docker: production image + Gitea Actions for monthly refresh

Dockerfile: self-contained image with corpus + Chroma + BM25 baked
in. Drawbar's compose pulls + runs without volume mounts. Built from
sources.json (labels schema), PRODUCT_NAME=crop_chem by default,
HYBRID_SEARCH=true (always-on for production quality). RERANK_URL +
OLLAMA_URL get set at compose time.

.gitea/workflows/refresh.yml: monthly cron (1st @ 06:00 UTC) does
full scrape → reindex → image push. Scrapes Bayer (~30 min) +
EPA PPLS row-crop filtered (~7h). Skips reindex+push if no corpus
diff. Tags pushed: :latest, :<sha12>, :corpus-<YYYY.MM.DD>.

.gitea/workflows/image-only.yml: on-demand or auto on code-only
pushes to main (paths: docs_mcp/, rag/, scrape/, requirements.txt,
Dockerfile, sources.json). Reindexes from committed corpus, builds
image, pushes. ~10 min vs ~9h full refresh.

.gitignore: corpus/ now COMMITTED (4,159 labels, 265 MB of .md +
sidecars). Lets image-only.yml rebuild indexes without re-scraping.
chroma/ + bm25/ still gitignored (regenerable binary indexes).

.dockerignore: drops venv, eval results, PLAN/README/CLAUDE.md,
deploy/, .git/ — keeps the image lean. corpus + chroma + bm25
explicitly NOT in dockerignore (those go INTO the image).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

corpus/epa_ppls/524-668.md

@@ -0,0 +1,187 @@
+# MON 94637
+
+- EPA Reg No: **524-668**
+- Registrant: BAYER CROPSCIENCE, LLC
+- Signal word: Caution
+- Active ingredients: Bacillus thuringiensis Cry1A.2 Protein and the genetic material (Vector PV -GMIR527237) necessary for its production in MON 94637 soybean event (0.0043%); Bacillus thuringiensis Cry1B.2 Protein and the genetic material (Vector PV -GMIR527237) necessary for its production in MON 94637 soybean event (0.002%)
+- Label accepted: 2025-09-16
+- Source PDF: https://www3.epa.gov/pesticides/chem_search/ppls/000524-00668-20250916.pdf
+
+---
+U.S. ENVIRONMENTAL PROTECTION AGENCY
+Office of Pesticide Programs
+Biopesticides and Pollution Prevention Division (7511M)
+1200 Pennsylvania Ave., N.W.
+Washington, D.C. 20460
+NOTICE OF PESTICIDE:
+܈Registration
+ ܆Reregistration
+(under FIFRA, as amended)
+EPA Reg. Number:
+524-668
+Date of Issuance:
+9/16/2025
+Term of Issuance: 
+Unconditional
+Name of Pesticide Product:
+MON 94637 Lepidopteran-
+Protected Soybean
+Name and Address of Registrant (include ZIP Code):
+Bayer CropScience LP
+800 North Lindbergh Boulevard
+St. Louis, MO 63167
+Note: Changes in labeling differing in substance from that accepted in connection with this registration must be submitted to and accepted by the 
+Biopesticides and Pollution Prevention Division prior to use of the label in commerce. In any correspondence on this product, always refer to the above EPA 
+Registration Number.
+On the basis of information furnished by the registrant, the above named pesticide is hereby registered 
+under the Federal Insecticide, Fungicide, and Rodenticide Act.
+Registration is in no way to be construed as an endorsement or recommendation of this product by the
+U.S. Environmental Protection Agency. In order to protect health and the environment, the 
+Administrator, on his or her motion, may at any time suspend or cancel the registration of a pesticide in 
+accordance with the Act. The acceptance of any name in connection with the registration of a product 
+under the Act is not to be construed as giving the registrant a right to exclusive use of the name or to its 
+use if it has been covered by others.
+This product is unconditionally registered in accordance with FIFRA section 3(c)(5) provided that you:
+1. Submit and/or cite all data required for registration or registration review of your product when 
+the EPA requires all registrants of similar products to submit such data.
+ 
+Signature of Approving Official:
+Michael Mendelsohn, Chief 
+Emerging Technologies Branch
+Biopesticides and Pollution Prevention Division (7511M) 
+Office of Pesticide Programs
+Date:
+9/16/2025
+EPA Form 8570-6 
+MICHAEL MENDELSOHN
+Digitally signed by MICHAEL 
+MENDELSOHN 
+Date: 2025.09.16 09:52:51 -04'00'
+
+Page 2 of 3 
+EPA Reg. No. 524-668 
+Action Case No. 00494632 
+ 
+2. The subject registration is limited to Bacillus thuringiensis Cry1A.2 and Bacillus thuringiensis 
+Cry1B.2 proteins and the genetic material (vector PV-GMIR527237) necessary for their 
+production in MON 94637 (OECD Unique Identifier: MON-94637-8) for use in soybean. 
+ 
+3. The subject registration may be combined through conventional breeding with other registered 
+plant-incorporated protectants that are similarly approved for use in combination, through 
+conventional breeding, with other registered plant-incorporated-protectants to produce 
+soybean varieties with combined pesticidal traits. 
+ 
+4. Seed containing this plant-incorporated protectant product may only be planted for the 
+purposes of agronomic evaluation, seed increase, and production in breeding nurseries as 
+specified in the terms of this registration and on the labeling. Commercial plantings of this 
+product for the purposes of grain production and controlling insect pests are prohibited. 
+ 
+5. The subject registration to a combined yearly total of 15,000 acres or less in the U.S. with no 
+more than 1,000 acres per county (per municipio in Puerto Rico). 
+ 
+6. Harvested seed are not allowed for sale as commercial seed in the U.S. under the terms of this 
+registration, but any non-treated seed containing event MON 94637 may be sold as grain. 
+ 
+7. Bayer CropScience LP must provide reference material, appropriate control substances and 
+technical support if requested by the Food and Drug Administration to support the analytical 
+method for the Cry1A.2 and Cry1B.2 proteins. 
+ 
+8. Upon request, Bayer CropScience LP must provide EPA by January 31st each year annual 
+reports on the acreage where MON 94637 Lepidopteran-Protected Soybean has been grown by 
+state and territory. 
+ 
+Should you wish to add/retain a reference to your company’s website on your label, then please be 
+aware that the website becomes labeling under FIFRA and is subject to review by the EPA. If the 
+website is false or misleading, the product will be considered to be misbranded and sale or distribution 
+of the product is unlawful under FIFRA section 12(a)(1)(E). 40 CFR § 156.10(a)(5) lists examples of 
+statements the EPA may consider false or misleading. In addition, regardless of whether a website is 
+referenced on your product’s label, claims made on the website may not substantially differ from those 
+claims approved through the registration process. Therefore, should the EPA find or if it is brought to 
+our attention that a website contains statements or claims substantially differing from statements or 
+claims made in connection with obtaining a FIFRA section 3 registration, the website will be referred to 
+the EPA’s Office of Enforcement and Compliance Assurance. 
+ 
+Your release for shipment of this product constitutes acceptance of these terms. If these terms are not 
+complied with, this registration will be subject to cancellation in accordance with FIFRA section 6. 
+ 
+A stamped copy of the labeling is enclosed for your records. Please also note that the record for this 
+product currently contains the following acceptable Confidential Statement of Formula (CSF):
+
+Page 3 of 3 
+EPA Reg. No. 524-668 
+Action Case No. 00494632 
+x Basic CSF dated November 6, 2023 
+ 
+If you have any questions, please contact Stephanie Kelly by phone at (202) 566-0890 or via email at 
+kelly.stephanie@epa.gov. 
+ 
+Sincerely, 
+ 
+ 
+ 
+ 
+Michael Mendelsohn, Chief 
+ E m e r g i n g T e c h n o l o g i e s B r a n c h 
+Biopesticides and Pollution 
+ Prevention Division (7511M) 
+Office of Pesticide Programs 
+ 
+Enclosure: MON 94637 Stamped Label 
+MICHAEL 
+MENDELSOHN
+Digitally signed by MICHAEL 
+MENDELSOHN 
+Date: 2025.09.16 09:53:26 
+-04'00'
+
+Plant-Incorporated Protectant Label
+MON 94637
+Lepidopteran-Protected Soybean 
+ (OECD Unique Identifier MON-94637-8) 
+Active Ingredients:
+Bacillus thuringiensis Cry1A.2 protein and the genetic material necessary for its 
+production (vector PV-GMIR527237) in MON 94637 soybean (OECD Unique Identifier 
+MON-94637-8) ................................................................................................... ”0.0043%*
+Bacillus thuringiensis Cry1B.2 protein and the genetic material necessary for its production
+(vector PV-GMIR527237) in MON 94637 soybean (OECD Unique Identifier MON-
+94637-8) .............................................................................................................. ”0.0020%* 
+* Maximum percent (wt/wt) of dry grain.
+KEEP OUT OF REACH OF CHILDREN
+CAUTION
+Net Contents __________
+EPA Registration Number 524-XXX
+EPA Establishment Number 524-MO-002 
+Bayer CropScience LP
+800 North Lindbergh Blvd. 
+St. Louis, Missouri 63167 
+ 
+ 
+M-844542-01-1
+ 
+ 
+09/16/2025
+524-668
+
+DIRECTIONS FOR USE 
+It is a violation of federal law to use this product in a manner inconsistent with its 
+labeling. 
+Seed containing this pesticide product may only be planted for the purposes of agronomic 
+evaluation, seed increase and production in breeding nurseries as specified in the terms and 
+conditions of this registration and on the labe ling. Commercial plantings of this product 
+for the purposes of grain production and controlling insect pests are prohibited. 
+MON 94637 soybean may be combined through conventi onal breeding with other 
+registered plant incorporated protectants that are similarly approved for use in combination, 
+through conventional breeding, with other regist ered plant-incorporated protectants to 
+produce soybean varieties with combined pesticidal traits. 
+Soybean has been transformed to express Bacillus thuringiensis Cry1A.2 and Cry1B.2 
+proteins for the control of the following lepidopteran soybean pests: 
+Soybean looper Chrysodeixis includens 
+Soybean velvetbean caterpillar Anticarsia gemmatalis 
+Plants that contain the pesticide product may be grown on 15,000 acres or less in total per 
+year in the U.S. with no more than 1,000 acres per county per year (per municipio in Puerto 
+Rico). 
+Harvested seed should not be allowed for sale as commercial seed in the U.S. under the 
+current conditions of this registration, but any non-treated seed containing MON 94637 
+may be sold as grain. 
+There are no refuge requirements for planting MON 94637 Soybean. 
+021/HSLGRSWHUDQ3URWHFWHG6R\EHDQ3HQGLQJ