chomp

Local food database CLI for AI-assisted nutrition tracking.

Built by @meimakes

Problem

AI assistants waste credits searching for nutrition data every time you log food. Your diet is repetitive — the same foods show up constantly. Why look up "ribeye" for the 50th time?

Solution

Local SQLite database that learns YOUR foods. AI queries it instead of web searching. Includes a web dashboard, REST API, and MCP server for Claude Desktop integration.

Commands

# Log food (default action)
chomp bacon                      # logs bacon (1 serving)
chomp ribeye 8oz                 # logs 8oz ribeye
chomp "bare bar"                 # logs bare bar
chomp "Ortiz Sardines" 0.5       # logs half a serving (bare number = serving multiplier)
chomp --date 2026-03-21 ribeye 8oz  # backdate to a specific day

# Manage foods (the database of what things are)
chomp add ribeye --protein 23 --fat 18 --carbs 0 --per 100g
chomp add ribeye -p 23 -f 18 -c 0 --per 100g --alias rib
chomp edit ribeye --protein 25 --fat 20
chomp delete "food name"         # removes food definition from DB

# Manage log entries (what you actually ate)
chomp unlog 42                   # delete log entry by ID
chomp unlog-last                 # delete most recent log entry
chomp edit-log 42 --amount 8oz   # fix a log entry

# Query
chomp search salmon              # fuzzy match
chomp today                      # show today's totals
chomp history                    # recent logs (default 7 days)
chomp history --days 30          # recent logs (30 days)
chomp stats                      # database stats

# Compound foods
chomp compound "breakfast" -i "3 eggs + 2 bacon"

# Import/Export
chomp export --csv               # for spreadsheets
chomp export --json              # structured output
chomp import usda                # seed from USDA database
chomp import csv --path foods.csv

# Server
chomp serve                          # MCP server (stdio)
chomp serve --transport sse          # HTTP server with SSE, REST API, dashboard
chomp serve --transport sse --auth-key mysecret  # with authentication
chomp serve --transport both         # stdio + HTTP simultaneously

All commands support --json for structured output.

Features

• Fuzzy matching — "rib eye" = "ribeye"
• Learned portions — "salmon" defaults to your usual 4oz (via default_amount field)
• Flexible amounts — bare numbers are serving multipliers (0.5 of a 4oz serving = 2oz), units work too (8oz, 3 tbsp)
• Aliases — "bb" = "bare bar"
• Compound foods — save multi-item meals as single entries
• Web dashboard — dark-themed nutrition dashboard with charts, target tracking, and daily breakdowns
• REST API — full CRUD API for foods and log entries
• MCP server — stdio and SSE transports for Claude Desktop / remote AI agents
• Authentication — bearer token + session cookie auth for the HTTP server
• Remote client — point the CLI at a remote chomp server instead of a local DB
• USDA import — seed from FoodData Central SR Legacy dataset (~7,800 foods)
• CSV import — bulk load from CSV (header: name,protein,fat,carbs,calories,serving)
• Docker/Railway ready — multi-stage Dockerfile with persistent volume support

Web Dashboard

When running with --transport sse, a dashboard is available at /dashboard.

Features: daily calorie/protein averages, macro ratio donut chart, target tracking bars, day-of-week breakdowns, today's entries table, top foods by frequency.

Dashboard targets are configurable via URL params:

/dashboard?protein=120&calories=1500&calorieMode=under

Param	Default	Description
protein	100	Daily protein target (g)
calories	2000	Daily calorie target
calorieMode	under	under or over — whether hitting target means staying below or above

AI Integration

CLI (for exec-based agents)

chomp "salmon 4oz" --json        # log + structured output
chomp search salmon --json       # nutrition lookup without web search

MCP Server

# stdio transport (Claude Desktop)
chomp serve
chomp serve --transport stdio

# SSE transport (remote agents, Railway)
chomp serve --transport sse
chomp serve --transport sse --port 3456 --host 0.0.0.0

# Both transports simultaneously
chomp serve --transport both --port 3000

MCP tools exposed:

Tool	Description
log_food(food, date?)	Log food, returns entry with calculated macros
search_food(query)	Fuzzy search with nutrition info
add_food(name, protein, fat, carbs, serving, ...)	Add new food to DB
edit_food(name, ...)	Edit an existing food
delete_food(name)	Delete a food from DB
get_today()	Today's macro totals
get_history(days?)	Recent log entries
unlog(id)	Delete a log entry by ID
unlog_last()	Delete most recent log entry
edit_log(id, ...)	Edit a log entry

REST API

All endpoints (except /health, /login, /logout) require authentication via Authorization: Bearer <key> header or session cookie.

GET    /dashboard          # web dashboard
GET    /health             # health check
GET    /login              # login page
POST   /login              # authenticate (sets session cookie)
POST   /logout             # clear session

GET    /api/today           # today's totals + entries
GET    /api/history?days=7  # log history
GET    /api/export?days=30  # CSV export
POST   /api/log             # log food  { "food": "ribeye 8oz", "date": "2026-03-21" }
DELETE /api/log/:id         # delete log entry
DELETE /api/log/last        # delete most recent log entry
PUT    /api/log/:id         # edit log entry

GET    /api/foods?q=salmon  # search foods
POST   /api/foods           # add food
PUT    /api/foods/:name     # edit food
DELETE /api/foods/:name     # delete food

GET    /api/stats           # database stats

Remote Client Mode

Point the CLI at a remote chomp server instead of using a local database:

export CHOMP_SERVER_URL=https://your-chomp.railway.app
export CHOMP_AUTH_KEY=your-secret-key
chomp today                      # queries the remote server

Deployment (Docker / Railway)

docker build -t chomp .
docker run -p 3000:3000 -e CHOMP_AUTH_KEY=secret -v chomp-data:/data chomp

On Railway:

1. Deploy from GitHub
2. Set CHOMP_AUTH_KEY env var
3. Add a volume mounted at /data (via Command Palette or railway volume add --mount-path /data)

The Dockerfile reads Railway's PORT env var automatically.

Environment Variables

Variable	Default	Description
CHOMP_DB_PATH	~/.chomp/foods.db	Database file path
CHOMP_PORT	3000	HTTP server port
CHOMP_HOST	127.0.0.1	HTTP server bind address
CHOMP_AUTH_KEY	(none)	Authentication key for HTTP server
CHOMP_SERVER_URL	(none)	Remote server URL (enables client mode)
PORT	(none)	Railway-injected port (maps to CHOMP_PORT)

Tech Stack

• Language: Rust (fast, single binary, no runtime)
• Database: SQLite via rusqlite (portable, no server)
• HTTP: Axum + Tower (SSE, REST API, static files)
• Charts: Chart.js (dashboard)

File Locations

• DB: ~/.chomp/foods.db (local), /data/foods.db (Docker/Railway)

Prior Art

• MyFitnessPal — bloated, cloud-only, privacy concerns
• Cronometer — good but no API, no CLI
• noms (Python) — nutrition data but not tracking-focused

License

MIT