da3f5faa91
FastAPI + React + PostgreSQL salary benchmarking tool with AI research pipeline. - Seed data for 25+ New York roles (junior/mid/senior levels) - Single + bulk lookup with location alias mapping (NYC -> New York, etc.) - Research pipeline: Serper -> Firecrawl -> Cohere Rerank -> Claude analysis - Editable validation UI for AI-proposed benchmarks - CSV export, Montserrat font, black/white/#FFC407 design - Fully Dockerized (app + db + frontend) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
54 lines
1.7 KiB
Python
54 lines
1.7 KiB
Python
import json
|
|
|
|
import anthropic
|
|
|
|
from app.config import settings
|
|
|
|
|
|
async def analyze_salary_data(
|
|
title: str, location: str, ranked_content: list[dict]
|
|
) -> dict:
|
|
sources_text = "\n\n---\n\n".join(
|
|
item["content"] for item in ranked_content if item.get("content")
|
|
)
|
|
|
|
prompt = f"""You are a compensation analyst. Based on the following salary data sources
|
|
for the role "{title}" in "{location}", produce a structured benchmark.
|
|
|
|
[SOURCES]
|
|
{sources_text}
|
|
[/SOURCES]
|
|
|
|
Return ONLY valid JSON in this exact format:
|
|
{{
|
|
"benchmarks": [
|
|
{{"level": "junior", "salary": <int>}},
|
|
{{"level": "mid", "salary": <int>}},
|
|
{{"level": "senior", "salary": <int>}}
|
|
],
|
|
"confidence_score": <float 0.0-1.0>,
|
|
"reasoning": "<Brief explanation of how you derived these numbers>",
|
|
"sources_used": ["<relevant source descriptions>"]
|
|
}}
|
|
|
|
Rules:
|
|
- Each salary value is a single annual USD integer representing the typical/median salary for that level
|
|
- Base your estimates on the provided data, not general knowledge
|
|
- If data is sparse, lower the confidence_score accordingly
|
|
- Return ONLY the JSON object, no markdown or explanation"""
|
|
|
|
client = anthropic.AsyncAnthropic(api_key=settings.anthropic_api_key)
|
|
message = await client.messages.create(
|
|
model="claude-sonnet-4-20250514",
|
|
max_tokens=1024,
|
|
messages=[{"role": "user", "content": prompt}],
|
|
)
|
|
|
|
response_text = message.content[0].text.strip()
|
|
# Strip markdown code fences if present
|
|
if response_text.startswith("```"):
|
|
response_text = response_text.split("\n", 1)[1]
|
|
if response_text.endswith("```"):
|
|
response_text = response_text[:-3].strip()
|
|
|
|
return json.loads(response_text)
|