Wesley Thompson
ChopMedia LLC
LEXARA-WTX-2026-001
Private Legal Intelligence · Demo
Welcome to Lexara
A private AI paralegal
that never acts without permission
This demo walks you through a real day in Lauren Hosler's practice. Every email you see is processed, classified, and surfaced for her review — on her server, with her data, going nowhere else.
The attorney
Lauren Hosler
Solo practitioner, Utah Bar. Family law, civil litigation. Her entire practice flows through Outlook. Three active cases. Two deadlines this week. One email she hasn't seen yet that changes everything.
What we'll see
1. An email arrives and gets classified in under 600ms
2. The rules engine finds a case number and FRCP deadline instantly
3. Eight micro-models classify, extract, and route
4. Lauren approves actions from her queue
5. Search recalls anything from the past instantly
6. The commitment tracker surfaces a forgotten promise
Demo note: All data is realistic but fictional. Lauren Hosler, Synergy Law PLLC, and all case details are illustrative only. The system processes locally — nothing shown here touches any external API.
Click "Begin demo" to start →
Step 1 — Intake
An email arrives
7:42am. Lexara's Graph API poll picks up a new message in Lauren's "Court Filings" folder from the U.S. District Court for the District of Utah. The original email is untouched.
What Lexara sees: New message in authorized folder. Source: Outlook via Graph API. Sender: ecf@utd.uscourts.gov. Before any AI runs, the envelope is normalized and handed to the rules engine.
The rules engine runs first — free, instant, zero AI cost
Step 2 — Rules Engine
Deterministic extraction
runs in microseconds · $0 · 100% confidence
Before any AI model loads, the rules engine scans for known patterns. Court domain, case number, FRCP citations, threatening language. What it finds, the classifiers treat as fact.
CT
Court domain confirmed
ecf@utd.uscourts.gov
confidence: 1.0 — exact registry match
CN
Case number extracted
2:24-cv-00847-JCB
confidence: 1.0 — federal pattern regex
DL
Explicit deadline found
"within 14 days of this Order"
confidence: 0.95 — deadline trigger pattern
FR
FRCP 37(b) citation
Sanctions reference — implicit threat
confidence: 1.0 — rule citation pattern
DL
Second deadline found
"status report within 30 days"
confidence: 0.90 — explicit deadline language
TH
Threat language flagged
"sanctions" detected
urgency auto-set: severity ≥ 4
Rules result: 6 extractions. Case number confirmed — case_matcher classifier will skip (case already known). Email type confirmed as court with confidence 1.0. Two deadlines found. Sanctions threat escalates urgency to 4. All results embedded in the envelope before a single model loads.
Now the AI classifiers run on what rules couldn't determine
Step 3 — Classification
Eight micro-models, one email
Privilege check runs alone first. Then email type + urgency + tone run in parallel. Then deadline, case, and action run sequentially. Total: under 600ms.
1
Privilege flag
qwen3:0.5b · :8014 · runs first, always
Result
Not flagged
Confidence
0.97
Action
Continue pipeline
2
Email type
qwen3:0.5b · :8010 · parallel with urgency + tone
Type
court
Confidence
1.0
Note
Rules confirmed domain
3
Urgency detector
qwen3:0.5b · :8011 · parallel
Urgent
YES
Severity
4 / 5
Reason
Court order + sanctions threat
4
Deadline extractor THINKING MODE
qwen3:4b · :8012 · /think prefix · sequential
Model reasoned: "The order was issued today. '14 days' → absolute date is [today+14]. '30 days' → [today+30]. FRCP 37(b) confirms this is a compliance deadline not a discovery deadline."
Deadline 1
+14 days · produce documents
Deadline 2
+30 days · file status report
Confidence
0.98
5
Case matcher
qwen3:1.7b · :8013 · SKIPPED — rules found case number
Case
Henderson v. Smith
Matter #
2024-0847
Source
rules_engine (regex)
6
Action detector
qwen3:1.7b · :8015
Action needed
YES
Type
task + 2× calendar
Suggested
Begin document production immediately
Watch the classifiers run...
Step 4 — Approval Queue
Lauren decides what happens next
Everything the system wants to do is waiting here. The alert fired immediately (severity 4). The rest wait for Lauren. Nothing executes automatically. Click Approve or Deny on each item.
Alert already fired: SMS sent to Lauren's phone at 7:42am — "URGENT: Court order received in Henderson v. Smith — 14-day document production deadline. Check Lexara."
calendar
Add deadline: Produce documents — Henderson v. Smith
Court order #47 · FRCP 37(b) compliance · 14 days from today's order date · failure may result in sanctions
✓ Added to calendar
calendar
Add deadline: File status report — Henderson v. Smith
Court order #47 · 30 days from today's order date · required by Judge Bennett
✓ Added to calendar
task
Begin document production — Requests for Production 3-7
Review documents responsive to RFP 3-7, coordinate with client, produce within 14 days per Order #47
✓ Task created
Approve the actions above, then continue
Step 5 — Dashboard
Lauren's full picture
Every email processed, every deadline tracked, every commitment monitored. One view of the entire practice.
3
Pending actions
1
Urgent today
4
Upcoming deadlines
$0
Per email cost
Recent messages
ECF Notification · utd.uscourts.gov
Order on Motion — Henderson v. Smith
court urgent
James Chen · opposing counsel
RE: Deposition scheduling — Martinez matter
opposing deadline
Sarah Henderson · client
Documents you requested — Henderson case
client case match
Utah State Courts · no-reply@utcourts.gov
Hearing scheduled — In re Estate of Williams
court
Upcoming deadlines
In 2 days
Respond to deposition notice
Martinez v. Clearfield · opposing counsel James Chen
+14 days
Produce documents RFP 3-7
Henderson v. Smith · Court Order #47
+21 days
Hearing — In re Estate of Williams
Williams Estate · 3rd District Court
+30 days
File status report
Henderson v. Smith · Court Order #47
The real power: search recalls anything instantly
Step 6 — Semantic Recall
Find anything, instantly
Lauren types a natural language question. The system searches 4 months of emails and messages using vector embeddings. Not keyword search — semantic understanding.
relevance 0.94 · Henderson v. Smith
RE: Discovery responses — motion to compel warning
...if Defendant fails to produce the responsive documents by Friday, we will be forced to file a Motion to Compel and seek sanctions under FRCP 37(b). This is your final notice before we proceed...
opposing_counsel · James Chen · 6 weeks ago
relevance 0.88 · Henderson v. Smith
Demand letter — Request for Production compliance
...Your client's failure to respond to Requests for Production 3-7 within the required 30-day period constitutes a violation of FRCP 34. We demand immediate production or we will seek court intervention...
opposing_counsel · James Chen · 9 weeks ago
relevance 0.71 · Henderson v. Smith
Court Order #47 — Order on Motion to Compel
...ORDER granting in part Motion to Compel. Defendant shall produce all responsive documents... failure to comply may result in sanctions pursuant to FRCP 37(b)...
court · ECF Notification · today
What just happened: "opposing counsel threats in Henderson" was embedded as a vector and compared against embeddings of all stored emails. It found the warning letter, the demand letter, and today's court order — in order of semantic relevance. Lauren can read back the full escalation timeline in seconds.
The feature that makes Lexara irreplaceable
Step 7 — Commitment Tracker
Promises Lauren made
that the system remembered
When Lauren approves a draft email, the commitment tracker scans it for promises before it sends. Those promises are tracked until fulfilled or overdue. This is the feature no other system has.
Open commitments requiring attention
Produce documents responsive to RFP 3-7 OVERDUE
Promised to: James Chen (opposing counsel) · Henderson v. Smith
Due: 3 weeks ago · From: "I will have our client's documents to you by month end" (Lauren's email, Nov 14)
Send draft settlement agreement for review
Promised to: Sarah Henderson (client) · Henderson v. Smith
Due: This Friday · From: "I'll have a draft agreement to you by end of week" (Lauren's email, Nov 22)
Follow up on deposition availability for Martinez
Promised to: James Chen (opposing counsel) · Martinez v. Clearfield
Due: Yesterday · From: "Let me check with my client and get back to you" (Lauren's email, Nov 25)
The connection: The first overdue commitment — "I will have our client's documents to you by month end" — is exactly what led to James Chen filing the Motion to Compel. Which led to today's court order. Which triggered a 14-day production deadline under threat of sanctions. Lexara saw the promise, tracked it, and surfaced it as overdue weeks before the court got involved. This is the feature that prevents malpractice.
One more: the system health view
Step 8 — System
Everything running locally
All 8 services, all models, all data — on Lauren's private server. Nothing external. Kill switch: revoke the OAuth token. Takes 30 seconds.
Privilege flag
● Online
63ms avg
Email type
● Online
79ms avg
Urgency
● Online
61ms avg
Deadline extractor
● Online
184ms avg
Case matcher
● Online
148ms avg
Commitment tracker
● Online
192ms avg
Storage
● Online
8ms avg
Queue
● Online
5ms avg
541ms
avg pipeline time
$0.00
per email processed
3.8GB
RAM used of 8GB
End of demo
Lexara processed one email in 541ms, found two deadlines and a sanctions threat, matched it to a case, queued three actions, and connected it to a broken commitment from 3 weeks ago — all locally, all free, all waiting for Lauren to decide what happens next.
Wesley Thompson · ChopMedia LLC · (385) 461-4164
Restart to walk through again