crop-chem-docs

34 Commits 1 Branch 0 Tags

3 Commits

Include renames

Author	SHA1	Message	Date
claudeandClaude Opus 5	c363a221e1	docs(deploy): make deploy/ describe the deployment that actually exists deploy/docker-compose.yml was unedited docs-mcp-template boilerplate — untouched since the scaffold commit, still carrying <product>, <registry> and <owner> placeholders — describing a standalone stack that has never existed. crop-chem-docs runs as the `chem-mcp` service inside Drawbar's parent compose. It also set MCP_ALLOWED_HOSTS, which no code in this repo reads. The knob is MCP_DISABLE_DNS_REBINDING_PROTECTION. Anyone who trusted the old file and set an allowlist would have gotten a 421 on every request with nothing in the logs to explain it. - deploy/docker-compose.yml: replaced with the real chem-mcp block, a copy of what runs in Drawbar/drawbar-backend. Verified structurally identical to the parent (image, environment, expose, extra_hosts, restart, labels all equal). Carries the why for each setting: the :latest-vs-corpus-tag Watchtower trap (#339), the rebind-protection rationale, and that the OLLAMA_URL override is load-bearing because Drawbar's own ollama service is commented out — the image default http://ollama:11434 does not resolve in that stack, so without the override every search_docs call fails to embed its query. - deploy/drawbar-compose-snippet.md: deleted. It was a second, differently-wrong copy (service name `crop-chem-docs`, ports 8001:8000, and "No environment block needed — the image's defaults handle it", which is false on both the rebind and Ollama counts). Its still-true content (verification commands) moved into the compose file; the tag scheme and deploy chain were already in the README. - deploy/rerank-docker.md: RERANK_URL said http://10.10.1.65:8082. In production the MCP reaches the sidecar by compose service name (http://llama-rerank:8080, baked into the image). Documents the network-attach gotcha that makes rerank fail silently, and keeps the host-IP form for local dev. - README.md: file tree updated for the deleted file; Watchtower poll interval corrected 5 min -> 60s (WATCHTOWER_POLL_INTERVAL=60, as configured on trashpanda). Verified: no <product>/<registry>/<owner> placeholders remain in deploy/ or README, and all six env vars set in the block are ones the server actually reads. Closes #5 Co-Authored-By: Claude Opus 5 (1M context) <[email protected]> Claude-Session: https://claude.ai/code/session_01FFBDnRWHispovmJVK9rXc9	2026-09-10 21:43:56 -04:00
justinandClaude Opus 4.7	af44d7a102	Phase 11 + Phase 6 GPU move ## 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) <[email protected]>	2026-05-24 12:10:09 -04:00
justinandClaude Opus 4.7	278fe5f456	Phase 6: reranker sidecar (jina-reranker-v2-base via llama.cpp) Wires the docs_mcp/server.py reranker hook into a real backend: ghcr.io/ggml-org/llama.cpp:server \\ -hf gpustack/jina-reranker-v2-base-multilingual-GGUF:Q8_0 \\ --reranking --host 0.0.0.0 --port 8080 Setup recipe at deploy/rerank-docker.md. The MCP server already honors RERANK_URL (added in Phase 7+8 commit); setting it to http://<host>:8082 turns on rerank automatically. ## Eval results (35 queries, k=5, pool=50) \| Retriever \| MRR \| Recall@5 \| nDCG@5 \| \|----------------\|-------\|----------\|--------\| \| dense \| 0.027 \| 0.086 \| 0.041 \| \| bm25 \| 0.544 \| 0.586 \| 0.524 \| \| hybrid-rrf \| 0.114 \| 0.114 \| 0.108 \| \| dense+rerank \| 0.171 \| 0.143 \| 0.149 \| \| hybrid+rerank \| 0.672 \| 0.638 \| 0.621 \| ← winner The reranker fixes hybrid's failure mode (dense noise polluting the fused pool) by scoring each (query, chunk) pair independently. Net: hybrid+rerank gives +24% MRR over BM25-only. Smoke test for the reranker itself (query: "soybean herbicide for waterhemp", 4 candidates): index=1 SENCOR metribuzin waterhemp soybean → score=0.84 ← right index=3 Headline wheat fungicide → score=-2.80 index=2 Lorsban corn rootworm → score=-2.91 index=0 Roundup fallow burndown → score=-3.44 Strong separation between the right doc and the rest. ## Production gotchas - CPU-only reranker is slow (~23s for a 50-doc pool). For interactive use put it on GPU (`--gpus all`); ~10-20× faster. - jina-reranker rejects the ENTIRE batch if any pair exceeds n_ctx_train=1024 — server truncates each doc to 2000 chars before sending. Already handled in _rerank_pool. Per-query rerank report at eval/results/with_rerank.md. Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>	2026-05-24 10:50:03 -04:00

Author

SHA1

Message

Date

claudeandClaude Opus 5

c363a221e1

docs(deploy): make deploy/ describe the deployment that actually exists

deploy/docker-compose.yml was unedited docs-mcp-template boilerplate —
untouched since the scaffold commit, still carrying <product>,
<registry> and <owner> placeholders — describing a standalone stack
that has never existed. crop-chem-docs runs as the `chem-mcp` service
inside Drawbar's parent compose.

It also set MCP_ALLOWED_HOSTS, which no code in this repo reads. The
knob is MCP_DISABLE_DNS_REBINDING_PROTECTION. Anyone who trusted the
old file and set an allowlist would have gotten a 421 on every request
with nothing in the logs to explain it.

- deploy/docker-compose.yml: replaced with the real chem-mcp block, a
  copy of what runs in Drawbar/drawbar-backend. Verified structurally
  identical to the parent (image, environment, expose, extra_hosts,
  restart, labels all equal). Carries the why for each setting: the
  :latest-vs-corpus-tag Watchtower trap (#339), the rebind-protection
  rationale, and that the OLLAMA_URL override is load-bearing because
  Drawbar's own ollama service is commented out — the image default
  http://ollama:11434 does not resolve in that stack, so without the
  override every search_docs call fails to embed its query.
- deploy/drawbar-compose-snippet.md: deleted. It was a second,
  differently-wrong copy (service name `crop-chem-docs`, ports
  8001:8000, and "No environment block needed — the image's defaults
  handle it", which is false on both the rebind and Ollama counts).
  Its still-true content (verification commands) moved into the compose
  file; the tag scheme and deploy chain were already in the README.
- deploy/rerank-docker.md: RERANK_URL said http://10.10.1.65:8082. In
  production the MCP reaches the sidecar by compose service name
  (http://llama-rerank:8080, baked into the image). Documents the
  network-attach gotcha that makes rerank fail silently, and keeps the
  host-IP form for local dev.
- README.md: file tree updated for the deleted file; Watchtower poll
  interval corrected 5 min -> 60s (WATCHTOWER_POLL_INTERVAL=60, as
  configured on trashpanda).

Verified: no <product>/<registry>/<owner> placeholders remain in
deploy/ or README, and all six env vars set in the block are ones the
server actually reads.

Closes #5

Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01FFBDnRWHispovmJVK9rXc9

2026-09-10 21:43:56 -04:00

justinandClaude Opus 4.7

af44d7a102

Phase 11 + Phase 6 GPU move

## 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) <[email protected]>

2026-05-24 12:10:09 -04:00

justinandClaude Opus 4.7

278fe5f456

Phase 6: reranker sidecar (jina-reranker-v2-base via llama.cpp)

Wires the docs_mcp/server.py reranker hook into a real backend:
  ghcr.io/ggml-org/llama.cpp:server \\
    -hf gpustack/jina-reranker-v2-base-multilingual-GGUF:Q8_0 \\
    --reranking --host 0.0.0.0 --port 8080

Setup recipe at deploy/rerank-docker.md. The MCP server already
honors RERANK_URL (added in Phase 7+8 commit); setting it to
http://<host>:8082 turns on rerank automatically.

## Eval results (35 queries, k=5, pool=50)

  | Retriever      | MRR   | Recall@5 | nDCG@5 |
  |----------------|-------|----------|--------|
  | dense          | 0.027 | 0.086    | 0.041  |
  | bm25           | 0.544 | 0.586    | 0.524  |
  | hybrid-rrf     | 0.114 | 0.114    | 0.108  |
  | dense+rerank   | 0.171 | 0.143    | 0.149  |
  | hybrid+rerank  | 0.672 | 0.638    | 0.621  |  ← winner

The reranker fixes hybrid's failure mode (dense noise polluting
the fused pool) by scoring each (query, chunk) pair independently.
Net: hybrid+rerank gives +24% MRR over BM25-only.

Smoke test for the reranker itself (query: "soybean herbicide for
waterhemp", 4 candidates):
  index=1 SENCOR metribuzin waterhemp soybean → score=0.84  ← right
  index=3 Headline wheat fungicide           → score=-2.80
  index=2 Lorsban corn rootworm              → score=-2.91
  index=0 Roundup fallow burndown            → score=-3.44
Strong separation between the right doc and the rest.

## Production gotchas

- CPU-only reranker is slow (~23s for a 50-doc pool). For
  interactive use put it on GPU (`--gpus all`); ~10-20× faster.
- jina-reranker rejects the ENTIRE batch if any pair exceeds
  n_ctx_train=1024 — server truncates each doc to 2000 chars
  before sending. Already handled in _rerank_pool.

Per-query rerank report at eval/results/with_rerank.md.

Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

2026-05-24 10:50:03 -04:00