Ship Log Search

Semantic + spatial + timeline search for vessel logs. Every entry is time-stamped and location-stamped. Search by meaning, by proximity, or by date range.

Runs on Cloudflare Workers (free tier) or locally with wrangler dev.

What it does

• Semantic search — "good chumming near Cape Edgecumbe" finds entries about successful baiting near that area, even if the exact words don't match
• Spatial search — "what happened within 50km of 56.6, -134.0?" returns entries sorted by distance
• Timeline search — "all maintenance logs from July 2026" returns time-ordered entries filtered by category
• Quick log — POST a single entry with text, category, lat/lon, and timestamp

Entry categories

Category	Example
catch	"32 sockeye, 600 fath on the slack, 45 min soak"
maintenance	"Port hydraulic ram weeping — topped off fluid, ordered seal kit"
weather	"NW 25 knots, 6-foot swell at the banks. Barometer dropping."
observation	"Humpbacks working the tide rip. Bait balls on the sounder at 25 fathoms."
navigation	"Anchored in Port Bazant. Good holding in mud. 60 feet."

Quickstart

# 1. Create the Vectorize index (one time)
npm run create-index

# 2. Deploy to Cloudflare
npm run deploy

# 3. Or run locally
npm run dev

API

POST /api/log — Quick log entry

{
  "text": "Chummed 45 min on the slack. 32 sockeye, 12 pinks. 600 fath.",
  "category": "catch",
  "lat": 56.6043,
  "lon": -134.4120,
  "location_name": "Cape Edgecumbe"
}

POST /api/ingest — Bulk import

{
  "documents": [
    {
      "id": "log-001",
      "text": "Port engine overheating. Shut down, inspected raw water intake. Cleared debris.",
      "metadata": {
        "timestamp": "2026-07-15T14:30:00Z",
        "lat": 56.6043,
        "lon": -134.4120,
        "category": "maintenance",
        "location_name": "Cape Edgecumbe"
      }
    }
  ]
}

GET /api/search — Semantic search

GET /api/search?q=hydraulic+failure&category=maintenance&from=2026-06-01T00:00:00Z&k=20

Params:

• q — search query (required)
• category — filter by category (catch, maintenance, weather, observation, navigation)
• from — ISO timestamp, entries after this
• to — ISO timestamp, entries before this
• k — max results (1-50, default 20)

GET /api/nearby — Spatial search

GET /api/nearby?lat=56.6&lon=-134.0&radius=50&k=20

Returns entries within radius km of lat, lon, sorted by distance.

GET /api/timeline — Time-ordered entries

GET /api/timeline?from=2026-07-01T00:00:00Z&to=2026-07-31T23:59:59Z&category=catch&k=50

Returns entries in the time range, most recent first.

GET /api/stats — Index statistics

{
  "entries": 1500,
  "model": "@cf/baai/bge-small-en-v1.5",
  "endpoints": ["/api/search", "/api/nearby", "/api/timeline", "/api/log", "/api/ingest"]
}

Free tier limits

• 10M vectors (you'll never hit this with log entries)
• 100K requests/day
• Workers AI embedding: free

Running on your laptop

# Install wrangler
npm install

# Run locally (connects to Cloudflare for AI + Vectorize)
npx wrangler dev

# Open http://localhost:8787

Importing existing logs

If you have existing fishing logs (CSV, spreadsheet, paper notes), write a script to:

1. Parse each entry
2. Build the text field (concatenate description, conditions, catch info)
3. Add timestamp, lat, lon, category as metadata
4. POST to /api/ingest in batches of 100

Example Python script:

import csv, requests, json

API = "https://your-ship-log.workers.dev/api/ingest"

docs = []
with open('fishing_logs.csv') as f:
    for row in csv.DictReader(f):
        text = f"{row['description']} | {row['catch_summary']} | {row['conditions']}"
        docs.append({
            "id": row['id'],
            "text": text,
            "metadata": {
                "timestamp": row['datetime'],
                "lat": float(row['latitude']),
                "lon": float(row['longitude']),
                "category": row.get('category', 'observation'),
                "location_name": row.get('location', ''),
            }
        })

# POST in batches of 100
for i in range(0, len(docs), 100):
    batch = docs[i:i+100]
    r = requests.post(API, json={"documents": batch})
    print(f"Batch {i//100}: {r.json()}")

License

MIT