recon/config.yaml

# RECON Configuration
# See PROJECT-BIBLE.md Section 11 for full documentation

# Root path for the PDF library (NFS mount from pi-nas)
library_root: /mnt/library

processing:
  max_pdf_size_mb: 2000         # Raised from 200MB default for large scanned books
  extract_workers: 4          # Concurrent PDF extraction threads
  enrich_workers: 16          # Concurrent Gemini enrichment threads (4 keys x 4)
  embed_workers: 4            # Concurrent embedding threads
  enrich_window_size: 5       # Pages per enrichment window (sent to Gemini)
  embed_batch_size: 500       # Vectors per Qdrant upsert batch
  rate_limit_delay: 0.1       # Delay between Gemini API calls (seconds)
  max_retries: 5              # Max retries for failed documents
  extract_timeout: 1800      # Max seconds per document extraction (30 min, allows vision OCR)
  page_timeout: 30           # Max seconds per page extraction
  enrich_max_retries: 5        # Max retries per enrichment window
  enrich_base_delay: 5.0       # Base backoff delay (seconds) — ~5s, 10s, 20s, 40s, 80s
  enrich_max_delay: 120.0      # Maximum backoff delay cap (seconds)

embedding:
  backend: tei                # "tei" (primary, ~1,711 emb/sec) or "ollama" (fallback, ~8 emb/sec)
  tei_host: 100.64.0.14       # TEI server (cortex)
  tei_port: 8090              # TEI HTTP port
  ollama_host: 100.64.0.14    # Ollama server (cortex) — fallback only
  ollama_port: 11434          # Ollama HTTP port
  model: bge-m3               # Embedding model name
  dimensions: 1024            # CRITICAL: bge-m3 is 1024-dim, NOT 384
  batch_size: 128             # Embeddings per TEI batch request

sparse_embedding:
  enabled: true
  host: 100.64.0.14            # Sparse embedding service (cortex)
  port: 8091                   # Sparse embedding HTTP port

vector_db:
  host: 100.64.0.14           # Qdrant server (cortex)
  port: 6333                  # Qdrant HTTP port
  collection: recon_knowledge_hybrid # Collection name

gemini:
  model: gemini-2.0-flash     # Gemini model for enrichment
  response_mime_type: application/json  # Force JSON output from Gemini

web:
  port: 8420                  # Dashboard HTTP port
  host: 0.0.0.0               # Bind address (all interfaces)

paths:
  base: /opt/recon             # Application root
  data: /opt/recon/data        # Data directory
  text: /opt/recon/data/text   # Extracted text output (data/text/{hash}/page_NNNN.txt)
  concepts: /opt/recon/data/concepts  # Enriched concept JSONs (data/concepts/{hash}/window_N.json)
  intel: /opt/recon/data/intel # ARGUS intel feeds
  logs: /opt/recon/logs        # Log files
  db: /opt/recon/data/recon.db # SQLite database (WAL mode)

book_server:
  base_url: https://files.echo6.co   # Public URL prefix for PDF downloads
  strip_prefix: /mnt/library         # Path prefix stripped when generating download URLs

upload_paths:                  # Category -> filesystem path mapping for uploads
  Survival Reference: /mnt/library/Survival-Companion-Library/Uploads
  Military Doctrine: /mnt/library/Army_Pubs/Uploads
  Gaming: /mnt/library/Gaming
  Reference: /mnt/library/Reference
  Technical: /mnt/library/Technical
  default: /mnt/library        # Fallback for unknown categories

web_scraper:
  words_per_page: 2000         # Target words per page chunk for web content
  fetch_timeout: 30            # HTTP request timeout (seconds)
  rate_limit_delay: 1.0        # Delay between URL fetches (seconds)
  max_batch_size: 50           # Max URLs per batch ingest
  user_agent: "Mozilla/5.0 (compatible; RECON/1.0)"

crawler:
  user_agent: "Mozilla/5.0 (compatible; RECON/1.0)"
  fetch_timeout: 30            # HTTP request timeout (seconds)
  rate_limit_delay: 1.0        # Delay between page fetches (seconds)
  max_pages: 500               # Max pages to discover per crawl
  max_depth: 3                 # Max link-following depth (BFS only, not sitemap)
  inter_site_cooldown: 30      # Seconds to wait between crawling different sites
  recrawl_interval_days: 7     # Skip sites crawled within this many days

  default_exclude:             # URL patterns always excluded from crawling
    - /search
    - /404
    - /login
    - /signup
    - /auth/
    - /api/
    - /assets/
    - /static/
    - /cart
    - /checkout
    - /account
    - /register
    - /subscribe
    - /membership
    - /shop
    - /store
    - /product
    - /wp-admin
    - /feed
    - /wp-json
    - /xmlrpc
    - /.well-known
    - /cdn-cgi

  # ─── Crawl Targets ─────────────────────────────────────────────
  # Sites are crawled by the scheduler loop in tier order (1 first).
  # Per-site delay overrides global rate_limit_delay for that site.
  # Per-site max_pages/max_depth override global defaults.

  # Disabled 2026-04-14 for refactor — see refactored-recon repo for context
  sites: []

  # sites:
  #
  # # ═══ TIER 1 — Free, authoritative, high-density ═══
  #
  # - url: https://hesperian.org/all-hesperian-health-guides
  # category: Medical
  # max_depth: 3
  # delay: 3.0
  # tier: 1
  # notes: "Free health guides — WTIND, midwives, community health"
  #
  # - url: https://swsbm.com
  # category: Medical
  # max_depth: 3
  # delay: 3.0
  # tier: 1
  # notes: "Michael Moore's entire free clinical herbal library — PDFs"
  #
  # - url: https://swsbm.henriettesherbal.com
  # category: Medical
  # max_depth: 3
  # delay: 3.0
  # tier: 1
  # notes: "Mirror of Moore's library — grab both"
  #
  # - url: https://nchfp.uga.edu
  # category: Sustainment Systems
  # max_depth: 3
  # delay: 2.0
  # tier: 1
  # notes: "USDA canning/preservation safety authority"
  #
  # - url: https://extension.uidaho.edu
  # category: Foundational Skills
  # max_depth: 3
  # delay: 2.0
  # tier: 1
  # notes: "Idaho-specific — soil, water, crops, livestock"
  #
  # - url: https://extension.usu.edu
  # category: Foundational Skills
  # max_depth: 3
  # delay: 2.0
  # tier: 1
  # notes: "Utah State — Idaho-adjacent climate"
  #
  # - url: https://attra.ncat.org
  # category: Sustainment Systems
  # max_depth: 3
  # delay: 3.0
  # tier: 1
  # notes: "ATTRA sustainable ag — hundreds of free publications"
  #
  # - url: https://pfaf.org
  # category: Sustainment Systems
  # max_depth: 3
  # delay: 3.0
  # tier: 1
  # notes: "Plants For A Future — 7,000+ edible/medicinal plant profiles"
  #
  # - url: https://eattheweeds.com
  # category: Sustainment Systems
  # max_depth: 3
  # delay: 3.0
  # tier: 1
  # notes: "Green Deane — 1,000+ foraging plant articles"
  #
  # - url: https://lowtechmagazine.com
  # category: Off-Grid Systems
  # max_depth: 3
  # delay: 3.0
  # tier: 1
  # notes: "Exceptional low-tech systems analysis"
  #
  # - url: https://appropedia.org
  # category: Off-Grid Systems
  # max_depth: 3
  # delay: 3.0
  # tier: 1
  # notes: "Appropriate technology wiki"
  #
  # - url: https://journeytoforever.org
  # category: Off-Grid Systems
  # max_depth: 3
  # delay: 3.0
  # tier: 1
  # notes: "VITA manuals, biodiesel, biogas, hand tools archive"
  #
  # - url: https://cd3wd.com
  # category: Off-Grid Systems
  # max_depth: 2
  # delay: 3.0
  # tier: 1
  # notes: "1,050+ appropriate technology eBooks — index pages only"
  #
  # - url: https://practicalselfreliance.com
  # category: Sustainment Systems
  # max_depth: 3
  # delay: 3.0
  # tier: 1
  # notes: "Ashley Adamant — foraging, preservation, homesteading"
  #
  # - url: https://open.oregonstate.edu/permaculture
  # category: Off-Grid Systems
  # max_depth: 3
  # delay: 3.0
  # tier: 1
  # notes: "Millison's free permaculture textbook"
  #
  # - url: https://open.oregonstate.edu/permaculturedesign
  # category: Off-Grid Systems
  # max_depth: 3
  # delay: 3.0
  # tier: 1
  # notes: "Millison's advanced permaculture textbook"
  #
  # - url: https://mushroomexpert.com
  # category: Sustainment Systems
  # max_depth: 3
  # delay: 3.0
  # tier: 1
  # notes: "Michael Kuo — mushroom ID, taxonomy, regional coverage"
  #
  # # ═══ TIER 2 — High value, second pass ═══
  #
  # - url: https://motherearthnews.com
  # category: Foundational Skills
  # max_depth: 2
  # max_pages: 200
  # delay: 8.0
  # tier: 2
  # notes: "50 years of homesteading archive — large commercial site, be polite"
  #
  # - url: https://permacultureresearchinstitute.com
  # category: Off-Grid Systems
  # max_depth: 3
  # delay: 5.0
  # tier: 2
  # notes: "Geoff Lawton — articles, case studies"
  #
  # - url: https://learnyourland.com
  # category: Sustainment Systems
  # max_depth: 3
  # delay: 5.0
  # tier: 2
  # notes: "Adam Haritan — foraging articles"
  #
  # - url: https://herbswithRosalee.com
  # category: Medical
  # max_depth: 3
  # delay: 5.0
  # tier: 2
  # notes: "Rosalee de la Foret — clinical herbalism articles"
  #
  # - url: https://commonwealthherbs.com
  # category: Medical
  # max_depth: 3
  # delay: 5.0
  # tier: 2
  # notes: "Katja and Ryn — clinical herbalism"
  #
  # - url: https://soilfoodweb.com
  # category: Off-Grid Systems
  # max_depth: 3
  # delay: 5.0
  # tier: 2
  # notes: "Elaine Ingham soil biology — archive before it goes dark"
  #
  # - url: https://rocketstoves.com
  # category: Off-Grid Systems
  # max_depth: 3
  # delay: 5.0
  # tier: 2
  # notes: "Ianto Evans — rocket mass heater designs and PDFs"
  #
  # - url: https://farmsteadmeatsmith.com
  # category: Sustainment Systems
  # max_depth: 2
  # delay: 5.0
  # tier: 2
  # notes: "Brandon Sheard — butchering articles (free content only)"
  #
  # - url: https://deeranddeerhunting.com
  # category: Sustainment Systems
  # max_depth: 2
  # delay: 5.0
  # tier: 2
  # notes: "Field dressing, processing, hunting technique library"
  #
  # # ═══ TIER 3 — Government (authoritative) ═══
  #
  # - url: https://plants.usda.gov
  # category: Sustainment Systems
  # max_depth: 2
  # delay: 2.0
  # tier: 3
  # notes: "USDA native plant database"
  #
  # - url: https://ars.usda.gov
  # category: Sustainment Systems
  # max_depth: 2
  # delay: 2.0
  # tier: 3
  # notes: "USDA Agricultural Research publications"
  #
  # - url: https://nrcs.usda.gov
  # category: Off-Grid Systems
  # max_depth: 2
  # delay: 2.0
  # tier: 3
  # notes: "Soil surveys, conservation practice standards"
  #
  # - url: https://ready.gov
  # category: Scenario Playbooks
  # max_depth: 3
  # delay: 2.0
  # tier: 3
  # notes: "FEMA emergency preparedness guides"
  #
  # - url: https://emergency.cdc.gov
  # category: Medical
  # max_depth: 3
  # delay: 2.0
  # tier: 3
  # notes: "Public health emergency references"
  #
  # - url: https://agri.idaho.gov
  # category: Foundational Skills
  # max_depth: 2
  # delay: 2.0
  # tier: 3
  # notes: "Idaho Dept of Agriculture — local relevance"
  #
  # - url: https://driveonwood.com
  # category: Off-Grid Systems
  # max_depth: 3
  # delay: 3.0
  # tier: 3
  # notes: "Wood gasification — FEMA manual + modern improvements"
  #
  # # ═══ TIER 4 — Selective scrape (specific sections only) ═══
  #
  # - url: https://richsoil.com
  # category: Off-Grid Systems
  # max_depth: 2
  # delay: 5.0
  # tier: 4
  # notes: "Paul Wheaton — rocket mass heaters, natural building"
  #
  # - url: https://wildfoodgirl.com
  # category: Sustainment Systems
  # max_depth: 3
  # delay: 5.0
  # tier: 4
  # notes: "Colorado foraging — Mountain West species"
  #
  # - url: https://foragersharvest.com
  # category: Sustainment Systems
  # max_depth: 3
  # delay: 5.0
  # tier: 4
  # notes: "Sam Thayer's site — articles"
  #
  # - url: https://mountainroseherbs.com/blog
  # category: Medical
  # max_depth: 2
  # delay: 5.0
  # tier: 4
  # notes: "Herb profiles and preparations — blog section only"
  #
  # - url: https://herbalprepper.com
  # category: Medical
  # max_depth: 3
  # delay: 5.0
  # tier: 4
  # notes: "Cat Ellis — grid-down herbalism"
  #
  # - url: https://prolongedfieldcare.org
  # category: Medical
  # max_depth: 3
  # delay: 5.0
  # tier: 4
  # notes: "PFC Collective — austere medical protocols"
  #
service:
  scan_interval: 3600          # Seconds between library scans (1 hour)
  stage_poll_interval: 30      # Seconds stages sleep when idle
  progress_interval: 60        # Seconds between progress log lines

peertube:
  api_base: http://192.168.1.170       # Internal PeerTube API (CT 110 nginx)
  public_url: https://stream.echo6.co  # Public URL for video links
  fetch_timeout: 30                     # HTTP timeout for API/VTT requests
  rate_limit_delay: 0.5                 # Delay between video ingestions (seconds)
  poll_interval: 1800                    # Seconds between PeerTube acquisition polls (30 min)

scraper:
  workspace: /opt/recon/data/scraper      # Working directory for wget mirrors + ZIM builds
  output_dir: /mnt/kiwix                  # Finished .zim files land here (kiwix-serve library)
  rate_limit_delay: 0.5                   # Seconds between wget requests (--wait)
  wait_random: 1.0                        # Random jitter added to wait (--random-wait range)
  default_language: eng                   # ISO 639-3 language code for ZIM metadata
  user_agent: "Mozilla/5.0 (compatible; RECON/1.0; +https://echo6.co)"
  poll_interval: 300                      # Seconds between checking for pending scrape jobs
  keep_workspace_on_failure: true         # Retain workspace for debugging when a job fails

  # Default URL patterns rejected by wget --reject-regex.
  # Covers common CMS junk across WordPress, Squarespace, Wix, Ghost, Drupal, etc.
  # Per-job overrides: additional_reject_patterns (appended) or skip_default_patterns (bypass).
  default_reject_patterns:
    # WordPress
    - '\?share='
    - '\?replytocom='
    - '\?like_comment='
    - '/feed/'
    - '/wp-json/'
    - '/wp-login'
    - '/wp-admin'
    - '/wp-cron'
    - '\?attachment_id='
    - '/xmlrpc'
    - '/trackback'
    - '/comment-page-'
    - '\?doing_wp_cron'
    # Squarespace
    - '\?format=json'
    - '\?format=rss'
    - '/api/'
    # Wix
    - '/_api/'
    - '/_partials/'
    # Ghost
    - '/ghost/'
    - '/p/'
    # Drupal
    - '\?q=comment'
    - '\?q=node'
    - '/user/login'
    - '/user/register'
    # General CMS / site chrome
    - '/login'
    - '/signup'
    - '/register'
    - '/cart'
    - '/checkout'
    - '/search\?'
    - '/tag/'
    - '/author/'
    - '\?print='
    - '\?pdf='
    - '\?format=amp'
    - '\?preview='
    - '/rss'
    - '/atom'
    - '/cdn-cgi/'

  # Pre-flight mode detection
  preflight:
    enabled: true
    timeout: 30                    # Seconds for single-page Playwright fetch
    min_static_size: 5120          # 5KB - wget HTML below this = suspect JS site
    min_browser_size: 20480        # 20KB - browser HTML above this confirms JS
    spa_markers:
      - 'div#root'
      - 'div#app'
      - 'div#__next'

  # SingleFile CLI settings (browser crawl mode)
  singlefile:
    executable: single-file
    chromium_path: "/usr/bin/chromium-browser"
    crawl_max_depth: 10

# Stream B: New Library Pipeline
new_pipeline:
  # Disabled 2026-04-14 for refactor — see refactored-recon repo for context
  enabled: false
  acquired_dir: /mnt/library/_acquired
  ingest_dir: /mnt/library/_ingest
  duplicates_dir: /mnt/library/_ingest/_duplicates
  failed_dir: /mnt/library/_ingest/_failed
  poll_interval: 60
  mtime_stability: 10
  pilot_domain: "Civil Organization"
  spaces_to_underscores: true

# Refactored pipeline configuration (2026-04-14)
# See https://forge.echo6.co/matt/refactored-recon for design
pipeline:
  acquired_root: /opt/recon/data/acquired
  processing_root: /opt/recon/data/processing
  # Subfolder name -> processor module mapping
  # Processors do not exist yet; this is scaffolding for Phase 3+
  dispatch:
    pdf: pdf_processor
    stream: transcript_processor
    html: html_processor
    text: text_processor
  # mtime stability threshold for picking up files from acquired/
  mtime_stability_seconds: 10
  # Language filter: skip non-English content before Gemini enrichment
  language_filter: true           # Enable langdetect-based filtering
  allowed_languages:              # ISO 639-1 codes allowed through enrichment
    - en