📝 nwhitehouse teaches the table reviewer to read the underlying documents 4 commits 19d ago contract-reviewsearch draft
The fork's tabular-review chat can now pull answers from the source PDFs, not just the cells in the table.
nwhitehouse/mike
nwhitehouse is turning Mike into Olava - a hosted legal-AI product for a hand-picked roster of law firms, built around tabular document review and a self-hosted reasoning model.
This fork, run by nwhitehouse, has quietly become one of the more ambitious Mike derivatives in the wild. It started as Finch, rebranded to Olava, and is now being prepared as a hosted product with signups locked to three named firms at the database level. The README doubles as a deployment playbook, and a sweeping security pass - auth, row-level access, rendering - signals that real client data is on the way.
The product centre of gravity is tabular review: running the same question across thousands of documents, with verified-cell state, real filters, clickable citations back to the source page, hide/rerun/delete column controls, and a drill-in document view for inspection. The bulk engine behind it has been rebuilt as a durable job system that survives tab closes, backend restarts, and four-hour proxy timeouts.
On the AI side, nwhitehouse has locked the fork to a single engine - a self-hosted Qwen-based reasoning model called Olava - and built a five-pass research orchestrator around it, with web search via Brave, a public US legal-corpus search tool, vision-mode PDF reading, persistent in-turn memory, and bounded tool-use loops. Thinking depth is a user-facing dial; the model's inner monologue is collapsed by default.
What's in it
- Tabular review, end to end Run one question across thousands of documents, mark cells verified, filter to what still needs eyes, hide or rerun columns, and drill into any single document without losing your place in the table.
- Citations you can actually click Extracted table cells and inline chat references jump straight to the exact page and phrase in the source PDF - and attachment chips in user messages open the doc too.
- Self-hosted reasoning model Olava is wired in as the fork's single AI engine: a self-hosted Qwen-class model with live streaming, tool-call recovery, and user-controllable thinking depth.
- Five-pass research orchestrator Tick a legal database or the web as a source and a multi-pass research engine kicks in automatically, with bounded tool dispatches so the assistant can't spin its wheels.
- Live legal and web sources Opt-in tools pull from four open US federal legal sources and from Brave web search, surfaced as clickable result cards in the chat.
- Vision-mode PDF reading Uploaded PDFs are shown to the model as page images rather than extracted text, pre-loaded the moment the file lands.
- Built to be hosted Dark mode on semantic tokens, a structured audit log of every AI turn, a deployment-grade README, persistent chat history, and a security hardening pass - all the plumbing of a real product.
Direction
discoverychat-uibranding
Activity
Launch-week plumbing brings the fork up as a hosted product, with signups locked to three named firms at the database.
📝 nwhitehouse wires up clickable citations in extracted tables 4 commits 19d ago contract-reviewdiscovery draft
Reviewers can now jump straight from a table cell to the exact page and phrase in the source document.
📝 nwhitehouse gives tabular review a verified column and real filters 4 commits 19d ago contract-reviewworkflow draft
Reviewers can now mark individual cells as checked, and filter the table down to what still needs eyes.
📝 nwhitehouse gives tabular review proper column controls 3 commits 19d ago discoverycontract-review draft
Once a review grows past a handful of columns, reviewers need to hide, rerun, or delete them without nuking the whole table - nwhitehouse just shipped that.
A fork-wide theme overhaul that swaps hard-coded colours for reusable tokens - so dark mode comes almost for free.
📝 nwhitehouse rebuilds Mike's bulk-document engine for scale 1 commit 20d ago discoveryinfrastructure draft
Tabular review - the bit that runs the same question across thousands of documents - now survives tab closes, backend restarts, and the kind of proxy timeouts that used to kill a four-hour run.
The fork now lets the underlying model think hard, think lightly, or skip thinking entirely - and hides the messy reasoning by default.
📝 nwhitehouse puts guardrails on the AI's tool-using loop 1 commit 20d ago infrastructureworkflow draft
A small safety layer that stops the assistant from spinning its wheels when it gets stuck calling the same tools over and over.
Every multi-step turn the AI takes now leaves a structured trail you can actually query after the fact.
📝 nwhitehouse teaches Mike to remember the conversation 3 commits 20d ago chat-uiinfrastructure draft
The chatbot used to re-read your documents from scratch on every reply. Not anymore.
Instead of extracting text from uploaded PDFs, this fork shows the document to the AI as images - and pre-loads them the moment you upload.
A small UX fix that turns a dead label into a working shortcut into the doc viewer.
📝 nwhitehouse tries a sturdier citation pipeline, then walks it back 2 commits 23d ago searchchat-ui draft
An experiment to make AI citations more reliable on a small in-house model hit a wall - but the surrounding polish stuck.
📝 Greeting picks up displayName once profile loads (no public page) 1 commit 23d ago not yet rewritten
📝 nwhitehouse builds a five-pass research engine for a small model 1 commit 23d ago searchchat-ui draft
The biggest single feature in this fork: a research pipeline that fires automatically the moment you tick a legal database or the web as a source.
📝 nwhitehouse makes every source visible, even when the AI goes quiet 2 commits 23d ago searchchat-ui draft
A small but pointed fix: search results now show up as clickable cards in the chat, whether or not the model bothers to write them up.
📝 nwhitehouse bolts web search onto Mike with a one-click toggle 1 commit 23d ago searchchat-ui draft
A globe icon next to the send button lets users pull in live web results alongside legal sources, message by message.
📝 nwhitehouse keeps the AI typing live during tool calls 2 commits 23d ago chat-uiinfrastructure draft
When the assistant pauses to look something up, it no longer goes silent the whole time.
📝 nwhitehouse catches a silent data loss bug in chat history 1 commit 23d ago chat-uiinfrastructure draft
User messages were quietly vanishing from the database - and nobody noticed because the screen looked fine.
Show 8 more
A new search tool lets the assistant pull live citations from four open federal sources - but only when the user explicitly turns them on.
A single sweeping hardening commit signals nwhitehouse is preparing to host actual user data.
# Vercel Web Analytics Installation
📝 nwhitehouse writes the operations manual for Mike 1 commit 25d ago infrastructureknowledge-management draft
The fork now ships with a 377-line README that doubles as the deployment playbook for a live legal-AI product.
The fork formerly known as Finch is now Olava - and it only talks to one model.
The fork adds a third AI option alongside the usual two - a self-hosted model the team can run on its own hardware.
📝 nwhitehouse turns tabular review into a real verify workflow 1 commit 27d ago contract-reviewdiscovery draft
A new drill-in view lets reviewers inspect, edit, and search inside each document without bouncing back to the table.
📝 Wire Vercel Web Analytics into the Next.js frontend (no public page) 0 commits — not yet rewritten
Threads of work (detailed view)
nwhitehouse teaches the table reviewer to read the underlying documents
The fork's tabular-review chat can now pull answers from the source PDFs, not just the cells in the table.
nwhitehouse opens Mike to a hand-picked roster
Launch-week plumbing brings the fork up as a hosted product, with signups locked to three named firms at the database.
nwhitehouse wires up clickable citations in extracted tables
Reviewers can now jump straight from a table cell to the exact page and phrase in the source document.
nwhitehouse gives tabular review a verified column and real filters
Reviewers can now mark individual cells as checked, and filter the table down to what still needs eyes.
nwhitehouse gives tabular review proper column controls
Once a review grows past a handful of columns, reviewers need to hide, rerun, or delete them without nuking the whole table - nwhitehouse just shipped that.
nwhitehouse ships dark mode the right way
A fork-wide theme overhaul that swaps hard-coded colours for reusable tokens - so dark mode comes almost for free.
nwhitehouse rebuilds Mike's bulk-document engine for scale
Tabular review - the bit that runs the same question across thousands of documents - now survives tab closes, backend restarts, and the kind of proxy timeouts that used to kill a four-hour run.
nwhitehouse puts the AI's inner monologue on a diet
The fork now lets the underlying model think hard, think lightly, or skip thinking entirely - and hides the messy reasoning by default.
nwhitehouse puts guardrails on the AI's tool-using loop
A small safety layer that stops the assistant from spinning its wheels when it gets stuck calling the same tools over and over.
nwhitehouse gives Mike a flight recorder
Every multi-step turn the AI takes now leaves a structured trail you can actually query after the fact.
nwhitehouse teaches Mike to remember the conversation
The chatbot used to re-read your documents from scratch on every reply. Not anymore.
nwhitehouse teaches Mike to read PDFs with its eyes
Instead of extracting text from uploaded PDFs, this fork shows the document to the AI as images - and pre-loads them the moment you upload.
nwhitehouse makes attachment chips actually open the document
A small UX fix that turns a dead label into a working shortcut into the doc viewer.
nwhitehouse tries a sturdier citation pipeline, then walks it back
An experiment to make AI citations more reliable on a small in-house model hit a wall - but the surrounding polish stuck.
nwhitehouse builds a five-pass research engine for a small model
The biggest single feature in this fork: a research pipeline that fires automatically the moment you tick a legal database or the web as a source.
nwhitehouse makes every source visible, even when the AI goes quiet
A small but pointed fix: search results now show up as clickable cards in the chat, whether or not the model bothers to write them up.
nwhitehouse bolts web search onto Mike with a one-click toggle
A globe icon next to the send button lets users pull in live web results alongside legal sources, message by message.
nwhitehouse keeps the AI typing live during tool calls
When the assistant pauses to look something up, it no longer goes silent the whole time.
nwhitehouse catches a silent data loss bug in chat history
User messages were quietly vanishing from the database - and nobody noticed because the screen looked fine.
nwhitehouse wires Mike into the public US legal corpus
A new search tool lets the assistant pull live citations from four open federal sources - but only when the user explicitly turns them on.
nwhitehouse's fork gets a real security pass
A single sweeping hardening commit signals nwhitehouse is preparing to host actual user data.
nwhitehouse writes the operations manual for Mike
The fork now ships with a 377-line README that doubles as the deployment playbook for a live legal-AI product.
nwhitehouse locks the fork to a single AI engine
The fork formerly known as Finch is now Olava - and it only talks to one model.
nwhitehouse wires in a self-hosted reasoning model
The fork adds a third AI option alongside the usual two - a self-hosted model the team can run on its own hardware.
nwhitehouse turns tabular review into a real verify workflow
A new drill-in view lets reviewers inspect, edit, and search inside each document without bouncing back to the table.
Pull requests (detailed view)
🟢 Open (1)
by
vercel[bot] · opened 24d ago