尚未检测到更新。GitHub 发布版本后会显示在这里。
按最近已完成的 8 小时观测回合间的 GitHub 星标增量排序
个人 GitHub 账号下公开仓库的所有者可以登录并验证所有权，自行管理介绍，无需管理员批准。
登录并生成验证文件，在 30 分钟内将其提交到此仓库的默认分支。其他分支或分叉仓库中的文件无效。然后在下方检查文件，即可无需管理员批准直接管理项目。请保留文件以维持管理权限。
如需登录，将前往登录页面。登录后请回到此处，再次点击所选的验证按钮。
在项目文件夹中运行 /nomorevibe verify，即可验证所有权并自行管理介绍。
一分钟完成验证Backend for esbvaktin.is — independent civic information platform for Iceland's EU referendum
翻译准备中
取自仓库 README
正在加载完整 README。 以下为已保存的摘录。
Backend for esbvaktin.is — an independent, data-driven civic information platform for Iceland's EU membership referendum (29 August 2026).
ESB Vaktin monitors Icelandic public discourse about EU membership, extracts factual claims, and assesses them against a curated evidence database. The platform tracks claims equally regardless of whether they are pro-EU or anti-EU.查看完整 README ↗Backend for esbvaktin.is — an independent, data-driven civic information platform for Iceland's EU membership referendum (29 August 2026).
ESB Vaktin monitors Icelandic public discourse about EU membership, extracts factual claims, and assesses them against a curated evidence database. The platform tracks claims equally regardless of whether they are pro-EU or anti-EU.
Pipeline: articles and speeches are scanned → factual claims extracted → matched against 390+ evidence entries → assessed on a five-point scale (supported, partially supported, unsupported, misleading, unverifiable) → published with full source transparency.
All AI prompts are published. The agent definitions in .claude/agents/ contain the exact instructions given to each AI model. The context preparation code in src/esbvaktin/pipeline/prepare_context.py shows how articles, evidence, and instructions are assembled. The evidence database is committed in data/seeds/.
See METHODOLOGY.md for the full methodological description, or Aðferðafræði on the site.
Python 3.12+ · uv · PostgreSQL 17 + pgvector · BAAI/bge-m3 embeddings · Claude (Anthropic) · GreynirCorrect · 11ty
git clone https://github.com/bgautijonsson/esbvaktin.git
cd esbvaktin
cp .env.example .env # DATABASE_URL defaults work for local Docker
docker compose up -d # Start PostgreSQL 17 + pgvector
uv sync --extra dev # Install dependencies + dev tools
uv run python scripts/init_db.py --seed # Create schema + seed evidence
| Extra | What it adds | When you need it |
|---|---|---|
| (none) | Core: trafilatura, psycopg, httpx, pydantic | Always |
dev | pytest, pytest-asyncio, ruff | Running tests and linting |
icelandic | GreynirCorrect, icegrams, islenska | Icelandic text generation/correction |
embeddings | FlagEmbedding, torch, transformers | Semantic search, evidence matching (~2GB model) |
email | requests | Mailgun email integration |
ghost | pyjwt | Ghost CMS publishing |
The speech analysis pipeline requires althingi.db from the althingi-mcp project. Set ALTHINGI_DB_PATH=/path/to/althingi.db in .env. The core article pipeline works without it.
Export scripts generate JSON for the esbvaktin-site frontend. They default to ../esbvaktin-site as sibling directory — override with --site-dir /path/to/site.
uv run --extra dev python -m pytest # All tests
uv run --extra dev ruff check src/ scripts/ # Lint
See CONTRIBUTING.md for setup details, dependency guide, and contribution guidelines.
Code is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0). Content and data in data/seeds/ are licensed under CC BY-SA 4.0.