Job Hunt Companion

An Obsidian plugin for students and early-career job seekers: paste a job posting and get a structured note (deadline, salary, location, requirements, keywords, the posting kept verbatim), move it through a status workflow with a dated timeline, see everything on a board with what is due soon and what has gone quiet, and let a local model (via Ollama) point out which of your resume bullets answer which requirement and draft likely interview questions. Everything runs on your machine; nothing is sent to a job board, a vendor or anyone else.

• A note per application, in your vault. Jobs/<Company> - <Role>.md with properties you can query with Dataview or Bases, the posting text as pasted, the requirements as a plain list (no checkboxes, so nothing lands in your task queries), your notes, and a timeline of every status change.
• Deterministic first, model second. Dates (six common formats), salary ranges, location, remote/hybrid/on-site and contact emails are read with regular expressions and win over anything the model says. The model only fills what the regexes cannot: company, role, requirements, nice-to-haves and keywords. Its dates are re-read by the deterministic parser, never trusted as written. Without Ollama the regex pass and the board still work.
• Grounded tailoring. The model may only quote bullets that exist in your resume note; a deterministic check drops anything it cannot find and the note says how many were dropped. Rewordings are marked as suggestions. The resume is never edited.
• Interview prep without invented answers. Eight likely questions drawn from the posting's requirements and keywords, each with what the interviewer is checking, and an empty slot for your own answer.
• Free and open source, MIT licensed. No accounts, no API keys, no telemetry.

Requirements

1. Obsidian 1.7.2 or later (desktop or mobile).
2. Optional: Ollama running locally (or on another machine on your network; set its URL in settings) with a chat model: ollama pull qwen3.5:9b (default; fits a 12 GB GPU), qwen3.5:4b for smaller cards. Without Ollama, Add a posting runs only the deterministic pass (dates, salary, location, emails, first line as the title) and the board, status workflow and calendar export work as usual; Tailor and Interview prep need the model.

Usage

1. Add a posting (ribbon icon or command). Paste the posting text and, optionally, its URL. The note opens as Jobs/<Company> - <Role>.md with status wishlist. Check the frontmatter: the deadline and salary are what the regexes found, so a posting that says "end of month" will have an empty deadline for you to fill.
2. Tick the requirements you meet under ## Fit; keep your own notes under ## Notes.
3. When you apply, run Change status with the note open and pick applied: applied_on is set to today and the timeline gets a dated line. Keep going through phone screen, interview, offer, rejected or withdrawn.
4. Open the board writes Jobs/Board.md: what is due in the next 14 days, what has had no update in 21 days, then a table per status. Re-running it rewrites only that note.
5. With a posting note open, Tailor resume to this posting compares your resume note (set its path in settings; default Resume.md) with the posting and writes matches, keyword gaps and suggested rewordings under ## Tailoring. Draft interview prep writes eight questions under ## Interview prep.

Commands

Note shape

---
company: Acme Analytics
role: "Software Engineer, Data Platform"
status: wishlist
applied_on: ""
deadline: 2026-11-30
location: "Denver, CO"
remote: hybrid
salary: $85k–$100k per year
url: "https://…"
source: posting
keywords:
  - python
  - airflow
---

# Acme Analytics — Software Engineer, Data Platform

## Posting
```text
(the pasted text, verbatim, in a fence so its headings, #tags and links stay inert)
```

## Requirements
- 2+ years writing production Python
…

## Nice to have
## Fit
- 2+ years writing production Python
## Notes
## Timeline
- 2026-11-07 — added
- 2026-11-12 — applied

## Tailoring and ## Interview prep are added (or replaced) by their commands. Everything else in the note is yours; the plugin edits only the frontmatter key it changes and the section it owns.

How it works

Deterministic pass. Dates are read in 2026-11-30, 11/30/2026 (day first when the first number cannot be a month), Nov 30, November 30, 2026, 30 November 2026 and 30th Nov forms; month-day without a year takes the next occurrence on or after today. Only lines with deadline wording ("deadline", "apply by", "applications close", "closing date", …) are considered, lines about start or posting dates are skipped, and vague wording ("end of month", "rolling") yields no deadline. Salaries need a currency symbol or code and an amount of at least 1,000 ($85k–$100k, £40,000, 85,000 - 100,000 USD). Location comes from a Location: label or a City, ST pair; remote/hybrid/on-site from the wording.

Model pass. One chat call (temperature 0.1, thinking off, JSON mode) with a fixed schema: company, role, location, remote, deadline (copied as written), salary, requirements (≤ 8), nice_to_have (≤ 5), keywords (≤ 12). The reply is parsed as JSON, every field is type-checked, lists are capped and de-duplicated, the deadline text is run through the same date parser, and the regex results win wherever both have a value. The first line of the posting is used as the role only when the model gives none.

Board. Reads every posting note under the jobs folder (frontmatter with company, role, status). "Last update" is the latest dated Timeline line. Due = active status, deadline within 14 days; stale = active status, no update for more than 21 days. Rejected and withdrawn postings are listed in their tables but never flagged. The .ics export (optional) writes an all-day event per upcoming deadline with a UID derived from the note path, so re-exports update rather than duplicate.

Tailoring. The resume's bullet lines (- , * , 1. ) and the posting's requirements and keywords go to the model (temperature 0.3, thinking off, JSON mode). Every quoted bullet is normalised (case, punctuation, markdown emphasis) and must appear in the normalised resume text; otherwise it is dropped and counted. At most 5 gaps and 3 rewordings are kept.

Settings worth knowing

• Jobs folder (default Jobs) and Resume note (default Resume.md).
• Export deadlines as a calendar file and its path (default Jobs/deadlines.ics).
• Ollama URL, Chat model, Context window.

Privacy and network use

The plugin makes network requests only to the Ollama URL you set (localhost by default), and only when you run Add a posting (with Ollama reachable), Tailor or Interview prep. It does not talk to job boards, does not fetch the posting URL you paste, and sends nothing anywhere else. It reads the notes in the jobs folder and your resume note, and writes under the jobs folder. No telemetry, no accounts.

Development

npm install
npm run dev           # watch build → main.js
npm test              # unit tests (vitest), no Ollama needed
npm run build         # type-check + Obsidian review lint + production bundle
npm run posting:live  # extraction prompt against a running Ollama on two sample postings (2 calls)

To release: bump version in manifest.json and package.json, add the version to versions.json, write release-notes/<version>.md, commit, then push a tag named exactly <version>. The release workflow builds, tests, attests and publishes the GitHub release.

src/core/ (extraction, posting schema and merge, note rendering and parsing, sections, status, board, ICS, tailoring, interview prep, Ollama client) has no Obsidian dependency and is tested against an in-memory vault; src/main.ts, the modals and the settings tab are the thin Obsidian layer.

Job Hunt Companion is part of the Companion suite of local-first Obsidian plugins by Tony Peonio for students and researchers: your machine does the work, every change is a diff you approve, and the numbers are published. The others are Local Vault Agent (local Ollama agent that searches, compiles, edits notes), Lecture Companion (live local lecture transcription, Canvas slides, Ollama summaries), Canvas Companion (Canvas LMS assignments, grades and announcements as notes), Exam Companion (day-by-day study plan, daily quizzes and flashcards), Paper Companion (Zotero search by meaning, page-cited reading notes), Research Log Companion (append-only lab notebook with git, parameters, data hashes), Fellowship Companion (grants and fellowships: deadlines, materials, letters, eligibility), Code Companion (commits in the daily note, weekly report, TODOs), Teaching Companion (local-model rubric grading with verbatim evidence) and Branching Table of Contents (parent/child properties as a rooted family tree). All at github.com/TonyPeonio.

License

MIT