Sabitlenmiş Tweet
Eugene Mi
1.4K posts

Eugene Mi
@emironic
Building open-source https://t.co/QFYJxUUOgz - up to x10 faster review of important documents.
Katılım Ekim 2010
170 Takip Edilen118 Takipçiler

@ProductHunt Check invoices x5 times faster revdoku.com (open-source)
English

@pmitu open-source app for visual review and check of important documents in finances, legal, logistics. Feedback appreciated! github.com/revdoku/revdoku
Use case: check invoice from vendor vs agreement
English

@emironic AGPL is worth flagging for enterprise users — if you deploy this as a service, modifications must be released under AGPL. Fine for internal self-hosted use, but friction if customers want to wrap it in a product. What's the model layer — local inference or API-backed?
English

I'm building Revdoku ( revdoku.com ), an open-source ( github.com/revdoku/revdoku , AGPL) tool for reviewing important documents (invoices, contracts, shipping paperwork, financial filings) with AI assistance.
The goal is to make a careful read-through roughly 3-5x faster without losing the reviewer in the loop! And, very important, to make it visual-first (unlike ChatGPT or Claude).
Why I built this: checking an invoice or agreement is fine with a chat-based AI tool, but I wanted:
- a list of reusable rules, not ad-hoc prompts
- to see with my own eyes exactly where the issue or incorrect calculation is
- when needed, also cross-check against the referenced document and see the exact matching places in both
- when a new revision is provided, re-run the same checks automatically and track all change outside the rules.
- the human must always stay in full control! While AI checks against rules, verifies the math etc BUT the operator always can edit results and override any flag, add manual checks.
What it does today:
- Create checklists (from template or turn an agreement/policy into a checklist)
- Open a single doc (or a set of docs) or just forward an email with attachment
- Run AI review via AI to detect issues, and it highlights exactly where issues are in the document (and where passed rules are!)
- Automatically runs cross-document checks, e.g. invoice vs. quote or agreement, or term sheet vs. final contract, flagging mismatches
- Reviewer can accept/reject each check with the reasoning shown inline rather than in a chat sidebar
- Keyboard shortcuts for fast and quick navigation. Report can be shared via link.
- Custom js based scripting, eg group and sum all amounts by category
- Works with both cloud AI and local LLMs (tested with Gemma 4, Qwen 3.6)
Tech Stack: Ruby on Rails, Sqlite, React, TS
I'd genuinely like feedback on:
- Cross-document checking: is the way mismatches are displayed (with a small inline viewer) - is it useful?
- Self-hosting: what would you need to actually run this on your own infra for sensitive docs? Currently it downloads and runs a Docker image from GitHub by default. Runs fine on MacBook Pro, can also run on a single Ubuntu server
- Industry fit: if you review documents professionally, what's the workflow this is missing? #buildinpublic
English

@code_codeforge github.com/revdoku/revdoku - open-source platform for visual review of important documents using AI.
English


@jitbit @cursor_ai Switched from Cursor (was a paid subscriber) after it repeatedly hangs the laptop on replacing strings in 1K+ files (VS Code and Antigravity not).
English

So @cursor_ai just deleted my Reddit post criticizing their vision. 177 upvotes. 76 comments. Mostly people agreeing. That says more than the post did.


English

@heyblake Add Summarize widget to your website (open-source) aicw.io/summarize-widg…
English


@dev_HarshMohan @NanouuSymeon not sure, depends on what it can do. but if something can run in browser locally but no download then it is interesting for sensitive documents
English

Yeah, it's a browser limitation - browsers aren't built to handle very large files in memory, and pushing past ~300MB causes crashes.
That's exactly why I'm building a desktop app - no browser constraints, so no hard limits on file size plus batch processing. Would that be something you'd use?
English

@dev_HarshMohan @NanouuSymeon is it a hard limit set by a browser, not a system? any way to increase it via some options?
English

Good questions! Here's where the current limits are:
1GB files: Not currently - the browser memory limitation is real. Practically, you're looking at 200-300MB max depending on your RAM.
Why: Client-side processing means everything loads into your browser's memory. A 1GB PDF would likely freeze/crash the tab.
The trade-off: Client-side = privacy + speed for typical use cases (2-5 files, <50MB each) while Server-side = can handle massive files and batch operations
English












