Self-hosted recipe management, honestly reviewed. No marketing fluff, just what you get when you run it yourself.

TL;DR

What it is: GPL-3.0 self-hosted recipe manager built on the philosophy that the app should do the structural work, not the user [README].
Who it’s for: Home cooks and food-focused founders who want to own their recipe collection without feeding it to a SaaS platform, especially those importing recipes from the web in bulk [README][1].
Cost savings: No SaaS pricing to compare against — Vanilla Cookbook has no paid hosted tier. Competing tools like Tandoor offer hosted plans at $2–5/mo; self-hosting either costs roughly $5–10/mo on a small VPS [2].
Key strength: Ingredient parsing intelligence that handles unit conversion, recipe scaling, and nutrition detection automatically from raw text — no manual data entry [README].
Key weakness: 145 GitHub stars as of this writing, no third-party reviews, and a solo developer working on it in spare time. Tandoor, the category leader, sits at 8,209 stars. This is a young, small project and carries the risks that come with that [1][2].

What is Vanilla Cookbook

Vanilla Cookbook is a self-hosted recipe manager with an unusual design principle: most recipe apps ask you to organize and structure your data; Vanilla Cookbook tries to do that work for you [README].

The problem it’s solving is real. Web recipes are chaos — inconsistent unit formats, inline instructions buried in ingredient lists, fractions written as text strings, nutrition data scattered across paragraphs. Most self-hosted recipe apps either ignore this and store whatever you paste in, or force you through a structured entry form that feels like accounting software for dinner. Vanilla Cookbook’s answer is text-string parsing: you import a recipe from a URL or paste in raw text, and the app figures out ingredient quantities, units, instructions, and nutritional structure on its own. The developer describes this as “complexity under the hood” [README].

The project is built in SvelteKit with a SQLite database and deployed via Docker. It’s a one-person spare-time project — the developer’s own words in the README: “I honestly don’t have the time to build it without the use of AI tools.” That’s an unusually honest disclosure, and the rest of the README has the same candid tone. All architecture was hand-built and reviewed before release; the AI assistance is on the coding side, not the feature design side [README].

At 145 GitHub stars with 9 forks, this is genuinely a small project. It was added to AlternativeTo in February 2026 and has no published reviews there yet [1]. That’s the honest baseline: interesting concept, early-stage community, limited production validation outside the developer’s own use.

Why People Choose It

This is where the usual review format runs into a wall: there are essentially no third-party reviews of Vanilla Cookbook. AlternativeTo lists it as an alternative to ManageMeals, Recipya, kitshn, Grocy, and about 47 other tools, but has zero published user reviews as of this writing [1]. The GitHub repository has 873 commits and an active project kanban, which signals ongoing development rather than abandonment — but the absence of community testimonials means the “why people choose it” question can only be answered through the README’s positioning and the gap it claims to fill.

The positioning argument is coherent. The developer explicitly frames the philosophy as anti-cluttered: most recipe managers require effort to get to cooking. Vanilla Cookbook’s bet is that removing data-entry friction is a competitive advantage for the casual home cook who wants to collect recipes from the web without becoming a recipe-database administrator [README].

The LLM assist features add a layer that most competitors haven’t reached: when a recipe page lacks proper Schema.org markup — which is most of them — the LLM can parse the raw HTML to extract structure. That’s a genuinely useful fallback for the long tail of recipe sites that don’t follow structured data standards [README].

What you won’t find here is the kind of multi-year user testimonial base or Reddit thread consensus that validates whether the parsing actually works reliably in practice across thousands of sites. That evidence simply doesn’t exist yet publicly.

Features

Based on the README and project documentation:

Core recipe management:

URL scraping via browser bookmarklet or paste-in (hundreds of supported sites) [README]
Raw text paste — useful for recipes from books, emails, or non-scraper-friendly pages [README]
SQLite storage with export support for data portability [README]
Multiple import formats [README]
Full-text and fuzzy search [README]

Smart ingredient parsing:

Automatic quantity, unit, and ingredient separation from raw strings [README]
US volumetric to imperial and metric weight conversion across thousands of ingredients [README]
Inline temperature conversion, fractions, and ranges handled automatically [README]
Recipe scaling — adjust serving count and ingredient quantities update accordingly [README]
Nutrition text parsed into structured tables with per-serving detection and scale-aware values [README]

LLM Assist (requires API key):

Supported providers: Google, OpenAI, Anthropic, and Ollama for local inference [README]
Scrape assist: LLM parses HTML when Schema.org data is absent or malformed [README]
Text analysis: paste raw text, LLM structures it into recipe format [README]
Ingredient tidy-up: strips verbose ingredient strings down to clean format [README]
Directions summarize: condenses wordy instructions [README]
Language detection and translation [README]
Recipe generation from prompt (e.g., “Original pesto genovese”) [README]
Image analysis: photograph up to three images and extract recipe data [README]
Semantic search: associative meaning search (“Japanese” surfaces noodles, miso etc.) [README]

Platform and access:

Progressive Web App — installable on mobile from a browser [README]
Share URL via browser to trigger scraping; select-and-share text parsing on mobile [README]
Shopping list with recipe integration and history [README]
SSO support [README]
Docker and Docker Compose deployment [README]
apt installation method available [README]

The feature list is genuinely ambitious for a solo spare-time project. The LLM integration in particular is more thoughtfully designed than many tools that bolt on an “AI button” — the use cases (HTML parsing fallback, image-to-recipe) address real pain points in the recipe import workflow.

Pricing: Self-Hosted Math

Vanilla Cookbook has no hosted SaaS tier. It’s GPL-3.0 software you run yourself, and the software cost is $0 [README][1].

Your actual costs to run it:

A Linux VPS: $4–8/mo (Hetzner CX22, Contabo, or similar)
Domain name (optional): $10–15/year
LLM API costs: variable, only if you use AI features. Ollama runs locally for $0 on the same VPS or your own hardware.

How this compares to the closest self-hosted alternatives:

Tandoor, the category leader with 8,209 stars, is also free to self-host. It offers a hosted tier at $2–5/mo [2]. Mealie, another major alternative, follows a similar model.

There’s no SaaS bill to escape here — this isn’t an Activepieces vs. Zapier situation where self-hosting saves you hundreds per year. The financial case for Vanilla Cookbook specifically is that it’s free software, which is table stakes in the self-hosted recipe category. The actual differentiation is on features and philosophy, not pricing.

Deployment Reality Check

The install path is Docker Compose, with an apt option also listed. SQLite as the database removes the need to manage a separate Postgres instance, which is a meaningful simplification for small self-hosted setups.

What you need:

A Linux VPS with 1–2GB RAM minimum (SQLite is lean, SvelteKit SSR doesn’t require much)
Docker and docker-compose
A domain and reverse proxy (Caddy or nginx) if you want HTTPS
No separate database setup — SQLite is bundled

What the documentation situation looks like: The project has a dedicated ReadTheDocs site at vanilla-cookbook.readthedocs.io. The README links to it consistently and the project has docs for installation, ingredient handling, import formats, AI assist, and mobile PWA setup [README]. For a 145-star spare-time project, the documentation coverage is better than typical.

Honest concerns:

First: bus factor. This is a one-developer spare-time project. The GitHub kanban is active and the commit history shows 873 commits [1], but if the developer moves on, you’re running unmaintained GPL-3.0 software. That’s fine if you know what you’re doing, less fine if you’re planning to rely on it indefinitely.

Second: no production scale data. There are no testimonials or Reddit threads describing this running with years of recipe data across a household. Tandoor has an established community with known failure modes. Vanilla Cookbook doesn’t yet [1][2].

Third: LLM API key dependency for advanced features. The semantic search, image analysis, and scrape-assist features require an API key from Google, OpenAI, or Anthropic — or a local Ollama setup. The core scraping and parsing work without it, but the feature list shrinks considerably if you want to avoid LLM API costs [README].

Realistic setup time for a comfortable Docker user: 30–45 minutes. For someone following a guide without prior Docker experience: 2–4 hours.

Pros and Cons

Pros

Zero-friction import philosophy. The core design bet — that parsing should happen automatically without structured data entry — is correct and differentiated [README].
LLM integration that solves real problems. Using an LLM as a fallback HTML parser for malformed recipe pages is practical, not gimmicky. Image-to-recipe is genuinely useful [README].
SQLite simplicity. No separate database container, no Postgres tuning, no backup complexity beyond copying a file.
PWA with mobile share integration. The ability to share a URL from your phone’s browser directly into the scraper is the kind of detail that makes daily use pleasant [README].
Ollama support. All AI features can run locally with no API costs if you have the hardware [README].
Active development. 873 commits, a project kanban with visible roadmap, and regular releases (24 releases on GitHub) [1][README].
Honest developer communications. The README acknowledges AI tool assistance, spare-time constraints, and reviewed-before-release standards — that kind of transparency is worth noting [README].

Cons

145 stars. This is the category’s loudest signal. Tandoor has 8,209 stars. Mealie has similar scale. Vanilla Cookbook is an unproven tool in a reasonably crowded category [1][2].
No user reviews anywhere. Zero published reviews on AlternativeTo, no Reddit threads found, no blog posts. You’re flying blind on real-world reliability [1].
One-developer bus factor. A spare-time solo project with no organizational backing. If development stops, you’re maintaining it yourself [README].
GPL-3.0, not MIT. If you want to embed this in a commercial product, GPL-3.0 has copyleft implications. For personal or internal use this doesn’t matter, but it’s worth knowing [1].
LLM features need a separate setup. The most interesting features require either paid API keys or running Ollama separately [README].
No validated hosting tier. Unlike Tandoor’s $2–5/mo cloud option [2], there’s no managed option if self-hosting is a blocker.
Community support is thin. No active forum, Discord, or large community to troubleshoot against. The docs site and GitHub issues are what you get [1].

Who Should Use This / Who Shouldn’t

Use Vanilla Cookbook if:

You have Docker experience and want a clean, minimal recipe manager where you’re not fighting the interface to add a recipe.
You import a lot of recipes from URLs and want the best available scraping-plus-parsing pipeline, including LLM fallback for malformed pages.
You want local LLM integration (Ollama) for AI features without paying OpenAI per query.
You’re technically comfortable maintaining a spare-time open-source project and willing to contribute if something breaks.
You’re an early adopter who wants a younger project with an interesting roadmap and doesn’t need five years of production validation.

Skip it (use Tandoor instead) if:

You want a proven, community-backed tool with thousands of users and an active forum.
You want meal planning, a calendar view, and tagging/filter infrastructure that scales to a large recipe library [2].
You want a managed cloud option with a small monthly fee instead of self-hosting.
Your household’s recipe ops are critical enough that bus factor matters.

Skip it (use Mealie instead) if:

You want a polished UI with a large community and wide recipe site support.
You want household user management and recipe sharing as first-class features.

Alternatives Worth Considering

From the AlternativeTo alternatives list and the broader self-hosted recipe category:

Tandoor — the category benchmark. 8,209 GitHub stars, Django-based, active community, optional hosted tier at $2–5/mo, strong feature set including meal planning and Nextcloud/Dropbox sync. Steeper setup than Vanilla Cookbook [2].
Mealie — arguably the most popular self-hosted recipe manager right now. Clean UI, strong URL scraping, household management, active community. Closest competitive match in terms of philosophy.
Recipya — listed as a common alternative on AlternativeTo [1]. Smaller community, more recent project.
Grocy — broader household management including pantry tracking and shopping. Overkill if you only want recipes, but powerful if you want inventory too [1].
kitshn — mobile-first recipe manager, also listed as an alternative [1].
ManageMeals — simpler option listed frequently alongside Vanilla Cookbook [1].

For someone who just wants to own their recipe collection and escape the chaos of browser bookmarks, the realistic shortlist is Vanilla Cookbook vs. Mealie vs. Tandoor. Pick Vanilla Cookbook if the parsing automation and LLM fallback are compelling and you’re comfortable with early-stage software. Pick Mealie for a safer choice with community backing. Pick Tandoor if you want meal planning and the most established option in the category.

Bottom Line

Vanilla Cookbook is a recipe manager with a genuine design insight: users shouldn’t have to be data-entry clerks to maintain their own recipe collection. The parsing engine, unit conversion, and LLM-assisted scraping are thoughtfully built and address real frustrations with web recipes. The AI features — specifically using an LLM to parse malformed HTML and photograph-to-recipe extraction — put it ahead of most alternatives in the scraping pipeline.

The honest caveat is that at 145 stars with no third-party reviews, “thoughtfully designed” is based on reading the README, not years of community use data. Tandoor has 57× more stars and a community that can tell you what breaks in production. Vanilla Cookbook cannot offer that yet. If you’re technically comfortable, willing to accept early-adopter risk, and the parsing philosophy resonates, it’s worth spinning up. If you need a recipe manager for serious household use and reliability matters, Tandoor or Mealie are the safer bets until this project builds more of a track record.

Sources

AlternativeTo — Vanilla Cookbook (149 stars, GPL-3.0, added Feb 2026). https://alternativeto.net/software/vanilla-cookbook/about/
AlternativeTo — Tandoor (8,209 stars; competitor comparison, pricing reference). https://alternativeto.net/software/recipes-1/about/

Primary sources: