Knesset Helper

Research assistant for Israeli Knesset (parliament) data. Ask questions in Hebrew or English about bills, votes, and committee discussions.

python3 -m agent.run "What education bills were proposed in the 25th Knesset?"
python3 -m agent.run "מה נאמר בוועדות הכנסת על חדשנות טכנולוגית בחברה הערבית?"

Note

Not affiliated with the Knesset or any Israeli government body. Data is sourced from the Knesset's publicly available OData API. Answers are AI-generated and should not be relied upon for legal, political, or official purposes.

Setup

Prerequisites: Python 3.10+, Docker, OpenAI API key

# Install
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env  # add your OPENAI_API_KEY

# Start OpenSearch and restore pre-built index
docker compose up -d
make setup

make setup downloads a pre-built snapshot from HuggingFace with all indexed protocols — no embedding costs.

To ingest from scratch instead:

PYTHONPATH=. python3 -m ingest.opensearch_setup
PYTHONPATH=. python3 -m ingest.ingest --knesset-num 25 --limit 20

Usage

CLI

PYTHONPATH=. python3 -m agent.run "your question here"

Web UI

PYTHONPATH=. python -m ui.app
# http://127.0.0.1:7860

Two tabs: Chat (agent Q&A with streaming progress) and Search (direct OpenSearch lookup). Supports Hebrew/English with RTL layout switching.

MCP Server (Claude Desktop)

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "knesset-helper": {
      "command": "/path/to/knesset-helper/.venv/bin/python3",
      "args": ["-m", "mcp_server.server"],
      "env": {
        "PYTHONPATH": "/path/to/knesset-helper"
      }
    }
  }
}

Exposes search_bills, get_bill_details, get_bill_votes, and search_protocols as tools.

How it works

A LangGraph agent runs an adaptive research loop:

Planner → Researcher → Judge ─── need more? → back to Planner (up to 3x)
                          │
                     sufficient
                          ↓
                     Synthesizer → Answer

The Planner breaks questions into research tasks. The Researcher executes them against the Knesset OData API and OpenSearch. The Judge filters irrelevant results and decides if more data is needed. The Synthesizer produces a cited answer.

Committee protocols are indexed into OpenSearch using hybrid search (kNN vectors + BM25 keywords) for retrieval-augmented generation.

Data source

The Knesset OData v3 API provides open access to parliamentary data — bills, votes, MKs, committees, and research documents. All content is in Hebrew.

License

MIT