Open source · AI-powered · RAG · noxtandav/hiremesh

Your talent pool, answers itself.

Hiremesh is the open-source, AI-powered talent base for recruitment agencies, hiring firms, and sourcers who keep their own pool. RAG over every resume and recruiter note, plain-English search across the lot, pipelines per job — all self-hosted on your stack.

hiremesh.local / ask
Talent base · Ask
Ask anything about the pool.
Routed to SQL aggregation, semantic retrieval, or both — automatically.
5+ years in payments Open to remote, India timezone Shortlisted but not contacted Count by stage, last 90 days
AnswerSemantic · 4 matches

Backend engineers with fintech experience based in Pune:

  • Arjun Mehta — 6 yrs, payments at Razorpay; cited from resume p.1, recruiter note 2025‑11.
  • Priya Iyer — 4 yrs, lending platform; cited from resume p.2.
  • Rohan Kulkarni — fintech consulting; cited from intake call notes.
  • Sneha Deshpande — 8 yrs, banking core; cited from resume p.1.

What it does

RAG‑powered, on the talent pool you already own — search, ask, pipeline.

Built for recruitment agencies, hiring firms, and independent sourcers running their own pool for themselves or their clients. AI + retrieval over every resume, note, and stage move — with audit trails you can defend.

Plain‑English semantic search

"Backend engineer with fintech experience in Pune." A RAG router decides between SQL aggregation and pgvector semantic retrieval — and tells you which it used.

Grounded Q&A with citations

Ask anything about a candidate or the whole pool. Answers come back retrieval-grounded in resume sections and recruiter notes — with the source attached.

Customizable Kanban pipelines

Per‑job stages you actually use. Drag, drop, move between jobs — no more sticky notes.

One pool, every client & job

For agencies running many clients, in-house teams, or solo sourcers — one canonical candidate record. Pipelines reference the pool, not copies, so every job benefits from every conversation.

AI resume parsing — PDF & DOCX

Drop a resume in. The parser turns it into structured candidate fields, with a sticky‑edit invariant: your manual fixes never get clobbered on re‑parse.

Permanent stage history

Every move, every reason, every recruiter — captured forever. Defensible audit trail without the spreadsheet archaeology.

Built for

Anyone who keeps a talent pool of their own.

Recruitment data is leverage. Hiremesh runs on infrastructure you control — no vendor lock‑in, no exfiltrated candidate pool, no per‑seat tax for asking your own database a question.

01

Recruitment agencies

Run dozens of clients off one pool. Per‑job pipelines, shared candidates, defensible audit trails.

02

In‑house hiring teams

Build a private talent base across years of inbound, referrals, and silver medalists — yours forever.

03

Independent sourcers

Solo recruiters and boutique firms who want their own AI desk without paying per seat for someone else's.

Data sovereignty

Your pool. Your data. Your house.

Your talent pool is your business. Hiremesh is self-hosted by design — every resume, every note, every embedding lives on your infrastructure. Not on a vendor's servers, not in a shared multi-tenant database, not one TOS update away from being held hostage.

Typical SaaS recruiting tool

Their cloud, their rules.

×Your candidates live on someone else's servers, in a database you can't see.
×The vendor decides retention, exports, and what "delete" actually means.
×Per-seat AI tax — every search, every Q&A, billed back to you forever.
×Stop paying and your pool — years of work — leaves with the subscription.
×A new privacy policy can change what they do with your data overnight.
Your pool is leverage. Theirs to monetize.
Hiremesh, self-hosted

Your stack, your sovereignty.

Every byte stays on your infrastructure — Postgres + pgvector you control.
You own retention, exports, deletes — it's just SQL, on your machine.
Bring your own LLM keys. Pay providers directly. No per-seat AI markup.
Yours forever. Cancel a vendor, fork the project — the pool stays put.
MIT-licensed code. Audit it, harden it, run it on an air-gapped box if you want.
Your pool. Your leverage. Your house.

Built on

A stack you'll recognize.

Boring infrastructure, exciting capabilities. Self-hosted with Docker, vector search on Postgres, and a single LLM gateway you can point at any provider you trust.

Docker‑native install

Single Compose file brings up the full stack — backend, parser worker, web UI, gateway. make up and you're live.

Postgres + pgvector

One database for both your structured candidate data and your embeddings. No separate vector store to babysit, no sync drift.

LiteLLM gateway

One unified gateway in front of every model — OpenAI, Anthropic, Bedrock, Ollama, your own. Swap providers without touching app code.

Caddy reverse proxy

Automatic HTTPS, single‑port ingress, zero config gymnastics. Production‑ready out of the box for self‑hosting.

RAG pipeline you can read

Plain‑English query → router decides between SQL aggregation and semantic retrieval → pgvector pulls the relevant resume chunks & notes → LiteLLM-routed model writes the cited answer. Every step in the repo, no black boxes.

Quickstart

Self‑host in three commands.

Hiremesh ships as a single Docker Compose stack — Postgres + pgvector, parser worker, web UI, LiteLLM gateway, and Caddy out front. Clone, set your secrets, bring it up. Full setup and architecture docs live in docs/.

~/projects — bash
# 1. Clone the repo
$ git clone https://github.com/noxtandav/hiremesh.git
$ cd hiremesh

# 2. Configure your secrets
$ cp infra/.env.example infra/.env
$ $EDITOR infra/.env   # add LLM keys, db creds

# 3. Boot the full stack
$ make up

→ Hiremesh is now live at http://localhost

What you'll get

A complete local talent base — Postgres + pgvector, parser worker, web UI, LiteLLM gateway, Caddy ingress, all wired up.

1
ClonePull the repo from GitHub.
2
ConfigureCopy infra/.env.exampleinfra/.env, add your LLM provider keys.
3
BootRun make up — Docker brings up the stack on http://localhost.
Read the full README

Stop sourcing.
Start asking your pool.

Hiremesh is in active development and built in the open. Star it, run it, fork it, shape it.