Pilot for moving the fleet off mcp 1.x before it goes away. The port is
three mechanical changes plus docs:
- import: mcp.server.fastmcp.FastMCP -> mcp.server.mcpserver.MCPServer
- constructor: transport options moved off it; MCPServer(name) only
- run(): host/port/stateless_http/transport_security are now kwargs,
and `mcp.settings` no longer exists
`@mcp.tool()` is unchanged — all six tools keep their decorators and
signatures untouched.
Verified, not assumed:
1. tools/list is byte-for-byte identical between mcp 1.27.1 and 2.x
(6 tools, same names, same inputSchema properties and required lists,
dumped with by_alias=True so the comparison is in wire format). This
was the real risk: 2.x switched the protocol types to snake_case
internally, and tool schemas are what decide routing.
2. The server boots for real over streamable-http with stateless_http=True
and DNS-rebinding protection disabled, and answers initialize 200 with
the same capability set.
3. CI now runs `import docs_mcp.server` inside the built image before push.
One behavior change: an unversioned server reports serverInfo.version ""
under 2.x instead of the SDK version, so Drawbar will see an empty version
string. Cosmetic — left as-is rather than inventing a version scheme.
Dependency note: 2.x has no [fastmcp] extra, replaces httpx with httpx2,
and requires opentelemetry-api. Our own httpx>=0.27 pin is unaffected —
httpx2 is a separate distribution.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01FnVuG79cYPcRLTp4pC8ujR
The scaffold-era README was out of sync with the shipped product:
- Vendor counts stale (recon estimates, not actual deployed counts)
- Trial data sources (gh_plot_reports + agripro_trials) entirely
unmentioned
- Tool list listed `corpus_status` (doesn't exist) and missed both
`lookup_variety` and `search_trials`
- Build-phase table showed everything as "pending" / "next" but
Phases 1-8 + 11 all shipped
Rewrite to reflect the deployed state:
- Corpus inventory: 760 variety records + 4,313 trial documents =
5,073 chunks across 6 sources
- All 6 MCP tools documented with their purpose
- Eval baseline table (hybrid+rerank wins 100%, P@1 90%, MRR 0.905)
with the surprising findings (dense alone is noise; hybrid w/o
rerank is WORSE than BM25 alone)
- Deploy mechanics: Watchtower chain, 4-GPU embedder pool, shared
llama-rerank sidecar with the network-attach gotcha
- Status table: ✅ on the phases that shipped, deferred work list
(becks_pfr, 2023 plot backfill, NK trials, Channel Seed brand)
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
claudeandClaude Opus 5
justinandClaude Opus 4.7