Dockerfile now sets OLLAMA_URL=http://ollama:11434 and
RERANK_URL=http://llama-rerank:8080 as image defaults, assuming the
MCP container shares a Docker network with services named `ollama`
and `llama-rerank` (typical compose pattern). Drawbar's stack
already runs both — no cross-host IPs to maintain, no off-stack
GPU dependencies. Stays inside the trashpanda compose.
deploy/drawbar-compose-snippet.md simplified: no environment
overrides needed for the common case. Override block shown only
for stacks with non-default service names. Pull tag updated to
:corpus-2026.05.24.
Per the new architecture call:
- MCP doesn't reach out to cross-host Ollama instances (192.168.0.2,
192.168.0.125 etc.) at serve time — only at index-build time in CI.
- All serve-time dependencies are in the same Docker network as
the consumer apps.
Code push touches Dockerfile → image-only.yml will rebuild + push.
Future-me note: the image-only.yml needs Ollama reachable from the
Gitea Actions runner for the reindex step; that still uses the LAN
endpoints (workflow env), which is correct since indexing is CI-side
not serve-side.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
## Phase 11 — Curated agronomy / label-handling knowledge layer
docs_mcp/lessons.md: 13 topic-anchored markdown sections covering
the LLM-side context a farmer-advisor needs alongside the raw
label corpus —
- how-to-use-this-corpus
- epa-signal-words
- rei-phi-fundamentals
- rup-handling
- supplemental-labels-24c-2ee
- tank-mix-fundamentals
- resistance-management-hrac-frac-irac
- glufosinate-application-rules
- dicamba-application-rules
- lake-erie-watershed-ohio
- scn-and-other-seed-treatment-context
- drift-management-essentials
- how-to-format-recommendations
Each Topic block is independently retrievable via the new MCP tool:
ppls_api_lessons(topic="rup-handling")
Or with no topic to get the full TOC, or with a substring to
match-and-return matching sections ("dicamba" → dicamba-application-rules).
Tool docstring instructs the LLM to call this proactively before any
pesticide recommendation so the recommendation lands with regulatory
framing, resistance-group callouts, RUP applicator language, and the
canonical recommendation format — not just a rate from a label.
## Phase 6 — Reranker moved to GPU on trashpanda
Stopped the local CPU container and started on trashpanda's Tesla P4
(8 GB VRAM) via:
docker run -d --name llama-rerank --restart unless-stopped --gpus all \
-p 8082:8080 \
ghcr.io/ggml-org/llama.cpp:server-cuda \
-hf gpustack/jina-reranker-v2-base-multilingual-GGUF:Q8_0 \
--reranking --host 0.0.0.0 --port 8080 -ngl 99
The :server-cuda image variant (not :server) is required for CUDA
backend; -ngl 99 offloads all layers to GPU.
Latency: 50-doc rerank dropped from ~23 s on CPU to ~0.7-1.5 s on
the Tesla P4 — production-grade interactive speeds.
deploy/rerank-docker.md updated with the trashpanda deploy recipe,
troubleshooting (mostly "did you use server-cuda?"), and a perf
reference table. The MCP server's RERANK_URL just points at
http://10.10.1.65:8082 now.
GPU eval still completing in background; results land in
eval/results/with_rerank_gpu.md as a follow-up commit.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
justin