3b.2 AI project JobRadar

← 3b Projects

Complete project ecosystem

For details see

• Documentation: #606_ai_ides_.docx
• Repo 602-2_D1-tool-LLM

Demo 2 jobradar_unified.py (what it does so far 26.0506)

Single app that combines:
1. Ingest — Pulls jobs from Greenhouse (configured companies), normalizes and dedupes into MongoDB, optional LLM scoring (OpenAI primary, Anthropic fallback) with a simple cost tracker.
2. Gmail path (CLI) — With GMAIL_* set, the same script run ingests LinkedIn-style job alerts from a Gmail label over IMAP, parses, dedupes, scores, and writes into the same jobs collection.
3. API (uvicorn jobradar_unified:app) — FastAPI with Google OAuth (session + email allowlist) protecting job routes. Exposes health, jobs (list/detail), status updates, summary, /login / /me / /logout.
4. Digest (Step 7) — GET /digest/preview builds the digest text without sending. POST /digest/send emails it (SMTP to Gmail by default, or Resend via DIGEST_SENDER=resend — S7e).
5. Google Doc digest (Step 7f) — POST /digest/google-doc creates a new, timestamp-titled Google Doc in Drive with the same digest body, using OAuth tokens with Docs + drive.file scopes; optional DIGEST_DRIVE_FOLDER_ID and title prefix. Config is env-driven (Mongo, LLM keys, Gmail, digest, Resend, Drive folder, Google OAuth secrets, etc.).

Cursor generated.

26.020 chat

i have no security clearance. and they often dont state at the beginning of a job ad that it requires. what would really be nice is if an AI app could 
1) read email (already) 
2) open the link (i think a bot could do this) 
3) search for keywords like "TS", "poly", "SCI", "clearance", etc.
4) not try to auto apply for a job, but give me a clean list with (summary) and URL 
5) interesting idea... could this be the data for the ledger? 


Yes — that is a very good JobRadar use case.
It becomes:
•	Job email
  ↓
Open job link
  ↓
Extract job ad text
  ↓
Detect clearance blockers
  ↓
Summarize
  ↓
Store in ledger
  ↓
Clean report
The app should not auto-apply. It should act as a filter/analyst.
Clearance keywords:
•	clearance
security clearance
active clearance
TS
Top Secret
SCI
TS/SCI
polygraph
CI poly
full-scope poly
FS poly
Secret
DoD
IC
Output could be:
•	Job: AI Product Manager
URL: ...
Clearance status: likely required
Evidence: "active TS/SCI with polygraph required"
Recommendation: skip
And yes — this is a great ledger example.
Each job becomes an append-only ledger entry:
•	Entry 001: email received with job link
Entry 002: page fetched
Entry 003: clearance analysis
Entry 004: user decision: skip / maybe / apply
Then AI can generate the current view:
•	Jobs worth applying to:
- ...
Jobs skipped because clearance required:
- ...
Jobs needing manual review:
- ...
This is probably more useful than the Nmap demo because it directly solves one of your real problems. 

It also fits 3b perfectly: Gmail + web link + AI + ledger + clean output.

26.0620 (0515)