeted, mathematically honest, and ATS-optimized career documents.

🧠 The Science: Why Praxis?

Using a single-shot prompt to ask an LLM to "write my resume" results in three catastrophic failures:

1. Summarization Loss: LLMs inherently compress facts, stripping away the exact metrics, technologies, and scale that actually get you hired.
2. Sycophancy & Hallucination: LLMs invent "AI-speak" (e.g., "Spearheaded synergistic paradigms") and hallucinate responsibilities to make you sound good, failing rigorous technical interviews.
3. The Blank Page Problem: Most professionals severely underestimate their own impact, scale, and day-to-day metrics.

Praxis solves this through a multi-agent architectural pipeline:

• Zero-Loss Relational Cataloging: Instead of summarizing your history into Markdown, Praxis iteratively ingests raw data (PDFs, GitHub .tar.gz exports, LinkedIn CSVs) into a strict .praxis/data/knowledge_base.json. This acts as a permanent, loss-proof relational database of your skills, mapping every tool to the exact metric and project where it was used.
• The Expert Inquisitor: Praxis analyzes your ingested data for weak points (missing metrics, vague responsibilities). It then conducts a multi-step, structured interview wizard to extract exact numbers and scale, enriching the database with authoritative context.
• Adversarial Drafting Loop (Pathos vs. Logos): When tailoring a resume for a specific job description (/praxis customize <url>), Praxis employs a two-agent adversarial loop.
  • praxis-pathos (The Visionary) drafts the resume using your saved Voice Profile and the STAR method.
  • praxis-logos (The Truth-Teller) acts as a brutal auditor, rejecting any bullet point that hallucinates facts or uses AI-speak not explicitly backed by the knowledge_base.json. They iterate until a mathematically honest, perfectly targeted document is produced.

---

🏗️ Architecture & Commands

Praxis installs directly into your local CLI environment (e.g., opencode or strata) as a skill.

1. The Intake Engine: /praxis

• What it does: Runs a deterministic Deep Harvest extraction script (ingest.py) across your root directory for raw exports (PDFs, TXTs), ignores system files, and syncs your public GitHub repositories (github_sync.py).
• How it works: It losslessly parses this data into string[] fact pools, pushes it into .praxis/data/knowledge_base.json, and then moves the raw files to a secure .praxis/sources/ directory. It then automatically drafts ATS-optimized baseline profiles (assets/Resume.md, assets/LinkedIn_Profile.md) utilizing a "Discrete Chronological Strategy".

2. The Fact Logger: /praxis history <fact>

• What it does: Quickly appends a specific accomplishment, metric, or bullet point to an existing role without requiring a full CV re-upload.
• How it works: You run /praxis history at ACME co., I managed a team of 50. Praxis parses the company, locates it in knowledge_base.json, and appends the fact to that role's bullets array (the "fact pool"), automatically regenerating your baseline resumes.

3. The Skill Enricher: /praxis skill <skill_name> <description>

• What it does: Replaces generic, scraped LinkedIn skills (e.g., "Kubernetes") with concrete contextual evidence of how you used that skill.
• How it works: You run /praxis skill Kubernetes Architected multi-region clusters.... Praxis pushes this rich context into the RelationalSkillsDatabase array for that skill, giving the Drafter agent highly specific material to work with.

4. The Forge: /praxis gen <job-url>

• What it does: Generates a highly tailored PDF resume specifically designed to pass the ATS for a target job.
• How it works: It runs a Skill Gap Analysis between the Job Description and your knowledge_base.json. It triggers the Pathos/Logos adversarial loop to strategically select the 3-4 most relevant facts from your "fact pools" (rather than dumping your whole history). It then generates a targeted PDF (e.g., assets/Company_User_Resume.pdf) and cleans up the temporary files.

---

🚀 Installation

# Clone the repository
git clone [email protected]:ksmeltzer/Praxis.git
cd Praxis

# Install the agents, skills, and commands to your local environment
chmod +x install.sh
./install.sh

🔒 Privacy & Security

Praxis is designed with absolute privacy in mind. Your raw data, API keys, and generated JSON databases are intentionally .gitignore'd. Your career data never leaves your local machine unless you explicitly configure an external model API.

---