docs: update prd.json and progress.txt for story 13

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
“BeeRad”
2026-01-29 15:43:27 +11:00
co-authored by Claude Opus 4.5
parent 0d2d4bf0fd
commit 883299d070
2 changed files with 25 additions and 2 deletions
+2 -2
View File
@@ -174,8 +174,8 @@
"Both server.js and stdio-server.js updated consistently", "Both server.js and stdio-server.js updated consistently",
"npm run type-check passes" "npm run type-check passes"
], ],
"passes": false, "passes": true,
"notes": "" "notes": "Removed 3 extraction tools (rah_extract_url, rah_extract_youtube, rah_extract_pdf) that called non-existent API endpoints. MCP servers now expose 11 working tools: node CRUD (add, search, update, get), edge CRUD (create, query, update), dimension CRUD (create, update, delete), semantic search. Both servers updated consistently."
}, },
{ {
"id": "14", "id": "14",
+23
View File
@@ -200,3 +200,26 @@ Goal: Strip ra-h_os to lightweight 2-panel UI + MCP server
- These were only used by contextBuilder.ts which was deleted in Story 8 - These were only used by contextBuilder.ts which was deleted in Story 8
- Type files are well-organized - each file has a clear purpose - Type files are well-organized - each file has a clear purpose
## Story 13: Audit and clean MCP server
- Completed: Thu 29 Jan 2026
- Files modified:
- apps/mcp-server/server.js (removed extraction tools)
- apps/mcp-server/stdio-server.js (removed extraction tools)
- Tools removed (called non-existent API routes):
- rah_extract_url (called /api/extract/url - doesn't exist)
- rah_extract_youtube (called /api/extract/youtube - doesn't exist)
- rah_extract_pdf (called /api/extract/pdf - doesn't exist)
- Remaining 11 tools (all verified working):
- Node CRUD: rah_add_node, rah_search_nodes, rah_update_node, rah_get_nodes
- Edge CRUD: rah_create_edge, rah_query_edges, rah_update_edge
- Dimension CRUD: rah_create_dimension, rah_update_dimension, rah_delete_dimension
- Semantic search: rah_search_embeddings
- Analysis:
- MCP servers are pure API clients - no direct code dependencies
- All tools use callRaHApi() to call Next.js API routes
- No chat/agent/delegation dependencies
- Both servers were kept in sync
- Learnings:
- Extraction tools were dead code - the API routes never existed in the open source version
- The MCP architecture is clean - just HTTP calls to local API