#!/usr/bin/env python3 """ tailor.py - tailor a resume to a job description using several models via OpenRouter, then have a model critique the result as a skeptical hiring manager. Setup (once): Put your full career history in resume.md (the more detail the better) Put the job description in job.md Optional: put a resume to use as a structural template in template.md Usage: python3 tailor.py # tailor with default models, then critique python3 tailor.py --critique-only # just critique resume.md against job.md python3 tailor.py -m openai/gpt-4o-mini -m anthropic/claude-sonnet-4 Outputs land in drafts/ with a timestamp, one file per model plus a critique. Requires OPENROUTER_API_KEY in your environment. Standard library only. """ import argparse import json import os import sys import time import urllib.request import urllib.error from concurrent.futures import ThreadPoolExecutor, as_completed from datetime import datetime from pathlib import Path API_URL = "https://openrouter.ai/api/v1/chat/completions" # Claude Code: verify these against https://openrouter.ai/api/v1/models and update. DEFAULT_MODELS = [ "openai/gpt-4o-mini", "meta-llama/llama-3.3-70b-instruct", "anthropic/claude-haiku-4.5", "google/gemini-2.5-flash", ] CRITIC_MODEL = "openai/gpt-4o-mini" TAILOR_PROMPT = """You are an expert resume writer for tech and startup roles. Below is a candidate's full career history, a job description, and an existing resume to use as a TEMPLATE. Write a tailored resume in Markdown for this role. Rules: - Follow the TEMPLATE's section order, headings, formatting, and voice exactly. Change the content, not the structure. - Use only facts present in the career history or template. Never invent metrics, titles, or dates. Respect any "Do Not Claim" notes in the career history. - Rewrite the summary and reorder or reword bullets so the most relevant experience for THIS job leads. Cut or shorten what isn't relevant. - Mirror the language and priorities of the job description where it's honest to do so. - Keep it to roughly one page. Plain, direct, verb-led. No em dashes. No buzzwords. - Output only the resume, no commentary. === CAREER HISTORY === {resume} === JOB DESCRIPTION === {job} === TEMPLATE === {template} """ CRITIQUE_PROMPT = """You are a skeptical hiring manager at the company in the job description below. You have 90 seconds with this resume. Be direct and specific. Give: 1. First impression in one sentence. 2. The three strongest signals for THIS role. 3. The three biggest gaps or doubts you'd raise, and how the candidate could address each in a cover letter or interview. 4. Any line that sounds generic, inflated, or unclear, with a suggested rewrite. 5. A one-line pitch the candidate should use to open their cover letter. === JOB DESCRIPTION === {job} === RESUME === {resume} """ def call(model: str, prompt: str, api_key: str, max_tokens: int = 2000) -> dict: body = json.dumps({ "model": model, "messages": [{"role": "user", "content": prompt}], "max_tokens": max_tokens, "usage": {"include": True}, }).encode() req = urllib.request.Request( API_URL, data=body, headers={ "Authorization": f"Bearer {api_key}", "Content-Type": "application/json", "X-Title": "tailor-cli", }, ) start = time.perf_counter() try: with urllib.request.urlopen(req, timeout=180) as resp: data = json.loads(resp.read()) except urllib.error.HTTPError as e: return {"model": model, "error": f"HTTP {e.code}: {e.read().decode()[:200]}"} except Exception as e: # noqa: BLE001 return {"model": model, "error": str(e)} return { "model": model, "text": data["choices"][0]["message"]["content"].strip(), "latency": time.perf_counter() - start, "cost": data.get("usage", {}).get("cost"), } def read(path: str) -> str: p = Path(path) if not p.exists(): sys.exit(f"Missing {path}. Create it first (see the top of this file).") return p.read_text().strip() def slug(model: str) -> str: return model.replace("/", "_").replace(":", "_") def main(): parser = argparse.ArgumentParser(description="Tailor and critique a resume via OpenRouter.") parser.add_argument("-m", "--model", action="append", dest="models") parser.add_argument("--resume", default="resume.md") parser.add_argument("--job", default="job.md") parser.add_argument("--template", default="template.md") parser.add_argument("--critic", default=CRITIC_MODEL) parser.add_argument("--critique-only", action="store_true") args = parser.parse_args() api_key = os.environ.get("OPENROUTER_API_KEY") if not api_key: sys.exit("Set OPENROUTER_API_KEY first.") resume = read(args.resume) job = read(args.job) template = Path(args.template).read_text().strip() if Path(args.template).exists() else resume out = Path("drafts") / datetime.now().strftime("%Y-%m-%d_%H%M") out.mkdir(parents=True, exist_ok=True) if not args.critique_only: models = args.models or DEFAULT_MODELS print(f"Tailoring with {len(models)} models...\n") prompt = TAILOR_PROMPT.format(resume=resume, job=job, template=template) results = [] with ThreadPoolExecutor(max_workers=len(models)) as pool: futs = {pool.submit(call, m, prompt, api_key): m for m in models} for f in as_completed(futs): results.append(f.result()) for r in results: if "error" in r: print(f" {r['model']}: ERROR {r['error']}") continue path = out / f"draft_{slug(r['model'])}.md" path.write_text(r["text"]) cost = "n/a" if r["cost"] is None else f"${r['cost']:.4f}" print(f" {r['model']}: {r['latency']:.1f}s, {cost} -> {path}") print() print(f"Critiquing with {args.critic}...\n") crit = call(args.critic, CRITIQUE_PROMPT.format(resume=resume, job=job), api_key) if "error" in crit: sys.exit(f"Critique failed: {crit['error']}") (out / "critique.md").write_text(crit["text"]) print(crit["text"]) print(f"\nSaved everything to {out}/") if __name__ == "__main__": main()