feat: add seed/query scripts, Dockerfile; fix vector dim 1024→768

Dockerfile

@@ -0,0 +1,50 @@
+# --- Stage 1: Build ---
+FROM python:3.12-slim AS builder
+
+ENV PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1 \
+    PYTHONPATH=/app/src \
+    PORT=8000
+
+WORKDIR /app
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    build-essential \
+    libpq-dev \
+    git \
+    && rm -rf /var/lib/apt/lists/*
+
+# Clone SDK sibling repo into WORKDIR so ../chatbot-plugin-sdk resolves
+RUN git clone https://github.com/Teng91/chatbot-plugin-sdk.git /chatbot-plugin-sdk
+
+COPY --from=ghcr.io/astral-sh/uv:0.10.12 /uv /usr/local/bin/uv
+
+COPY pyproject.toml uv.lock ./
+RUN uv sync --frozen --no-dev
+
+# --- Stage 2: Run ---
+FROM python:3.12-slim
+
+ENV PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1 \
+    PYTHONPATH=/app/src \
+    PORT=8000
+
+WORKDIR /app
+
+COPY --from=builder /app/.venv /app/.venv
+
+# Editable install needs the source checked out at the sibling path
+COPY --from=builder /chatbot-plugin-sdk /chatbot-plugin-sdk
+
+COPY src/ ./src/
+COPY alembic/ ./alembic/
+COPY alembic.ini .
+
+RUN addgroup --system appuser && adduser --system --group appuser
+RUN chown -R appuser:appuser /app
+USER appuser
+
+EXPOSE 8000
+
+CMD ["sh", "-c", ".venv/bin/uvicorn chatbot_plugin.main:app --host 0.0.0.0 --port ${PORT:-8000}"]

alembic/versions/002_toolbox_schema.py

@@ -57,7 +57,7 @@ def upgrade() -> None:
     )
 
     # Cast dense_vector to proper vector type
-    op.execute("ALTER TABLE article_chunks ALTER COLUMN dense_vector TYPE vector(1024) USING dense_vector::vector")
+    op.execute("ALTER TABLE article_chunks ALTER COLUMN dense_vector TYPE vector(768) USING dense_vector::vector")
     # Cast sparse_vector to proper sparsevec type
     op.execute("ALTER TABLE article_chunks ALTER COLUMN sparse_vector TYPE sparsevec(250002) USING sparse_vector::sparsevec")
 

scripts/query.py

@@ -0,0 +1,38 @@
+"""Minimal query: search DB directly with a fixed query vector."""
+import asyncio
+from sqlalchemy import select, text
+from sqlalchemy.ext.asyncio import create_async_engine
+
+ENGINE = create_async_engine(
+    "postgresql+asyncpg://postgres:postgres@localhost:5432/chatbot_plugin",
+    echo=False,
+)
+QUERY_VEC = [1.0] + [0.0] * 767  # should match v0 most closely
+
+async def main():
+    from chatbot_plugin_sdk.models import Article, ArticleChunk
+
+    async with ENGINE.begin() as conn:
+        await conn.execute(text("CREATE EXTENSION IF NOT EXISTS vector"))
+
+    async with ENGINE.connect() as conn:
+        stmt = (
+            select(
+                ArticleChunk.id,
+                ArticleChunk.chunk_index,
+                ArticleChunk.content,
+                Article.title,
+                Article.url,
+                ArticleChunk.dense_vector.cosine_distance(QUERY_VEC).label("dist"),
+            )
+            .join(Article, ArticleChunk.article_id == Article.id)
+            .order_by(ArticleChunk.dense_vector.cosine_distance(QUERY_VEC))
+            .limit(3)
+        )
+        rows = (await conn.execute(stmt)).all()
+
+        print("=== query vector = [1.0, 0, 0, ...] ===")
+        for r in rows:
+            print(f"  chunk {r.chunk_index}: dist={r.dist:.4f} | {r.content[:40]}")
+
+asyncio.run(main())

scripts/seed.py

@@ -0,0 +1,32 @@
+"""Minimal seed: insert test article + chunks directly via SDK."""
+import asyncio
+import uuid
+from chatbot_plugin_sdk import RagArticleProcessor
+
+article_id = uuid.UUID("11111111-1111-1111-1111-111111111111")
+
+# 768-dim vectors with intentional similarity pattern
+v0 = [1.0] + [0.0] * 767   # article chunk 0
+v1 = [0.9] + [0.1] * 767   # article chunk 1
+v2 = [0.0] * 768           # noise chunk
+
+async def main():
+    p = RagArticleProcessor()
+    p.configure(dbname="chatbot_plugin", user="postgres", password="postgres")
+    await p._save_article_and_chunks(
+        article_id=article_id,
+        metadata={
+            "url": "https://example.com/rag-intro",
+            "title": "RAG 介紹",
+            "source": "example.com",
+            "metadata": {"author": "test", "tags": ["RAG"]},
+        },
+        chunks_data=[
+            {"chunk_index": 0, "content": "RAG 是一種結合檢索與生成的技術。", "dense_vector": v0},
+            {"chunk_index": 1, "content": "它透過外部知識庫增強語言模型的回答品質。", "dense_vector": v1},
+            {"chunk_index": 2, "content": "這是一句不相關的填充文字。", "dense_vector": v2},
+        ],
+    )
+    print("Inserted 1 article + 3 chunks.")
+
+asyncio.run(main())