SamProject/sam-kd
6
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
16a54cef2d7c87ccfd79e4d0803e13c9db98b6e2
sam-kd/chatbot/implementation_plan.md
hskwon aca1767eb9 초기 커밋: 5130 레거시 시스템 
- URL 하드코딩 → .env APP_URL 기반 동적 URL로 변경
- DB 연결 하드코딩 → .env 기반으로 변경
- MySQL strict mode DATE 오류 수정
2025-12-10 20:14:31 +09:00

1.7 KiB
Raw Blame History

Implementation Plan - Advanced Vector RAG (Parallel Deployment)

We will build the "Advanced Vector RAG" system as a separate module to compare with the existing "Live Search RAG".

User Review Required

Note

New Menu Item: "운영자 vertex RAG 챗봇" will be added to the SAM menu. Initial Setup: You must run chatbot/rag/ingest.php once to populate the vector database.

Proposed Changes

1. New UI & Entry Point

[MODIFY] myheader.php

  • Add new menu item: "운영자 vertex RAG 챗봇" linking to chatbot/rag_index.php.

[NEW] chatbot/rag_index.php

  • Copied from index.php.
  • Updated to call rag_api.php instead of api.php.
  • Updated title/branding to indicate "RAG Version".

2. Vector Backend

[NEW] chatbot/rag_api.php

  • Copied from api.php.
  • logic: NotionClient->search() replaced with VectorSearch->search().

3. Vector Engine (Core Logic)

[NEW] chatbot/rag/ingest.php

  • EXTRACT: Fetch all Notion pages.
  • EMBED: Generate vectors using Gemini text-embedding-004.
  • STORE: Save to chatbot/rag/data/vectors.json.

[NEW] chatbot/rag/search.php

  • LOAD: Read vectors.json.
  • SEARCH: Calculate Cosine Similarity between Query Vector and Stored Vectors.
  • RETRIEVE: Return top 5 matches.

Verification Plan

  1. Menu Check: Verify new menu item appears and leads to the new page.
  2. Ingestion: Run ingest.php (via browser or CLI) and check vectors.json size.
  3. Comparison Test: Ask the same question to both chatbots and compare answers.