Data Migration Engine · Sprint 2 · 19 August – 2 September 2026

Sprint 2

End to end in dev

Draft · numbers as of 1 Sep
Lawhive

The sprint in one picture

A real matter made the whole trip

1real v2 matter 36files 92staff records
v2 database
the real Lawhive Legal data, on dev
✓ copied out whole
Stage 1Sourcing
Import bucket · raw/
source data, exactly as it came
✓ landed
Stage 2Transformation
Import bucket · target/
ready-to-load packets, made by one click
✓ exported
Stage 3Importing
Platform v3
the matter, its people, and every file, in the product
✓ imported on dev
Stage 4Activation
A working firm
logins, previews, AI search, case context
being wired up now

Sprint 1 built the parts. Sprint 2 connected them on a deployed environment. One honest gap: the 36 files still await their virus-scan results, held up by a storage security-policy decision set for 2 September. The fix is written and reviewed.

Lawhive

Stage 1 · Sourcing · Daniele

Source system → raw/

The records
v2 database, copied out whole
Everything the firm's old system knows, in one dump. Rehearsals read a copy of the data, never the live system.
→
The files · new this sprint
Bulk file copy, with proof
An automated job copies the firm's actual file contents across, fingerprinting every file (SHA-256) so we can prove nothing was corrupted. It resumes where it stopped, and an operator starts it with a single input.
→
Landing zone
Import bucket · raw/
A locked-down storage area with a written retention policy, one folder per firm. Source data stays untouched from here on.

Where it stands

Records copied into raw/Local ✓Dev ✓ provenProd
Files copied in bulkthe approach validated against the real v2 estate: 3 TB, 4 million filesn/aDev ◐ built, first run nextProd
Lawhive

Stage 2 · Transformation · Cesar & Daniele

raw/ → packets, one click

Read
Every kind of v2 record
Matters, contacts, files, notes, tasks and the rest, read from the raw dump into the data warehouse.
→
Reshape
Into 17 agreed shapes
Whatever the source system, its data is reshaped into the same 17 standard forms. A future source like LEAP slots in without rebuilding anything downstream.
→
Export
One click, whole caseload
A single job reshapes a firm's entire caseload and writes the ready-to-load packets into target/, unattended.
→
Handoff
target/ packets
Exactly the shape the importer expects. A rehearsal run ends with a readable summary of everything that failed its checks.

Where it stands

Whole-caseload reshape + exporton Aug 19 this barely existed; 19 merged changes later it is a working pipelineLocal ✓Dev ✓ provenProd
The 17 shapes stay in sync automaticallyevery day the pipeline's shapes are re-checked against the platform's; any drift opens a change for reviewRunning daily ✓n/a
Matters classified live / recent / archiveper-firm cutoff date; this drives how much activation work and cost each matter getsLocal ✓Dev ✓ landedProd
Lawhive

Stage 3 · Importing · Adolfo & Lukas

Packets → v3 rows and files

Read
target/ packets
Read in order: a matter first, then everything that belongs to it.
→
Live on dev this sprint
The import service
Deployed in the sprint's first two days. It checks every row before writing it, records where each row came from, and can undo an entire run. Dry runs never touch real data.
→
The records
Into the product's databases
Matters, contacts, notes, tasks, file records, and the firm's members.
→
The files · new this sprint
Every file copy, tracked
Each file is copied into the matter store, virus-scanned, and quarantined if suspect, with every step recorded in a ledger operators can drive and watch from the admin screen.

Where it stands

Records + members into v3a 714-matter rehearsal in Sprint 1, the real matter this sprintLocal ✓Dev ✓ provenProd
Files: copy, scan, quarantineall 36 files copied on dev; virus-scan results held up by the security-policy decision, fix writtenLocal ✓Dev ◐ scans blockedProd
Lawhive

Stage 4 · Activation · new this sprint

Imported rows → a working firm

People
Logins & invites
An imported person becomes a real user: a working login, a welcome email, and a clean undo if the run is torn down. The machinery and its admin panel are built and merged.
+
Files
Previews & AI search
One operator action wakes a run's files, newest matters first: document previews and AI search indexing, running on a dedicated lane that never slows down live customers.
+
Matters
Case context
Built per matter once its files are done, so the AI can answer questions about a migrated matter the way it does for a native one.

Where it stands

Member loginsLocal ✓Dev ◐ deployed, pilot nextProd
The dedicated processing lanea real document went through the whole lane into search while live traffic was untouchedLocal ✓Dev ✓ provenProd
File wake-up + case contextseven changes built, reviewed and approved, ready to mergeLocal ✓Dev ○ in reviewProd
Lawhive

Around the edges

Also this sprint

Nine design docs
The files-and-activation planning arc: measured dry-run lessons, the file-migration framework and its agreed tech plan, firm provisioning, staff activation, file activation, and the economics of activating at scale.
One trace per file
For every file the document pipeline processes, one page now shows its whole journey: every stage it went through and what each produced. Proven on three live runs. Migration waves become individually inspectable. In review
The virus-scan incident
A security-policy change on dev accidentally locked out the virus scanner itself, so new files could never pass. The squad diagnosed it and wrote the fix; the policy direction is decided 2 September. Decision Sep 2
Lawhive
Demo · importing a firm's people Recorded on dev · 28 Aug
Lawhive

What's next

Next sprint

Pending sprint planning
Close the file loop
Settle the virus-scanning policy, re-scan the demo's 36 files, and run the first full-book file copy. Plus the checks that guarantee packets only reference files that really arrived.
First activation in dev
Merge the wake-up chain, give the demo run's files previews and search, build its case context, and run the one-member login pilot.
Scale toward the full load
The full Lawhive Legal load on dev, run reports a business reader can act on, and measurements that predict how long a firm's run will take.
Production preparation
Finish per-file tracing, then start the production checklist for the dedicated lane: workers, configuration, signing keys.
Lawhive

Thank you

Full report → Sprint 2, Data Migration Engine Design docs → firm-migrations Linear → Build & Test Migration Engine
Lawhive